Opus Growth

Campaign setup

Ready-made prompts for Claude, ChatGPT and other assistants connected to your ad accounts. Copy one, fill in the brackets and send.

Campaign setup

Build a local Google Search campaign from scratch, created paused

A fully structured Search campaign with keywords, ads and negatives, shown for approval before it is created paused.

Set up a new Google Search campaign for my [business type] in [city]. It must be created paused.

1) Run Keyword Planner with 5-10 seed terms for my main services. Keep ideas with at least 50 average monthly searches. Drop job, DIY, free and training queries.
2) Group the remaining ideas into 2-4 tightly themed ad groups, one service per group. Use phrase and exact match only, with no broad match at launch.
3) Find the location ID for [city]. Set the location option to people present in the location, not people interested in it.
4) Write one responsive search ad per ad group with at least 12 headlines and 4 descriptions. Include the city and a clear call to action. The final URL is [landing page URL].
5) Add campaign-level negative keywords: jobs, salary, free, how to, course, used.
6) Set the daily budget to [daily budget] with Maximize Clicks to start.

Before you create anything, show me the full structure as a table. Include ad group, keyword, match type, monthly searches, headlines with character counts, negatives, budget and location. Wait for my approval, then create the campaign paused.

Campaign setup

Demand Gen campaign using my best-retaining YouTube videos

Shortlists your strongest short YouTube videos by retention and builds a paused Demand Gen campaign around them. It adds country and language targeting, a 30-day remarketing audience and ready-to-run ad copy, and previews every step before anything is created.

I want a Demand Gen campaign that uses my own YouTube videos. It must be created paused.

Inputs:
- Landing page: [product or service page URL]
- Country: [country]
- Language: [language]
- Daily budget: [daily budget]
- Business name (max 25 characters): [business name]

1) Video shortlist.
   - List my YouTube channels. If there is more than one, ask me which to use.
   - Pull the channel overview for the last 90 days to get the channel's average view percentage as a benchmark.
   - Pull my top 25 videos for the same period.
   - Rank them by average view duration in seconds, not by views. Show each video's title, views, watch time and average view duration.
   - The report does not include video length or per-video view percentage. Take the top 6 by average view duration and ask me to paste each one's length, or the "average percentage viewed" from YouTube Studio. Then calculate view percentage as average view duration divided by length.
   - Shortlist 2-3 videos that are under 60 seconds and above 40% average view percentage. If none qualify, say so and stop.
   - For each shortlisted video, ask me for its YouTube link or video ID if you do not already have it.
2) Audience check. List my audiences with their definitions. Look for a remarketing list of visitors to [product or service page URL] with a 30-day membership.
   - If one exists, use it.
   - If not, propose one: name, URL rule (CONTAINS [product or service page URL path]), membership 30 days, lookback 30 days.
3) Campaign plan:
   - Campaign: Demand Gen, daily budget [daily budget], bid strategy Maximize Conversions unless I say otherwise.
   - One ad group, named after the product or service.
   - Ad group location: [country]. Ad group language: [language]. Demand Gen keeps location and language on the ad group. Without this step the campaign targets every country, so it must be applied before anything is enabled.
   - Remarketing audience on the ad group. A newly created list starts empty and fills over the following days. Recommend whether to attach it as Observation or Targeting given its current size, and explain why.
4) Ad copy. For each shortlisted video, write:
   - 3 headlines (max 40 characters each)
   - 1 long headline (max 90)
   - 2 descriptions (max 90)
   - A call to action
   Show character counts. The final URL for every ad is [product or service page URL] and the business name is [business name].

Preview everything first: the shortlist with metrics, the audience rule, the campaign and ad group settings, the targeting and all ad copy. Wait for my approval. Then create things in this order, showing a dry run (confirm=false) for each step and waiting for my OK before confirming:
a) The remarketing audience, if a new one is needed.
b) The Demand Gen campaign (paused).
c) The ad group. First check whether the campaign already has one by listing its ad groups. If it does not, run the ad group creation as a dry run. If the tool refuses for a Demand Gen campaign, tell me to add one ad group in the Google Ads interface and paste its ID, then continue.
d) Ad group location and language targeting. After applying it, read the ad group's targeting back and confirm only [country] and [language] are set.
e) Attach the audience to the ad group in the agreed mode.
f) One video ad per shortlisted video.

Finish with a summary of the IDs created and confirm that the campaign is still paused.

Campaign setup

Final pre-launch review of my paused campaigns before turning them on

Audits every recently created paused campaign for targeting, ad strength, policy, extensions, schedule, budget and landing page, then proposes enabling only the ones with zero fails.

Review every campaign I created in the last 14 days that is still paused, before I turn any of them on. Use the account snapshot including inactive campaigns to find them by creation date, and list their ad groups and ads.

My intended daily budgets: [campaign name: daily budget, one per line]
My opening hours: [opening hours]

For each campaign, check the following:

1) Targeting. The location option should be people present in the location, not people interested in it. Confirm the language is correct and the device settings make sense. For Demand Gen campaigns, read location and language from each ad group, because Google keeps them there.
2) Ads. Every ad group should have at least one ad with Ad Strength of Good or better. List Google's improvement items for the rest. For Performance Max, check the asset group ad strength instead. Also flag ads that are themselves paused: a paused ad stays paused when the campaign is enabled and serves nothing.
3) Policy. Check for any disapproved or limited ads, and give the reason and the triggering text Google reports.
4) Extensions. Search campaigns should have at least 4 sitelinks and 4 callouts. For other campaign types, mark this check as not applicable rather than failed.
5) Schedule and budget. Check that the ad schedule matches my opening hours, and that the daily budget matches the number I gave above. If I gave no number for a campaign, show the current budget and mark it "to confirm".
6) Landing page. Take the final URL from the ads. Run a mobile SEO and tag audit and check mobile Core Web Vitals. Fail the page if mobile LCP is over 2.5 seconds (use real-user field data when available, otherwise lab data, and say which) or if no Google tag (Google Ads or GA4 / GTM) is found among the loaded measurement tags.

Deliver one table per campaign with these columns: check, pass/fail/not applicable, detail, fix.

Decision rule: recommend enabling only the campaigns with zero fails. Show that list as a dry run (confirm=false), including any paused ads inside them that would also need enabling, and wait for my approval before changing any status.

Campaign setup

Launch a Meta lead generation campaign with an instant form

Reuses or drafts a qualifying instant form, then builds a paused Leads campaign, ad set and form-linked ads, with dry-run previews before creation and rendered Feed and Stories previews after.

Set up a lead generation campaign in Meta ad account [client ad account] for [offer], targeting [city or country].

1. List the Facebook Pages this ad account is actually allowed to advertise for, and the instant forms already on the correct Page with the questions each asks. Reuse a form only if it collects full name, phone and email and includes at least one qualifying question. Otherwise, draft a new form with full name, phone, email, two qualifying questions as choice lists (budget range: [budget options]; timeline: [timeline options]) and the privacy policy URL [privacy URL].
2. Draft one campaign with the Leads objective, created paused. Ask me whether [offer] falls under housing, employment, credit or another special ad category before building. Include one ad set with broad targeting in the location above, ages 25-65, automatic placements (no placement restrictions), lead generation optimisation with the Page as the promoted object. Ask me for the daily budget before building.
3. Draft link ads attached to the form, using the image [image link or attached file]. Write 3 primary text options under 125 characters, a headline under 40 characters, and a Sign Up call to action. Use a clear, plain style with no hype. Ask me whether to build one ad per primary text (3 ads, so Meta can compare them) or one ad with the option I pick.

Before you create anything, show me the full plan as a table with the form fields, campaign, ad set and ad copy, plus the dry-run (confirm=false) output of each step. Wait for my approval, then create everything paused in this order: form, campaign, ad set, ads. Then render ad previews for Feed and Stories (Facebook and Instagram) from the new creatives so I can check them while nothing is spending. Return all IDs.

Campaign setup

Set up a Meta dynamic product retargeting campaign from the catalog

Checks pixel, catalog, product set and Page, builds only the missing retargeting audiences, then drafts a paused Sales campaign with a catalog ad set and a dynamic ad, with dry-run previews and approval before anything is created.

Set up a dynamic product retargeting campaign for e-commerce client [client ad account].

Pre-checks: stop and report if any of these fail.
- The pixel has fired recently. Read its last fired time and stop if it is older than 24 hours or empty. If the value is only precise to the day, say so and tell me what it shows.
- The catalog appears among the catalogs on the ad account's Business Manager and its product count is above zero.
- Product set [product set name] exists in that catalog. If it is missing, list the available product sets. If the listing shows a product count, confirm it is not empty; if it does not, say so.
- The Facebook Page this ad account is actually allowed to advertise for (check with the ad account, not just my personal Page list).

Audiences: check the existing custom audiences first, then propose only what is missing. Each website audience is built from one pixel event, so the retargeting pool is two audiences used together in the ad set:
- Viewed a product (ViewContent), 14-day retention
- Added to cart (AddToCart), 14-day retention
- Purchasers (Purchase), 30-day retention, as the exclusion
If the audience list returns sizes, warn me when the 14-day audiences look too small to deliver.

Draft the following:
- One campaign with the Sales objective, created paused, with no campaign budget so the ad set holds it.
- One catalog ad set bound to the catalog and product set, targeting both 14-day audiences, excluding purchasers, with automatic placements (no placement restrictions), conversion optimisation on the pixel's Purchase event and a daily budget of [daily budget], created paused.
- One dynamic ad on the Page with a Shop Now button and the fallback link [store homepage], created paused. The catalog fills the product name; put the price in the description with {{product.price}}. Write a primary text that reminds the viewer of the product they looked at, with no discount claims unless I add them.

Step 1: show the full plan with the confirm=false output for every audience you will create and for the campaign. Wait for my approval.
Step 2: after I approve, create the audiences and the paused campaign. Then show the confirm=false output of the ad set and the ad (they need the new IDs). Wait for my approval again.
Step 3: once approved, create them paused. Run a delivery estimate on the new ad set and flag a small audience. Take the creative id from the create response (or read it from the new ad's creative) and render previews for Facebook Feed, Instagram Feed, Facebook Stories and Instagram Stories, so I can review them before anything spends. Return all IDs.

Campaign setup

Turn Instagram Reels into WhatsApp click-to-message ads

Resolves the ads-usable media ID for each Reel, checks that the target ad set really sends people to WhatsApp, and builds paused ads that keep the original post's engagement. You approve the drafts before anything is created.

Turn these Instagram Reels into WhatsApp click-to-message ads in our client's Meta ad account [client ad account]: [Reel links, one per line]. The client's Instagram account is [client Instagram username].

1. Access: Confirm that [client Instagram username] appears among the Instagram accounts usable from this ad account. Then list the Pages for this ad account and tell me which Facebook Page is actually assigned to it. A Page that is only visible to me does not count. If the Instagram account or an assigned Page is missing, stop and tell me what has to be fixed.

2. Media IDs: Try to resolve each Reel link to its ads-usable media ID. A link alone does not always return that ID. When it does not, list the client's Instagram posts and Reels and match each link by its shortcode or permalink, then take the media ID from that list. Never use the internal media_pk as the ad ID. If a Reel is not in the list, or its owner is not [client Instagram username], mark it and ask me to paste its media ID from Ads Manager (Use existing post > Instagram).

3. Ad set check: Read the settings of ad set [ad set name]: optimisation goal, destination type, placements, Page and status. Also list the ads already in it and report any delivery issues, especially error 2490279 (button does not match the ad set's messaging destination). Tell me whether the ad set suits a WhatsApp ad. That means a WhatsApp messaging destination, ideally with conversations optimisation, and placements that include Instagram Reels. If it does not suit, propose the fix but do not apply it. An existing ad set's messaging destination and optimisation goal cannot always be changed after creation, so the fix may have to be a new ad set. In that case, give me its full spec: name, CONVERSATIONS optimisation, WhatsApp destination, assigned Page, daily budget [daily budget], targeting copied from [ad set name], and placements. Create it only after you show me a preview and I approve it, and create it paused. Do not put WhatsApp ads into an ad set whose destination is not WhatsApp.

4. Drafts: For each Reel, draft an ad built from the existing Instagram post so that its likes, comments and views stay on the ad. Use the WhatsApp message button, this pre-filled message: [pre-filled message], and status PAUSED. Run every draft as a preview first and create nothing yet.

Show me a table with these columns: Reel link, media ID, how the ID was found (from the link, from the posts list, or pasted by me), views and likes if returned, ad name (format CLIENT_REEL_DATE), button, pre-filled message. Wait for my approval before you create the ads.

After creation, do these checks while the ads are still paused:
- List the ads in the ad set. Confirm that each new ad is paused and has no delivery issues, and report its ID.
- Read each ad's creative. Confirm that the WhatsApp button and the pre-filled message are there and that the ad is still linked to the original Instagram post.
- If a creative ID is available, render placement previews in INSTAGRAM_REELS and INSTAGRAM_STORY formats and give me the links. Previews can only be rendered once the ads exist, which is why this step comes after creation.

Campaign setup

Pre-launch readiness check for a new client's Meta account

Before the first campaign, checks Page and Instagram access, pixel activity, the GA4 property, custom conversions against GA4 key events, Facebook and Instagram traffic tagging, thank-you pages, audiences and instant forms. Then proposes the missing custom conversions, which are created only after your approval.

We just onboarded a new client. Before we build any campaign in Meta ad account [client ad account], run a readiness check. The client's website is [client website].

Check each item and mark it pass, fail or needs manual check. Only use what the tools return. When a field you need is not returned, mark the item needs manual check and say what is missing.

- Access: The ad account is listed and reachable. List the Pages for this ad account and confirm which Facebook Page is actually assigned to it, not just visible to me. Then list the Instagram accounts usable from this ad account. If no Page is assigned, mark it fail.
- Pixel: At least one pixel exists and its last-fired time is within the last 24 hours. The tools only show the last-fired timestamp, not individual events. So even if this passes, recommend an event-level check in Events Manager (for example, a test purchase or a test lead).
- GA4 property: List the GA4 properties and their web data streams. Find the one whose stream URL matches [client website]. If no stream matches the client's domain, flag it and skip the GA4 items below, because they would describe the wrong site.
- Custom conversions vs GA4: List the Meta custom conversions. Flag any that have never fired, since that usually means the rule does not match what the pixel sends. Check that URL rules point to pages on [client website]. List the GA4 key events, then pull a 28-day GA4 report by event to see how often each key event actually occurred. List every GA4 key event with real volume that has no Meta custom conversion equivalent.
- Facebook and Instagram traffic in GA4: Pull a 28-day source/medium report. Find the rows whose source looks like Facebook or Instagram (facebook, fb, ig, instagram, l.facebook.com, m.facebook.com, lm.facebook.com and similar), and report their sessions and key events. If there are none, or they only show up as referral traffic, UTM tagging is probably missing. This is a judgement based on source names, so list the rows you matched.
- Thank-you pages: From 28-day GA4 landing page and page reports, list the pages that look like confirmation pages (thank you, order received, success, tesekkurler) with their sessions. These are the candidates for URL-based custom conversions.
- Audiences: List the custom audiences. Judge from the fields actually returned (name, type, size, and any date fields) which ones look too small, outdated or duplicated. If size or dates are not returned, mark the item needs manual check rather than guessing. List the saved audiences we can reuse, with their approximate size.
- Instant forms: List the instant forms on the assigned Page with their status and the questions each one asks. If no Page is assigned, mark this needs manual check.

Deliver the checklist as a table with these columns: item, status, evidence, fix.

Then, for each missing custom conversion, propose a definition: name, pixel, URL contains rule and/or standard event, category, and a default value if the client gave one ([conversion values, if any]). Show me a preview of every custom conversion and create nothing until I approve. The default value cannot be set at creation, so after the conversions are created, show me a second preview that adds the default values and wait for my approval again before you apply it.

Campaign setup

Rebuild your best Google Search campaign in Microsoft Ads

Copies only the ad groups that earn their keep from a Google Search campaign into a paused Microsoft Ads campaign. Keywords, RSA copy, negatives and locations are mapped to Microsoft's rules, and the build runs in three approval rounds so every new ID exists before the next step needs it.

Build a Microsoft Ads version of my Google Search campaign [Google campaign name] in Microsoft account [Microsoft account].

1. Find the campaign ID in Google Ads, then pull the last 90 days for this campaign:
   - Ad groups with cost, clicks, conversions and conversion value. If the standard report does not return conversion value, get it with a read-only query per ad group. ROAS = conversion value / cost.
   - Keywords with match type, cost, clicks, conversions and average CPC. If average CPC is not returned, calculate it as cost / clicks.
   - The RSA headlines and descriptions of every enabled ad in each ad group.
   - Every negative keyword that applies to the campaign: ad group level, campaign level and the shared negative lists attached to it, with the keywords inside each list.
   - The campaign's location, radius and language targeting.
2. Carry over only ad groups that meet at least one condition: ROAS at or above [target ROAS], or at least 3 conversions. If the campaign records no conversion value, use the conversion rule alone and say so. List the excluded ad groups with the reason for each.
3. Check that [Microsoft account] is connected and that no campaign named [Microsoft campaign name] already exists there. Then show the planned build in one table with these columns:
   - ad group, with its Microsoft language set to [ad language]
   - keywords grouped by match type (Broad, Phrase, Exact), original match types kept
   - proposed default max CPC, set at 70% of that ad group's Google average CPC
   - RSA copy: 3 to 15 headlines (max 30 characters) and 2 to 4 descriptions (max 90 characters), checked against these limits, with the final URL and paths. Shorten anything over the limit and mark it. Note that pins do not carry over.
   - negatives, at campaign or ad group level as they were in Google. Microsoft negatives only accept Exact and Phrase, so map Google broad negatives to Phrase and flag each one you mapped. Shared list negatives go to campaign level.
   - locations, as Microsoft location IDs found by name search. If the Google campaign uses radius targeting, say it cannot be copied and propose the closest cities instead. If a name returns several matches, pick the one in the right country and show it.
4. Use a daily budget of [daily budget] for a Search campaign named [Microsoft campaign name].
5. After I approve the table, build it in three rounds, because ad groups need the new campaign ID and keywords, ads and ad group negatives need the new ad group IDs. In each round run every write with confirm=false first, show me the previews together, wait for my approval, then apply with confirm=true and report the new IDs:
   - Round 1: the Search campaign (created paused) and its location targets.
   - Round 2: the ad groups under the new campaign, with language and default max CPC. They are created paused.
   - Round 3: keywords (one call per ad group and match type), RSAs, campaign-level negatives and ad-group-level negatives using the new ad group IDs.
   If any preview differs from the approved table, for example a location that was not found or a headline that fails validation, stop and tell me before applying anything in that round.

Everything stays paused. Do not activate the campaign, the ad groups or the ads.

Campaign setup

Pre-launch QA for new TikTok, LinkedIn and Microsoft builds

Runs a launch checklist on paused or draft builds across three platforms: tracking, optimisation, budget, structure, destinations and leftovers, with an honest pass, fail or cannot verify per check and a go or no-go verdict per platform.

I built new campaigns on three platforms and they are all paused or in draft. Before anything goes live, run a launch QA on these campaigns:
- TikTok: [TikTok campaign]
- LinkedIn: [LinkedIn campaign] (this is the campaign as Campaign Manager shows it, so treat it as a campaign group and check every ad set inside it)
- Microsoft: [Microsoft campaign]

If the TikTok campaign does not appear in list_tiktok_campaigns, check list_tiktok_smart_plus before calling it missing.

Check each campaign for:
1. Conversion tracking:
   - TikTok: the pixel set on each ad group exists in list_tiktok_pixels, has the optimisation event configured and shows activity_status ACTIVE.
   - LinkedIn: a conversion rule from list_linkedin_conversions is enabled and lists every ad set in its associated campaigns. A MAX_CONVERSION ad set without one is a fail.
   - Microsoft: a UET tag and a conversion goal exist. Report the tag or goal status only if the data includes it.
   None of these tools show recent event counts, so mark "recorded events recently" as cannot verify unless the returned data proves it, and tell me where to confirm it (TikTok Events Manager, LinkedIn Campaign Manager, Microsoft UET Tag Helper).
2. Optimisation: the TikTok ad groups, LinkedIn ad sets and Microsoft bid strategy optimise for purchases or [agreed lead event], not clicks or reach. If a list does not return the optimisation goal or bid strategy, mark it cannot verify.
3. Budget: daily budgets at the level each platform sets them add up to no more than [total daily budget]. If the accounts use different currencies, show each in its own currency and ask me for the rate instead of guessing.
4. Structure: every TikTok ad group, LinkedIn ad set and Microsoft ad group has at least 2 ads. The Microsoft campaign has at least one sitelink and one callout attached.
5. Destinations: compare the landing page or final URL returned for each ad, plus any UTM parameters in it, across platforms against [expected landing page and UTM pattern]. Where an ad's URL is not returned, or where tracking is set in a template these tools cannot read, mark it cannot verify and ask me to paste the URLs.
6. Leftovers: list every campaign, ad group or ad set and ad in the three accounts that is draft, has a test-like name or is active but not part of this build, and flag anything that could spend money.

Deliver one table with these columns: platform, check, status (pass, fail or cannot verify), evidence, fix. End with a go or no-go verdict per platform.

Do not fix or activate anything yourself. For each platform marked "go", list exactly what has to be switched on (campaign, ad groups or ad sets and ads, since new ad groups and ads are often created paused on their own), run those status changes with confirm=false and show me the previews. Apply nothing until I approve.

Campaign setup

Pre-launch readiness check and go-live for a ChatGPT Ads campaign

Checks a paused ChatGPT Ads campaign, its ad groups and ads against a launch checklist, then gives a go or no-go verdict. Activates only after your approval.

My ChatGPT Ads campaign [campaign name or cmpn_ id] is built but not live. My service area is [service area]. Run a pre-launch check before I activate it.

1. Confirm which ad account you are using with manage_chatgpt_ads_accounts (list), and run a check on it if any call returns an authorization error.
2. If I gave a name instead of an id, find the campaign id with chatgpt_ads_insights at campaign level with include_zero=true. Then list every ad group and ad under it with chatgpt_ads_insights, scope_id set to the campaign, include_zero=true, at ad_group level and again at ad level, so entities with no impressions are included.
3. Read the full details of the campaign, every ad group and every ad with chatgpt_ads_details. Verify each of the following:
   - the budget is set (daily or lifetime)
   - locations are set and limited to my service area. Translate any location ids into names with chatgpt_ads_geo_lookup so I can read them.
   - the schedule (start and end) is set
   - each ad group has a max bid and at least 5 context hints, written in the language my buyers use. Page through the hints if needed.
   - each ad has a title, body, image and destination URL, and its review status is not rejected
   - a lead conversion event (lead_created or my equivalent) is attached to the campaign. Cross-check with chatgpt_ads_conversion_setup, including whether the pixel received any recent events.
   - account spending limits will not block delivery (chatgpt_ads_spend_limits). If the tool says the key lacks admin billing permission or the account is not on invoice billing, mark this item as a warning and say why.
4. For every unique destination URL, load it with site_read_url and confirm it returns a 200 status without a redirect chain.
5. Render each ad with chatgpt_ads_ad_preview and list any serving issues and review status OpenAI reports. Then run diagnose_chatgpt_ads_delivery for this campaign with check_landing_pages=true.

Output a checklist table with these columns: item, status (OK, warning, blocker), evidence, and the fix. Give a final go or no-go verdict:
- Any blocker means no-go.
- Warnings are acceptable only if you explain the risk.

If the verdict is go, prepare the activation with chatgpt_ads_status using confirm=false for the campaign, each ad group and each ad that is still paused, and show me that dry run. Do not run anything with confirm=true until I explicitly approve it.

Campaign setup

Check the landing page before sending ChatGPT Ads traffic to it

Audits every destination page in a ChatGPT Ads campaign for status, call and form presence, ad message match, speed and tracking, compares GA4 results with your best page, and proposes URL changes for approval.

Before I scale my ChatGPT Ads campaign [campaign name or cmpn_ id], check that its destination pages turn visitors into calls and quote requests.

1. List every ad in the campaign with chatgpt_ads_insights (scope_id = the campaign, level = ad, include_zero=true), then read each ad's title, body and destination URL with chatgpt_ads_details.
2. For each unique URL:
   - load the live page with site_read_url and confirm it returns a 200 status
   - read the page outline and text, and check that a tel: phone link and a quote form exist and appear early in the page, before the first long content section. If the page is hosted on my Opus Growth server, also take a mobile screenshot with site_preview so we can see what is actually above the fold. If it is not, tell me this check is based on page order, not on a rendered view.
   - check that the city and the service named in the ad also appear in the page title, H1 or opening text
   - run core_web_vitals on mobile and seo_audit on mobile, and report which measurement tags (GA4, Google Ads, others) the audit finds loaded on the page
3. Pull GA4 data for the last 30 days with ga4_report, dimension = landing_page and dimension2 = source_medium: sessions and key events. Calculate the key event rate (key events per session) for each page and compare it with my best-converting service page [alternative URL]. If ChatGPT Ads traffic can be identified in the source/medium (for example chatgpt.com or my UTM tag [utm_source value]), report that slice separately. If it cannot, say so and compare the pages on all traffic.

Give a verdict per URL: keep, fix, or replace. Use this rule:
- Replace if mobile LCP is above 4 seconds, if there is no call link, or if the key event rate is less than half that of the alternative page.

List the fixes in priority order. If a replacement is warranted, prepare the destination change with update_chatgpt_ads_ad using confirm=false and show me the current and proposed values. Remind me that changing the destination sends the ad back to OpenAI review and it stops serving until the review passes. Do not apply any change until I approve.

Campaign setup

Build a better-qualified Meta instant form from your existing ones

Reviews the instant forms on a Facebook Page, measures 30-day lead volume and answer quality per form, then drafts a new form with qualifying questions and creates it only after your approval.

Our Meta lead ads bring volume, but sales says too many leads are unqualified. Work on Facebook Page [page name] in ad account [ad account name or id].

1. Find the Page id with list_meta_pages, passing the ad account so you can confirm the Page is assigned to it. Then list the Page's instant forms with list_meta_lead_forms and keep the active ones. For each active form, show its questions.
2. For each active form, pull leads with download_meta_leads using since = the date 30 days ago, and count them to get the 30-day lead count per form. If a pull hits the limit, raise it or say the count is a minimum.
3. Take the form with the most leads and analyse its leads:
   - How answers are distributed on any existing custom questions.
   - Your estimate of the share of incomplete or obviously fake entries (empty or placeholder fields, malformed phone numbers or emails, repeated junk text, the same contact submitted more than once). Explain the rules you used, since this is a judgement call. Do not paste personal data, only aggregates.
4. Draft a new form with these elements:
   - Contact fields: FULL_NAME, PHONE and EMAIL only.
   - Two multiple-choice qualifying questions: [question 1] and [question 2], written as 'Question?|Option 1|Option 2|...'. Use answer options that clearly separate a fit from a non-fit.
   - intro_headline and intro_description that state who the offer is for.
   - privacy_policy_url: [privacy policy URL].
   - thank_you_text that tells people when they will be contacted, and follow_up_url: [website URL].
   - locale matching the language of my audience: [language].
   - Consider higher_intent=true (a review step before submitting) and tell me whether you recommend it for this case.
5. Show the draft side by side with the current top form and explain the expected effect: fewer leads, but a higher share of qualified leads.

Then call create_meta_lead_form with confirm=false and show me the preview. Do not create the form until I approve it. Once it is created with confirm=true, return the form id so I can attach it to the lead ad.

Campaign setup

Audit extension coverage across Search campaigns and draft the gaps

Maps which extension types every enabled Search campaign has, measures their clicks and CTR, then drafts the missing sitelinks, callouts and snippets from your own website for approval.

Review the extension assets on every enabled Search campaign in my Google Ads account [account name].

1. Find the account, then list the enabled Search campaigns.
2. For each campaign, list which of these are linked: sitelinks, callouts, structured snippets, call, location, image, business name and logo.
   - Use list_ad_extensions and list_extension_assets for sitelinks, callouts, snippets and call.
   - Use google_ads_query on campaign_asset and customer_asset to check image, business name and logo, and on customer_asset_set and campaign_asset_set to check location. Assets linked at account level count for every campaign, so mark them as "account level".
3. Pull extension performance for the last 30 days with extension_asset_performance.
4. Deliver a table with these columns: campaign, missing types, number of linked sitelinks, and clicks and CTR per extension type.
5. Flag any campaign with fewer than 4 sitelinks, no callouts or no structured snippet.

For each gap, read my website [website URL] with site_read_url. Use outline=true to see page titles and headings, and read the homepage body to find the real page URLs in the navigation. Then draft the missing assets:
- Sitelinks: enough to reach 4 per campaign. Link text up to 25 characters, two description lines up to 35 characters each, and each one pointing to a real page that matches the campaign. Do not repeat a sitelink the campaign already has.
- Callouts: 4 per campaign, up to 25 characters each.
- Structured snippet: one, with a header from Google's predefined list (such as Services, Types, Brands, Destinations) and at least 3 values.

Use only claims you can find on the website: delivery times, guarantees, service areas. Avoid generic phrases like "best quality". If the site does not support a claim, leave it out.

I will add any missing image, location, business name or logo myself, so just list those gaps.

Show the full draft as a preview with manage_extension_assets, grouped by campaign. Do not add anything until I approve each campaign separately.

Campaign setup

Find weak or broken sitelinks and rewrite them in place

Checks 90 days of sitelink performance, opens every final URL to catch broken or mismatched pages, flags low CTR links and edits them in place so their history is kept.

Check the sitelinks on my Google Ads campaigns for the last 90 days.

1. List every sitelink with its campaign, asset id, link text, description lines, final URL, impressions, clicks, CTR and conversions.
   - Get the sitelinks and their URLs with list_extension_assets.
   - Get the numbers with extension_asset_performance, field_type SITELINK and days=90. Run it per campaign with campaign_id, so each campaign has its own sitelink averages.
2. Open each final URL with site_read_url and outline=true. Flag a sitelink if:
   - the page returns an error status (4xx or 5xx),
   - the page ends up on the homepage (check the status, the redirect headers and the canonical),
   - the title and headings do not match the link text. Give a one-line reason for each mismatch.
3. Flag any sitelink that has at least 500 impressions and a CTR below half of the average sitelink CTR in its campaign.

Deliver one table sorted by impressions. Add a "problem" column with one of these values: broken, mismatched page, low CTR, or OK.

For every flagged sitelink, propose a fix:
- New link text up to 25 characters.
- Two description lines up to 35 characters each. Google shows them only as a pair, so write both.
- A correct URL from the same site. Find it by reading the homepage body for the real navigation links, then confirm the page works with site_read_url.

Edit the existing sitelinks with update_sitelink rather than deleting them, so their history is kept. If a sitelink is linked to more than one campaign, tell me, because the edit will change it everywhere.

Show me the old and new versions side by side as a preview. Wait for my approval before changing any of them.

Campaign setup

Pre-launch checklist for a paused Google Search campaign

Runs eight launch checks on a paused Search campaign, from targeting and budget to ad strength, extensions, negatives and live landing pages, then previews each fix before enabling.

Run a pre-launch checklist on the paused Search campaign [campaign name] in [client account]. Do not change anything yet. The client's business hours are [business hours].

Return a table with the columns Check | Pass/Fail | Evidence:
1. Locations and language are correct. The location option is set to Presence, not Presence or interest.
2. The ad schedule matches the business hours and has no leftover bid modifiers.
3. The daily budget is at least 2x [target CPA].
4. Bid strategy: if it is Maximize conversions or Target CPA, the primary conversion actions recorded conversions in the last 30 days.
5. Every enabled ad group has at least one enabled RSA with Ad Strength Good or Excellent, and no ad is disapproved or limited.
6. At least 4 sitelinks and 2 callouts are linked.
7. Campaign-level negatives or a shared negative keyword list are attached to this campaign.
8. Every final URL used by the ads returns 200. If a URL returns a 3xx, read the Location target and follow it until you reach the final page, and fail the check if there is more than one hop or the end page is not 200.

Below the table, split the failures into Blockers (must fix before launch) and Improvements. For each fix you can make here, show me a preview and wait for my approval. Fixes to the landing pages themselves are for me to handle. Enabling the campaign is the last step, and only after I confirm.

Campaign setup

End-to-end conversion tracking check before launch

Verifies the conversion chain across Google Ads, Tag Manager, GA4 and the live website, flags duplicate tags and unpublished changes, and gives a single Go or No-go verdict.

Before we launch new campaigns for [client account], verify the conversion chain end to end for [website URL]. This is read-only.

Check each layer:
1. Google Ads: list the conversion actions with status, primary/secondary, counting (One for leads, Every for purchases), lookback window and default value. Flag any primary action with zero conversions in the last 14 days.
2. Tag Manager: find the container this site uses. Confirm which version is live, and list any unpublished workspace changes. If there are none, the workspace tags match the live version; if there are, say which tags may differ from what is live. Then confirm the AW conversion ID and label of each primary action match a Google Ads conversion tag, and that its trigger fires on the real success event or thank-you page, not on every page.
3. GA4: list the key events and confirm the property is linked to this Google Ads account.
4. Website: if the site is connected as a WordPress site, list every place GTM or GA4 fires. Otherwise read the live homepage and [thank-you page URL] and list every GTM container ID and gtag ID in the HTML. Flag duplicates, such as GTM loaded twice, two different containers, or a hard-coded gtag next to GTM, and confirm the container on the site is the one checked in step 2.

Output a table with the columns Layer | Check | Status | Evidence, then a single verdict: Go, or No-go with the blocking items. If something needs fixing, describe the fix and show a preview. Do not publish a GTM container or edit the site without my explicit approval.

Campaign setup

Meta campaign pre-flight: pixel, delivery estimate and creative review

Checks a paused Meta campaign before launch: pixel and custom conversion health, delivery estimates against the learning threshold, audience size, creatives, placement previews, delivery issues and landing pages, with fixes previewed for approval.

Review the paused Meta campaign [campaign name] in [Meta ad account] before launch. The client's goal is [goal, e.g. purchases, leads, booked calls]. Do not change anything without my approval.

Account level first:
- Note whether the campaign has its own daily budget (CBO) or the budgets sit on the ad sets, because the learning check below depends on it.
- List the pixels with their last fired time. Flag any pixel the ad sets rely on that has not fired in the last 24 hours or has never fired. Last fired time is the only freshness signal here, so if it looks ambiguous, say so rather than guessing.
- List the custom conversions. Flag any that the ad sets use and that appear under never_fired, because the rule probably does not match what the pixel sends.

For each ad set:
- Check that the optimization goal and conversion event match the client's goal, and that the pixel and any custom conversion it relies on exist and are live. If the ad set data does not show which pixel or event it uses, say so.
- Run the delivery estimate. Multiply the expected daily results by 7 and flag the ad set if that is under about 50 optimization events per week at the current budget, because it will likely stay in learning. If the estimate gives no results figure for this goal, say so and judge from audience size and budget instead.
- For prospecting ad sets (no custom audience attached, if the ad set data shows audiences), flag an estimated audience under 500,000. If the delivery estimate returns no audience size, use the reach estimate with the same countries, ages and interests.

For each ad:
- Read the creative: primary text, headline, CTA, destination URL, Page and Instagram account. Flag a destination URL with no UTM parameters.
- List which Advantage+ creative enhancements are switched on, so I can confirm them with the client.
- Using the creative id from the creative read, generate preview links for Feed (MOBILE_FEED_STANDARD), Stories (INSTAGRAM_STORY) and Reels (INSTAGRAM_REELS). If no creative id comes back, or a format does not render for that creative, note it in the table.
- Report any delivery issues Meta lists on the ad (issues_info), with the error code and message.
- Read the destination URL live: it should return 200. If the response shows a redirect or a different final URL, note it.

Output two tables (ad sets, ads) with the columns Item | Status | Issue | Suggested fix. Then list the blockers.

For any fix to an ad set or a creative, show me the confirm=false preview first and wait for my approval. If the creative read shows an ad was built from an existing Instagram post, change its message or button only through the Instagram post ad update so it stays linked to the post. Keep the campaign paused.

Campaign setup

LinkedIn B2B campaign pre-launch: audience, bids, conversions, copy

A campaign-by-campaign check of audience size, bid versus LinkedIn's suggestion, conversion setup, lead forms and ad copy before launch.

Before launch, audit every campaign in the campaign group [campaign group name] in [LinkedIn ad account].

For each campaign:
1. Audience: get the forecast. Flag the campaign if the audience is under 50,000 members, which is too narrow for Sponsored Content, or over 1,000,000, which is too broad for a B2B offer.
2. Bid and budget: compare the bid with LinkedIn's suggested bid and range. Flag bids below the floor or under 80% of the suggestion. Flag any daily budget below the minimum.
3. Conversions: confirm that the right conversion rule is attached and that the rule exists on the account. For lead gen campaigns, confirm that the lead form is active.
4. Copy: read each ad. Flag intro text over 150 characters, because it gets truncated in the feed. Also flag a missing CTA and a headline that does not name the offer.

Campaign setup

Go/no-go check before enabling a paused campaign

A pass/fail launch checklist for a paused Search, PMax, Demand Gen or App campaign, with fixes, and activation only after your approval.

Before I enable the paused campaign [campaign name] in [account], run a launch check and give me one table with these columns: check, status (pass / fail / warning), evidence, fix.

Checks:
1. Conversions: Purchase is a primary goal and has recorded data in the last 14 days. It uses transaction-specific values rather than one default value, is counted as "every", and no secondary or micro conversions (add to cart, page views) are primary.
2. Bidding fits the data: a ROAS target only if the campaign or account has at least 30 purchase conversions in the last 30 days.
3. Targeting: correct countries and language, and location set to presence rather than presence or interest.
4. Creative: for Search, Ad Strength is at least Good. For PMax, every asset group has all image ratios, a logo and at least 5 headlines. For Demand Gen and App, the ads exist.
5. Policy: no disapproved or limited ads.

End with a verdict: GO, GO after fixes, or NO-GO. If any fix needs a change, preview it and wait for my approval. Enable the campaign only after I explicitly confirm.

Campaign setup

Launch a Meta lead campaign with a qualifying instant form

Builds a paused Meta Leads campaign for a local service: a qualifying instant form, city or radius targeting, two ads made from your own Page post photos, preview links and a delivery check before anything goes live.

Set up a Meta lead generation campaign for [service, e.g. boiler repair] in [city], with a daily budget of [daily budget].

1. Account and Page: find my ad account with list_meta_ad_accounts. Then list my Pages, passing that account_id, and use a Page that is actually assigned to the ad account. List that Page's existing instant forms and the questions each one asks.

2. Form: if no existing form fits, draft a new instant form with these questions:
- FULL_NAME and PHONE (standard, pre-filled fields)
- "Which district are you in?" as a free-text custom question (Meta has no standard district field)
- "Which service do you need?|[option 1]|[option 2]|[option 3]"
- "How soon do you need it?|Today|This week|Just researching"
Use [privacy policy URL], form language [form language], a short intro headline and a thank-you text.

3. Campaign: OUTCOME_LEADS objective, status PAUSED, with no campaign-level budget so the budget sits on the ad set. Leave special ad categories empty unless my service falls under housing, employment, credit or finance. If it does, tell me.

4. Ad set:
- Location: find the Meta location key for [city] with search_meta_interests (kind adgeolocation), and check the country code. If I give you my business coordinates [latitude, longitude], use custom_locations with a [radius] km radius around them instead of the whole city.
- Audience and placements: ages 25-65. Set no placement fields, so placements stay automatic.
- Delivery: optimization_goal LEAD_GENERATION, billing_event IMPRESSIONS, my page_id, daily_budget [daily budget], status PAUSED.

5. Ads:
- Images: pull my recent Page posts with list_meta_page_posts and pick two with strong photos. Upload each with upload_meta_image using its object_story_id (not a copied image link, which can expire) to get an image_hash.
- Copy: draft two link ads attached to the form through lead_gen_form_id. Make one price/offer-led and one trust-led. For the trust ad, use only facts I confirm: [years in business], [rating and review count].
- Settings: primary text under 125 characters, a short headline, cta_type GET_QUOTE, link [website URL], status PAUSED.

Approval: show me the form questions, campaign, ad set targeting, the two chosen images and both ad copies as dry runs (confirm=false). Do not create anything until I approve. Once I approve, create the form, campaign, ad set and ads in that order, all PAUSED.

After creation:
6. Give me preview links with preview_meta_ad for each ad's creative_id, in MOBILE_FEED_STANDARD and INSTAGRAM_STORY.
7. Run meta_delivery_estimate on the paused ad set. This only works once the ad set exists. If expected leads are under 1 per day, propose either a wider radius (or the whole city) or a higher budget. Show the change as an update_meta_adset preview and wait for my OK. Keep everything paused. I will activate it myself.

Campaign setup

LinkedIn pre-flight: audience size, bids and budget fit before building

A comparison of three LinkedIn targeting variants by reach, suggested bid and affordable clicks, with one recommendation ready to send to management.

I'm planning a LinkedIn lead generation campaign for [product or offer], aimed at [job titles or functions] in [countries], with a monthly budget of [monthly budget]. Run a pre-flight check before anything is built.

1. Find the targeting URNs for those titles and functions, seniority (Manager and above unless I say otherwise), company size 51+ employees, and the listed geographies.
2. Forecast audience size for three variants: Narrow (job titles only), Core (titles + seniority + company size), Broad (functions + seniority).
3. For each variant, get LinkedIn's suggested bid, the bid floor and ceiling, and the minimum daily budget.

Deliver one table with these columns: variant, members reached, suggested bid, minimum daily budget, and estimated monthly clicks at my budget (budget divided by suggested CPC). If a suggestion is CPM-based, convert it using a 0.5% CTR assumption and state that you did.

Decision rule: reject any variant under 50,000 members, and any variant where my budget buys fewer than 300 clicks a month. Recommend one variant, with a two-sentence rationale I can forward to management.

Do not create anything yet. If I approve, show the campaign group and campaign settings as a preview and wait for my confirmation.

Campaign setup

Plan a Search campaign from Search Console gaps where organic is weak

A paused Search campaign built only on commercial queries where your site ranks 4-20, so ad spend covers real organic gaps.

Use my Search Console data from the last 90 days to decide where paid search should fill organic gaps.

Pull queries with at least 100 impressions, together with their ranking page. Sort them into three buckets:
(a) Average position 1-3 and CTR above 5%. Skip these, because organic already wins.
(b) Position 4-20 with commercial intent (price, service, near me, buy, book). These are paid candidates.
(c) Informational queries. Exclude these.

For bucket (b), get monthly searches and competition from Keyword Planner. Deliver a table with these columns: query, impressions, position, CTR, monthly searches, proposed ad group, match type (exact or phrase).

Then propose one Search campaign:
- 2-4 ad groups.
- One responsive search ad per group, pointing to the page that already ranks for those queries.
- My brand term [brand name] added as a negative, because organic covers it.
- Daily budget of [daily budget].

Show the complete preview and wait for my approval. Then create the campaign paused.

Campaign setup

Pre-launch conversion tracking check before creating any new campaign

A go/no-go table showing whether your conversion setup can support a new campaign and its bidding strategy.

I want to launch a new campaign whose goal is [main conversion, e.g. form submit or phone call]. First check whether my tracking can support it.

1) Run a conversion tracking health check for the last 30 days. Flag every action that is paused, removed or has recorded zero conversions.
2) List my conversion actions and goals. Tell me which ones are primary, meaning they feed bidding.
3) Check the settings of each primary action. Lead actions should count "one", not "every". Note the lookback windows.
4) Check whether GA4 is linked to Google Ads. Look for duplicates, where the same event is counted by both the Google Ads tag and a GA4 import and both are primary.

Deliver a table with these columns: check, status (pass/fail), evidence, fix.

Decision rule:
- If no primary action recorded a conversion in the last 30 days, say "do not use conversion-based bidding yet." Recommend starting on Maximize Clicks.
- If there are 15 or more conversions, Maximize Conversions is acceptable.

For any fix, show me the exact change as a preview and apply nothing until I approve.

Campaign setup

Validate Meta interest targeting before launching a prospecting ad set

Sized interest clusters with a go or no-go verdict, then a paused ad set with Meta's expected daily results before you spend anything.

I want to add a prospecting ad set to Meta campaign [campaign name]. Check whether the targeting is viable before we spend.

Audience brief: [describe the audience, e.g. homeowners 30-55 interested in solar energy]. Location: [location].

1. Search Meta interests that match the brief and return each interest's ID and estimated size. Drop interests that are only loosely related.
2. Group the remaining interests into 2-3 clusters. Estimate the reach of each cluster in the location above.
3. Apply this decision rule. Reject clusters below 200,000 people. Flag clusters above 15 million as too broad to add value over broad targeting.
4. Deliver a table with these columns: cluster, interests, estimated reach, verdict, and the reasoning in one line.

Recommend one cluster and describe the ad set settings: optimisation event, placements, age, and a daily budget of [daily budget]. Wait for my approval. Then create the ad set paused. Run a delivery estimate on it and report the expected daily results. If the estimate shows fewer than 1 conversion per day, tell me before we activate anything.

Campaign setup

Clone a proven Meta campaign into a new market

Picks the winning ad sets and ads from a source campaign, deep-copies it paused, archives the losers in the copy, retargets it to the new country, rebalances budget within Meta's safety cap and localizes flagged creatives, with approval gates throughout.

We want to replicate Meta campaign [source campaign] in [new country] for the same client, in the same ad account [ad account].

1. Find the account and the source campaign. Note whether the budget sits at campaign level (CBO) or on the ad sets, and the current daily budgets (they come back in cents, so convert them).
2. Pull performance from [start date] to [end date] (the last 60 days) at ad set level and at ad level for this campaign. Count conversions as [conversion action, e.g. purchase or lead] from the actions data and calculate CPA as spend divided by those conversions.
3. Apply this selection rule. Keep ad sets with at least 10 conversions and a CPA no more than 20% above the campaign average. Within those ad sets, keep only ads with at least 5 conversions. List everything you excluded and why.
4. Read the creative of every ad you keep. Flag anything tied to the old market: currency, city names, phone numbers, shipping claims, or a destination link that is not [new market landing URL]. Also flag any ad built from an existing Instagram post: its text and media cannot be rewritten without unlinking it from the post, so propose either keeping it as is (only the prefilled message can change) or replacing it with a new ad.
5. Render previews of the current creatives you plan to keep (Feed and Stories), so we can see what we are starting from.
6. Propose the build:
   - Deep-copy the whole campaign with the prefix [prefix, e.g. "DE - "]. The copy comes out paused. Meta has no "copy only some children" option, so after copying I will archive the excluded ad sets and ads in the copy instead of deleting them.
   - Set the location of every kept ad set in the copy to [new country]. If the source ad sets also target regions or cities, point that out, because those need to be replaced too.
   - Split a total daily budget of [daily budget] across the kept ad sets in proportion to their source conversions. If the copy runs on a campaign budget instead, set [daily budget] at campaign level. Budget increases above +50% of the copied amount are blocked by the safety cap: for any ad set (or campaign) that would need more, set it to the maximum allowed now and list the remaining steps to raise it after launch.
   - Draft replacement primary text, headline and link for each flagged creative.
   Show this full plan and the confirm=false preview of the deep copy. Wait for my approval.
7. After I approve, run the copy and list the new campaign, ad set and ad IDs. Then show confirm=false previews for every follow-up change: archiving excluded ad sets and ads, the location changes, the budget changes and the creative replacements. Wait for my approval again.
8. Once approved, apply them, keeping everything paused. Check the new targeting with a delivery estimate for each kept ad set and warn me if the audience looks too small for the budget. Render previews of the edited creatives for Feed and Stories (Facebook and Instagram).

Do not activate anything. Return the final list of new IDs with their status, location and daily budget.

Campaign setup

Plan a LinkedIn campaign for corporate gifting and bulk orders

A forecast-checked LinkedIn campaign spec with audience size, suggested bid, break-even cost per lead and attached conversion tracking.

We sell [product category] online and want B2B bulk orders (corporate gifts, employee kits) from LinkedIn. Plan the campaign in account [LinkedIn account].

1. Targeting: find job titles and functions for HR, office management, procurement and employer branding. Limit to [country] and companies with 50+ employees.
2. Forecast: check the audience size. Aim for 50,000 to 400,000 members, and adjust titles or seniority if the audience falls outside that range.
3. Bids: get LinkedIn's suggested bid, the bid range and the minimum daily budget for this audience.
4. Conversions: list the conversion rules and pick the one that tracks the bulk-order quote form. Tell me if no such rule exists.
5. Break-even: calculate it as [average B2B order value] × gross margin × an assumed 20% quote-to-order rate. Tell me if LinkedIn's suggested bid makes this cost per lead unrealistic.

Deliver a spec table with these columns: campaign group, campaign, objective, targeting, audience size, bid, budget, conversion. Then give a go or no-go line.

Create the campaign group and the campaign in DRAFT, and attach the conversion, only after I approve.

Campaign setup

Build a first ChatGPT Ads campaign from proven Google Ads search terms

A ready-to-approve ChatGPT Ads structure built on the service themes that already bring leads in Google Ads.

Campaign setup

Set up and verify ChatGPT Ads lead conversion tracking before launch

A verified lead conversion event on ChatGPT Ads, tested with a real form submission.

Campaign setup

Turn organic and Maps search queries into ChatGPT Ads context hints

New ChatGPT Ads ad groups with context hints and bids based on how locals actually search for the service.

Campaign setup

Add a qualifying Google Ads lead form to a Search campaign

A drafted Google Ads lead form with one qualifying question, previewed and attached to your campaign only after your approval.

In Google Ads account [account name], I want to add a lead form to Search campaign [campaign name] so mobile searchers can leave their details without loading the website.

1. Use list_extension_assets and list_assets to check whether a lead form already exists on this campaign or in the asset library. If one exists, show me its fields and stop. We should not run two forms side by side.
2. If none exists, draft a new form with these elements:
   - Headline and description.
   - Fields: name, phone and email only.
   - One multiple-choice qualifying question: [qualifying question, e.g. monthly budget], with 3 to 4 answer options.
   - Call to action and submission message.
   - Privacy policy URL: [privacy policy URL].
3. Show the full form as a preview, including the character count for each text field, and state whether it will attach at campaign level.
4. Wait for my approval before creating or attaching anything.

After creation, confirm the form is attached. Then explain how I pull submissions with list_lead_form_submissions so sales can work them daily.

Campaign setup

Build tiered Meta lookalike audiences from your best customers

Checks the seed audience, previews 1%, 3% and 5% lookalikes with non-overlapping tiers, sizes them against the country audience and sets up a paused 2-week tier test.

I want to prospect on Meta using people who resemble our real customers, not just page engagers.

1. Find ad account [ad account name] with list_meta_ad_accounts and list its custom audiences with list_meta_audiences. Show the type, approximate size and last update for each one, as far as the data includes them.
2. Evaluate [source audience name] as the seed. Meta needs at least 100 people from one country in the seed, and well above that gives better results. Use the approximate size if it is returned; the per-country match count for [country] is not exposed here, so ask me to confirm it from the audience's details in Ads Manager. Also judge from its type and name whether it reflects customers or qualified leads rather than all website visitors. If it is too small or too broad, recommend a better seed from the list.
3. Meta builds each lookalike from 0% up to the chosen ratio, so a 3% list already contains the 1% list. Propose three lookalikes in [country] at 1% (ratio 0.01), 3% (0.03) and 5% (0.05), with clear names, and plan the tiers so they do not overlap at ad set level: the 1 to 3% ad set targets the 3% list and excludes the 1% list, and the 3 to 5% ad set targets the 5% list and excludes the 3% list.
4. Get the total Meta audience in [country] with meta_reach_estimate (countries=[country], age [min age] to [max age]) and use it with each ratio to give a rough reach estimate per tier. Say clearly that this is an approximation; the real size appears in list_meta_audiences once Meta finishes building the list. Also check with list_meta_adsets whether any existing ad set already targets a lookalike from the same seed. If the ad set data does not show its custom audiences, say so instead of guessing, and ask me to check in Ads Manager.
5. Show the three lookalikes as a preview with create_meta_lookalike_audience using confirm=false, and wait for my approval before creating anything.

After creation, suggest a 2-week test: one ad set per tier with the exclusions above (custom_audience_ids and excluded_custom_audience_ids), the same creative and an equal daily budget of [daily budget per ad set]. Ask me whether to use existing campaign [campaign name] (find it with list_meta_campaigns) or a new one; if new, preview create_meta_campaign with status PAUSED and no campaign budget, so each ad set keeps its own equal budget. Preview each ad set with create_meta_adset using confirm=false and status PAUSED, and create nothing until I approve. Once created, run meta_delivery_estimate on each ad set to sanity-check the estimated audience and results.

When the test ends, pull results per ad set with meta_campaign_insights at adset level. Judge the winner on cost per qualified lead, not on CTR: Meta reports leads, not lead quality, so ask me for the qualified lead count per ad set from our CRM and calculate it from that.

Campaign setup

Audit every audience list across ad platforms before next quarter

A read-only inventory of audience lists on Google Ads, Meta, TikTok, LinkedIn and Microsoft. It covers size, status and usage, and flags small, broken, duplicate and unused lists, with ranked cleanup recommendations.

Before next quarter's planning, I need a clean inventory of our audience lists on every connected ad platform: Google Ads, Meta, TikTok, LinkedIn and Microsoft. This is read-only. Do not create, delete or change anything.

1. Find the connected account on each platform. Skip any platform that is not connected and say so.
2. Pull the audiences:
   - Google Ads: list_audiences with definition=true. This shows each list's rule, its lookalike seed, its membership duration and its Search and Display sizes.
   - Meta: list_meta_audiences.
   - TikTok: list_tiktok_audiences.
   - LinkedIn: list_linkedin_audiences. Use usable_for_targeting and below_delivery_minimum.
   - Microsoft: list_microsoft_audiences.
3. Check which audiences are actually in use:
   - Google Ads: read the audience criteria of all campaigns with get_campaign_targeting, and keep only enabled campaigns (check status with list_entities). For Demand Gen campaigns, read each ad group with ad_group_id, because the audiences live there. For Performance Max, read the audience signals with get_pmax_assets.
   - Meta and TikTok: list the active ad sets and ad groups. If the listing includes the custom audience ids they target or exclude, use them. If it does not, mark usage "not visible".
   - Microsoft: run microsoft_report_query with report_type AudiencePerformance for the last 30 days, with columns such as CampaignName, AdGroupName, AudienceId, AudienceName, Impressions, Clicks and Spend. An audience with delivery in that report counts as used. If the report is rejected or returns nothing, mark usage "not visible".
   - LinkedIn: list the ad sets with list_linkedin_campaigns. If the rows show the matched audience URNs in their targeting, use them. Otherwise mark usage "not visible" instead of guessing.

Build one table with these columns:
- Platform
- Audience name
- Type (website, customer list, lookalike, engagement, other)
- Size
- Status
- Membership duration, if the platform returns it
- Used by an active campaign: yes, no, or not visible

Flag an audience if any of these is true:
- On Google, it is below 1,000 users for Search or below 100 for Display.
- Its status is closed, failed, archived or expired, or the platform marks it as too small to serve. On TikTok, treat a customer list under about 1,000 matched users as too small. On LinkedIn, use below_delivery_minimum and usable_for_targeting.
- It repeats another list's rule on the same platform. Only Google returns rule definitions. On the other platforms, flag lists with near-identical names and the same source as "possible duplicate".
- It is a lookalike whose seed list is missing, closed or tiny, where the platform shows the seed.
- It is not used by any active campaign. Flag this only where usage is visible.

Deliver three things:
1. Total audience count per platform, and how many of them are usable.
2. The flagged lists, grouped by problem type.
3. The top 5 cleanup or rebuild actions, ranked by impact on targeting quality. List them as recommendations only.

Add one line per platform on what could not be checked. Then write the summary so I can paste it straight into a management update.

Campaign setup

Add a Google lookalike segment to a Demand Gen ad group

Picks the best seed list, builds a lookalike segment sized from recent ad group data and attaches it to a Demand Gen ad group, with a preview before each step.

Our Demand Gen campaign [campaign name] runs only on broad targeting. I want to test a lookalike segment in ad group [ad group name].

1. Find the campaign and the ad group ids. Read the ad group's current audiences with get_campaign_targeting and ad_group_id, so we do not add something that is already there.
2. List the account's audiences with definition=true, and rank the candidate seeds by size and intent:
   - Customer Match lists
   - Converters (lists built on a thank-you or purchase page)
   - Lead form submitters
   - High-intent page visitors (pricing, contact, product pages)
   Use each list's rule to decide which group it belongs to. Exclude any seed that is under 1,000 members or whose membership status is not open.
3. Recommend the best one or two seeds, and explain your pick in one sentence each.
4. Pull the ad group's last 30 days with get_report at ad group level: impressions, conversions and cost per conversion. Based on that, recommend NARROW, BALANCED or BROAD for a lookalike in [country], and explain why.
5. Preview the lookalike with create_lookalike_audience. Wait for my approval before creating it.
6. After it is created, preview attaching it to the ad group as targeting with manage_audience_targeting. Wait for my approval again. A new lookalike can take a while to fill, so check its size with list_audiences and tell me if it is still building.

After both steps are applied, tell me which metrics I should compare after 14 days against the previous 14 days, using get_report with start and end dates:
- Conversions
- Cost per conversion
- Conversion rate

Also state the minimum conversion count I need before I call a winner.

Campaign setup

Build an ad schedule from 90 days of day and hour performance

A day-by-hour performance grid with clear rules for bid adjustments or exclusions, ready to apply after approval.

Analyze campaign [campaign name] in my Google Ads account for the last 90 days.

1. Read the current ad schedule and the bidding strategy.
2. Pull clicks, cost, conversions and cost per conversion by day of week and by hour.
3. Group the hours into 3-hour blocks so each block has enough data.

My target cost per conversion is [target CPA]. Apply these rules only to blocks with at least 30 clicks:
- Cost per conversion above 1.5x target: bid -30%.
- Spend above 2x target with zero conversions: bid -50%, or exclude the block if it is overnight.
- Cost per conversion below 0.7x target: bid +15%.
- Everything else: no change.

If the campaign uses Smart Bidding, tell me that bid adjustments are mostly ignored. In that case, only consider full exclusions for clearly wasteful blocks.

Deliver:
- A grid of day by time block with cost per conversion.
- The proposed schedule with a reason for each change.
- The estimated monthly spend that moves.

Preview the new schedule and wait for my approval before applying it.

Campaign setup

Copy proven Google sitelinks and callouts to Microsoft Ads

Ranks a Google campaign's sitelinks and callouts by real 90-day performance, finds what is missing in Microsoft Ads, and previews adding the best ones without duplicates.

My Microsoft Ads campaigns have fewer extensions than my Google Ads campaigns. I want to reuse what already works.

1. In Google Ads, list the sitelinks and callouts on campaign [Google campaign name], with each sitelink's link text, description lines and final URL.
2. Pull per-asset extension performance for that campaign over the last 90 days. Keep only assets with at least 300 impressions, calculate CTR as clicks divided by impressions, and rank by clicks, then by CTR.
3. In Microsoft Ads, find campaign [Microsoft campaign name] and list the sitelinks and callouts already attached to it.

Deliver one table with these columns: extension text, URL, Google impressions, Google clicks, Google CTR, and whether it already exists in Microsoft. Mark exact or near duplicates, because duplicates are a common cause of disapproval in Microsoft Ads.

Propose adding up to 6 sitelinks and 6 callouts that are missing in Microsoft, taking the best Google performers first. Rules:
- Sitelink display text and callouts must stay within 25 characters. If one needs shortening, show the new version next to the original.
- Carry sitelink descriptions over only as a pair (both lines or none).
- Point each sitelink to the same landing page as in Google.
- Skip any asset that mentions a Google-only offer or a phone number.

Show the exact list as a preview and wait for my approval before adding anything to Microsoft.

Campaign setup

Launch a dated promotion asset and a matching Business Profile post

A time-limited promotion asset for your Search campaign plus a matching Business Profile post, both previewed before going live.

I am running a time-limited offer: [offer details]. It runs from [start and end dates] and the landing page is [landing page URL].

1. Check campaign [campaign name] for existing promotion assets. Flag any that are expired or that conflict with this offer.
2. Draft a promotion asset with:
   - the occasion, if one fits;
   - the discount type, either percent or amount;
   - the item text, up to 20 characters;
   - start and end dates, so it stops showing automatically;
   - the landing page above;
   - a minimum order value or promo code, only if my offer includes one.
3. Draft 2 callouts that support the offer, up to 25 characters each, for example the deadline or a free extra.
4. Draft a Business Profile update post for my location. Use the same offer, up to 300 characters, with a "Learn more" button pointing to the same page.

Make sure the offer, the dates and any conditions are worded identically in the ad assets and the post.

Show all drafts together as a preview. Wait for my approval before adding the assets or publishing the post.

Campaign setup

Build a paused non-brand Search campaign from converting search terms

A ready-to-approve Search campaign build: themed ad groups, keywords, RSA drafts and brand negatives, created paused only after your approval.

In Google Ads account [account], analyse search terms and keyword data from the last 90 days. Find search terms with at least 2 conversions and a ROAS above [target ROAS] that are not yet added as keywords. Leave out anything containing our brand name.

Group these terms into 3 to 6 tightly themed ad groups by product category. Give me a table for each ad group with these columns:
- Keywords in exact and phrase match, with the 90-day conversions, conversion value and ROAS of each source term
- Suggested final URL. Ask me to confirm it if the data does not show a clear landing page.
- One RSA draft with 15 headlines (max 30 characters) and 4 descriptions (max 90 characters). Mention price, shipping or returns only if I confirm the offer.

Also list our brand terms to add as campaign-level negative keywords, so the campaign stays non-brand. Use a daily budget of [daily budget] and Maximize conversion value without a ROAS target until the campaign has 30 conversions.

Show the full build as a preview. After I approve it, create the campaign PAUSED. Do not change any existing campaign.

Campaign setup

Demand Gen retargeting campaign using your best YouTube product videos

Finds your strongest product videos by retention on YouTube, then builds a paused Demand Gen retargeting campaign aimed at product page visitors, with ad group level targeting and one video ad per shortlisted video.

From my connected YouTube channel (check the channel list first if there is more than one), list the top 10 videos of the last 90 days by views, with views, watch time and average view duration for each.

Average percentage viewed per video is not available from the YouTube data here. If I give you each video's length, estimate it as average view duration divided by video length. Otherwise rank retention by average view duration and say so.

Based on the titles, mark which videos look like product videos, then ask me to confirm which ones show a product we sell and which product page each one belongs to. From the confirmed videos, shortlist the 2 with the strongest retention and explain the choice. If the video IDs or URLs are not in the data, ask me to paste the links.

Then prepare a Demand Gen retargeting campaign in Google Ads account [Google Ads account]:
- Daily budget [daily budget], bid strategy Maximize conversions, created PAUSED.
- One ad group named [ad group name]. Create it inside the new campaign. If Google rejects the ad group for a Demand Gen campaign, tell me, I will create it in the Google Ads interface, and you find its ID from the campaign's ad group list before continuing.
- Targeting on that ad group: [country] and that country's main language. Set it at ad group level, because Demand Gen keeps location and language there and otherwise serves in every country.
- Audience: website visitors whose page URL contains [product URL path], 30-day membership. First check the existing audiences with their rule definitions and reuse one whose rule matches instead of creating a duplicate. Attach the audience to the ad group. If the list is new or small, warn me that it needs visitors before the campaign can deliver.
- One video ad per shortlisted video: business name [business name], headline (max 40 characters), long headline (max 90), description (max 90), a call to action fitting a purchase, and a final URL pointing to the product shown in the video. For the required square logo, reuse a square logo already in the asset library if there is one; otherwise ask me for a logo image URL.

Show the complete setup as a preview first, including the audience rule, the retention numbers behind the video choice and character counts for every text. Wait for my approval before creating the audience, campaign, ad group, targeting or ads, and create each one with a preview step so I can see exactly what is applied.

Campaign setup

Paid Search campaign for product queries stuck on organic page two

A ranked list of high-impression commercial queries ranking 8 to 20 organically, with no paid coverage, turned into a paused Search campaign draft.

In Search Console property [property], pull the last 90 days of queries for product and category pages. Keep queries with an average position between 8 and 20, at least 200 impressions, and clear buying intent (product names, models, "buy", "price", "order"). Drop informational and brand queries.

Cross-check these against the Google Ads search terms report for the same period and remove any query that already gets paid clicks. For the rest, get average monthly searches, competition and top-of-page bid ranges from Keyword Planner.

Give me a ranked table with these columns: query, landing page, organic position, impressions, monthly searches, top-of-page bid range, priority. Rank by monthly searches multiplied by commercial intent, and flag any query whose high bid range would make a [target ROAS] ROAS unrealistic.

Then group the top queries into 2 to 4 ad groups. Draft exact match keywords and one RSA per ad group (15 headlines, 4 descriptions), using the ranking page as the final URL. Use a daily budget of [daily budget]. Show it as a preview, and create the campaign PAUSED only after my approval.

Campaign setup

Mirror a proven Google Search campaign into Microsoft Ads

A paused Microsoft Ads campaign built only from the Google keywords, ads, negatives and locations that already convert, previewed before creation.

Use the Google Ads campaign [Google campaign name] as the template for a new Microsoft Ads campaign with a daily budget of [daily budget]. Use the last 90 days of data.

What to carry over:
- **Keywords:** those with at least 1 conversion, plus those with 100+ clicks and a CTR above the campaign average. Leave out any keyword whose spend exceeds 2x the campaign's average cost per conversion with zero conversions.
- **Match types:** keep exact and phrase as they are. Convert broad match keywords to phrase.
- **Ad groups:** keep the same ad group structure.
- **Ads:** one RSA per ad group, built from the best-performing headlines and descriptions. Drop any asset that has "Low" performance or a CTR below the ad group average.
- **Negatives:** all campaign-level and ad-group-level negatives, plus the shared negative lists attached to the campaign.
- **Locations:** the same location targets, matched to Microsoft LocationIds.

Before creating anything, show me a preview table: ad group, keywords with match type, RSA headlines and descriptions, negatives count and locations. Also list anything that could not be mapped.

Wait for my approval. Then create everything in the paused state and give me a one-paragraph summary I can share with management. The summary should cover what was launched and why this subset was chosen.

Campaign setup

Build a standard Meta retargeting and lookalike audience set

Checks existing audiences and pixel health, then proposes and builds a deduplicated set of website, converter, engagement, lead form and lookalike audiences with a consistent naming scheme.

Build our standard audience set for Meta ad account [client ad account].

First, list the existing custom and lookalike audiences and the pixels with their last fired time. Flag any audience that already covers the same rule and window so we do not create duplicates. Also flag any pixel that has not fired in the last 24 hours. If the main pixel is stale, stop and tell me. Then find the Facebook Page, the Instagram account and the instant forms on that Page, so you have the source IDs for the engagement audiences.

Then propose these audiences:
- Converters in the last 180 days, from the pixel's [Purchase or Lead] event (the preferred lookalike seed)
- Website visitors in the last 30 days and in the last 180 days
- Visitors of [high-intent path, e.g. /pricing] in the last 30 days. The website audience tool builds audiences from a pixel event, not a URL rule, so: if a standard event such as ViewContent or InitiateCheckout fires only on that page, use it and say so; otherwise give me the exact URL-contains settings to create this one in Ads Manager.
- Instagram profile engagers and Facebook Page engagers in the last 90 days
- People who opened a lead form but did not submit it, in the last 90 days, from each relevant form, if the lead form source offers that event. If it does not, tell me which lead form events are available.
- Lookalikes in [country] at 1% and at 3%, seeded from the strongest source: converters, or the 180-day visitors if the converter audience is too small. Note that the 3% lookalike includes the 1% band; if I want a pure 1-3% band, exclude the 1% lookalike in the ad set.

Name each audience CLIENT_TYPE_SOURCE_WINDOW. Deliver a table with these columns: name, source, rule or event, window, purpose (retarget, exclude, or seed).

Do not create anything until I approve the table. Show each creation as a dry run (confirm=false) first. Create the seed audiences before the lookalikes. After creation, list the audiences again. Flag any that the listing shows as below 1,000 people or still populating.

Campaign setup

Grade Meta lead quality by ad set and send qualified leads back to Meta

Cost per qualified lead by ad set, a tighter form proposal, and a Conversions API upload of qualified leads, all held for approval.

Campaign setup

Build a purchase-optimised TikTok campaign from scratch

Verifies the pixel's purchase setup, then builds a website conversions campaign, a targeted ad group and three video ads from your library, previewed in full and kept off until you approve launch.

Set up a TikTok campaign for [product or collection] in ad account [TikTok account], optimised for purchases.

1. Pixel check: list the pixels and confirm that one has activity status ACTIVE and has a purchase event (CompletePayment, or the purchase event as TikTok lists it) configured as an optimisation event. TikTok's API does not expose event fire counts, so if the setup passes, remind me to confirm recent purchase events in Events Manager. If there is no active pixel or no purchase event, stop and tell me exactly what is missing.
2. Identity: use my existing identity from the identities list. If there is none, stop and tell me.
3. Campaign: website conversions objective, daily budget of [daily budget]. Note that TikTok stops accepting new campaigns with this objective on 1 January 2027. If the campaign comes back enabled after creation, disable it right away.
4. Ad group:
   - Location: [country], using the location ID found for the website conversions objective.
   - Ages 18 to 44 (18-24, 25-34, 35-44).
   - Interests: search and suggest 3 to 5 relevant interest categories with their IDs. If the daily budget is under 50 times my target cost per purchase of [target cost per purchase], recommend broad targeting with no interests instead, and explain why.
   - Optimisation goal: conversion, with the pixel above and the purchase event as the optimisation event.
5. Ads:
   - From my video library, pick the 3 newest videos that are 9 to 30 seconds long, based on the duration returned for each video.
   - For each video, write ad text of up to 100 characters with no price claims I have not given you.
   - Call to action: Shop now. Landing page: [landing page URL] with UTM parameters utm_source=tiktok, utm_medium=paid, utm_campaign=[campaign name].
   - Use the identity from step 2.

Show a full preview covering campaign settings, ad group targeting, budget, optimisation event and each ad (video, text, CTA, URL). Create everything only after I approve. The ad group is created disabled; if any ad comes back enabled, disable it. Keep the campaign, ad group and ads off until I confirm launch.

Campaign setup

Write ChatGPT Ads chat cards from my services and customer reviews

Turns your Business Profile services and the phrases customers repeat in reviews into 3 ChatGPT Ads chat card variants with an image, then creates the approved ones paused and checks their preview and delivery blockers.

Write ad copy for my ChatGPT Ads ad group [ad group id, adgrp_...]. The copy must sound like my business, not like generic ads.

1. Find my listing with gbp_list_locations and read it with gbp_location_details: services, business hours, description and the city in my address.
2. Read my Google reviews with gbp_reviews and keep the ones from the last 12 months. Note my average rating and review count. Pull out the phrases customers repeat most, such as speed, price clarity or cleanliness. Ignore anything I cannot prove.
3. Read the ad group with chatgpt_ads_details: its context hints and parent campaign, so the copy matches the conversations it is set to match.
4. Sort out the image, because a chat card is refused without one. If I give you an image URL [image URL, optional], use it. Otherwise list my profile photos with gbp_list_photos and suggest the 3 best fits (exterior, at work or team shots usually beat interiors). Turn the one I choose into a public URL with upload_image using source_url.
5. Draft 3 chat card variants. Each one should take a different angle:
   - fast response or same-day service, only if my hours or reviews support it
   - trust, using the rating and review count
   - a clear offer: [offer, e.g. free quote]
   Give each variant a title, body text, the destination URL [landing page URL] and the image.

Copy rules:
- Plain and factual.
- No superlatives I cannot back up.
- Mention my city.
- The card has no call button, so put the action in the body text itself, e.g. "Call us or request a quote on our site."
- Keep the title short. If a dry run rejects the length, shorten it and show me the new version.

Show the 3 variants in a table next to the review phrases they use. After I pick, run create_chatgpt_ads_ad with confirm=false for each chosen card and show me the exact dry run. Creating an ad submits it for OpenAI review, so text, image and URL must be final. Create them only after my approval, with confirm=true and without activate, so they stay paused. Then render each ad with chatgpt_ads_ad_preview and list the review status and any delivery blockers OpenAI reports.

Campaign setup

Performance Max asset and signal readiness check

Audits a paused Performance Max campaign asset group by asset group: text, image, logo and video counts, search themes, audience signal, final URL, primary conversion goals, presence targeting and disapprovals, then drafts missing text assets for approval.

Check whether the paused Performance Max campaign [campaign name] in [client account] is ready to launch. Do not change anything without my approval.

Read every asset group in the campaign with its assets and signals. For each one, report the current count against the target:
- Headlines: 15. Long headlines: 5. Descriptions: 5, with at least one of 60 characters or fewer.
- Landscape, square and portrait images, and at least one logo. Count portrait images only if the asset data shows the field type; otherwise mark it as not readable.
- Videos: if there are none, flag it, because Google will auto-generate video from the images.
- Search themes: present and relevant to the final URL.
- Audience signal: includes a customer list or a remarketing audience, not only interests. Report each list's match size and flag any customer list under 1,000 members.
- Final URL: fetch it, it must return 200.
- Ad strength of the asset group.

At campaign level, check three things:
- Which conversion goals are primary for this campaign. Flag page views or micro-conversions set as primary.
- Whether the location setting is presence only (PRESENCE) rather than presence or interest.
- Whether any asset is disapproved or limited. Check the ad policy report, and query the asset group assets' policy approval status with a GAQL query if the report does not cover PMax assets.

Output a gap table with the columns Asset group | Item | Have | Need.

Then read the landing page text and draft the missing text assets from it: headlines up to 30 characters, long headlines and descriptions up to 90. Use only claims that appear on the page or that I give you here: [offers, USPs, proof points]. Images, logos and videos are listed as gaps for me to supply, not created.

Do not add anything yet. Show the confirm=false preview of the text asset changes per asset group and wait for my approval. Keep the campaign paused.

Campaign setup

TikTok launch readiness: account, identity, pixel and ad groups

Pre-launch checklist for a paused TikTok campaign. It covers Business Center qualification, ad identities, pixel events, ad group location and budget, creative length, landing pages and possible Smart+ overlap. Any fix needs your approval first.

Run a launch checklist for the TikTok campaign [campaign name] in [TikTok ad account]. It is not live yet. Target country: [target country]. Target cost per result: [target CPA in account currency].

Check:
1. Account: list the Business Centers I can access and the advertisers under each one. If this ad account is under one of them, read its qualification status: business verification, industry qualification and account standing. If it is not under any Business Center, say this check cannot be run here and I should check it in TikTok Ads Manager.
2. Identity: take the identity id and type that each ad uses and match it to the identity list to get the display name. Confirm the display name is the client's brand. If the identity list comes back from existing ads with no display name, say so. The avatar cannot be inspected here, so ask me to confirm it in TikTok Ads Manager.
3. Pixel: a pixel exists, its activity status is ACTIVE, and the optimization event used by each conversion ad group is configured on that pixel.
4. Ad groups: read the campaign objective. Then search [target country] in the region search with that same objective and compare the returned location ids with each ad group's location ids. Flag any id that is not the target country or a place inside it, and give its name if the search shows it. Also report age range, schedule, bid type and bid, and budget. If the budget is set at campaign level, check it there. Flag any daily budget below the 20 USD equivalent, and any budget below 20x the target cost per result.
5. Ads: for each ad, use its video_id to find the video in the video library. For Spark ads, use its tiktok_item_id to find the organic post on that identity. Confirm the duration is between 9 and 60 seconds. Ad text is 100 characters or fewer (Spark ads take their caption from the post, so note that separately). For each unique landing page URL, confirm UTM parameters are present and read the live page once to confirm it returns 200.
6. Overlap: list the Smart+ campaigns and ad groups in the same account. If the listing shows their location, age, gender or audience settings, compare them with this campaign and flag likely overlap. If it does not, list them with their status and ask me to check their targeting in Ads Manager. Do not guess the overlap.

Output a table with the columns Check | Status | Detail, then a list of blockers ranked by launch risk.

If an ad group's targeting, bid or budget needs fixing, edit it in place (do not rebuild it). Show me the confirm=false preview and wait for my approval. Budget increases are limited to +50% per change. Do not enable the campaign, ad groups or ads.

Campaign setup

Microsoft Ads imported campaign check before enabling

Checks a campaign imported from Google Ads into Microsoft Advertising before it goes live. It covers UET and conversion goals, bid strategy, locations, negatives, ad groups, keywords and extensions. Every fix comes as a preview for your approval.

We imported the Google Ads campaign [Google campaign name] into Microsoft Advertising as [Microsoft campaign name] for [client account]. Before enabling it, check what the import broke or missed. Do not change anything without my approval.

1. Tracking: a UET tag exists and its reported status is active. At least one conversion goal uses that tag and its status shows it is recording. Take the final URL of one ad and read that live page. If the UET tag id or bat.bing.com appears in the HTML, say so. If it does not, say the tag may load through Tag Manager and cannot be confirmed from the page source. As a cross-check, report conversions on the account's other Microsoft campaigns over the last 30 days. If there are none, say tracking is unproven.
2. Bidding: read the campaign's bid strategy from the account snapshot. If it is Target CPA, Max Conversions or another conversion-based strategy and no goal is recording, flag it. In that case recommend Enhanced CPC until conversions come in.
3. Locations: read the Google campaign's location targets and exclusions. Microsoft location targets cannot be read here, so ask me to paste them from the Microsoft campaign settings. Then compare the two and flag extra or missing locations. For missing targets, look up their Microsoft location ids so they can be added after a preview. Extra targets and any missing exclusions I will fix in the Microsoft interface.
4. Negatives: collect the Google campaign's ad group and campaign negatives. List the shared negative lists attached to the campaign and read each one in full, raising the limit if a list is cut short. Compare them with the negatives on the Microsoft campaign and on each Microsoft ad group, matching ad groups by name. List every negative missing on Microsoft and the level it belongs to. Microsoft negatives accept only Exact and Phrase, so map Google broad negatives to Phrase and say so.
5. Ad groups, ads and keywords: every ad group is active and has at least one active RSA. Flag paused ad groups. Flag keywords that came over paused or with a zero bid. If the ad group data shows an end date or a network setting, flag any end date that would stop the campaign early.
6. Extensions: at least 4 sitelinks and 2 callouts are attached to this campaign, with no duplicates.

Output a table with the columns Check | Status | Detail.

Then show the confirm=false previews one at a time and wait for my approval on each:
- the missing negatives, as one batch for the campaign plus one batch per ad group
- any bid strategy switch
- any missing locations
- any paused ad group I should switch on

Change the campaign status to Active only after I confirm, and show that preview too.

Campaign setup

Turn top Instagram posts into WhatsApp message ads

Ranks your recent Instagram posts and Reels by engagement and runs the top 3 as they are in a paused, local WhatsApp conversation campaign with a pre-filled message.

Many of my customers prefer to message rather than fill in a form. Build a WhatsApp conversation campaign from my best Instagram content.

1. Find my Facebook Page and the Instagram account linked to it. List my recent posts and Reels with likes, comments and, for Reels, views. If the data shows publish dates, keep the last 90 days. Rank them by engagement and skip posts whose caption only promotes an offer that has clearly ended.
2. Pick the top 3 and explain each pick in one line.
3. Draft the setup:
- a campaign with the Engagement objective
- one ad set optimised for conversations with WhatsApp as the destination, my Page as the promoted Page, a [radius] km radius around these coordinates: [latitude, longitude of the business], ages 25-65, [daily budget] per day
- three ads that run the existing Instagram posts as they are, so their likes and comments stay on them
- on each ad, the WhatsApp button with this pre-filled message: "[pre-filled message, e.g. Hi, I need a quote for...]"
4. Remind me that my WhatsApp Business number must be connected to the Page for these ads to deliver. You cannot check this directly, so confirm it later from the ad status.

Show the full settings as a preview first. Create everything paused, and only after I approve. Then, for each paused ad, take its creative id (from the creation result or by reading the creative) and give me preview links for Instagram feed, Instagram Reels and Instagram Story. Finally, check each ad's status and report any delivery issue Meta lists with the fix. If the button does not match the WhatsApp destination, propose correcting it on the existing Instagram post ad so the likes and comments are kept, and wait for my approval.

Campaign setup

Set up Meta conversion tracking for form leads and phone clicks

Checks your Meta pixel, WordPress site and GTM container for gaps, then drafts and publishes, step by step with approval, a Lead conversion for Contact Form 7 submissions and a Contact event for tel: clicks.

Before I spend on Meta ads that send traffic to my website, make sure website forms and phone clicks are tracked as conversions.

1. List my Meta pixels with their last fired time, and my existing custom conversions. Point out any custom conversion that has never fired. Last fired time is the only pixel health signal available here, so if a pixel looks stale, tell me to check Events Manager for event details.
2. On my WordPress site [site URL]:
- show every place GTM or GA4 fires and flag duplicate containers. Note the GTM container id the site loads.
- check for a Meta pixel installed outside GTM: a pixel plugin in the plugin list (for example PixelYourSite or Facebook for WooCommerce), "fbq(" in page content, a WPCode snippet containing fbq, or fbq in the active theme's header file
- list my Contact Form 7 forms with their required fields and notification email
- check whether a submission goes to a thank-you page: search page content and snippets for "wpcf7mailsent" redirects or a link to [thank-you page URL], check that page exists and is published, and if possible fetch the URL to confirm it loads. If you cannot tell for sure, say so and ask me.
3. In GTM, find the container whose id matches the one on my site (if they differ, stop and tell me). List the current tags, triggers and variables in the workspace I should use. Read the Custom HTML tags' code to see whether the Meta pixel base code (fbq('init', ...)) is there, and read the trigger conditions to see whether anything fires on tel: link clicks.
4. Draft the fixes:
- if forms go to a thank-you page: a Meta custom conversion on my pixel for URLs containing [thank-you page path], categorised as Lead
- if forms only show an on-page message: say that a URL rule will never fire, and draft a Custom HTML listener tag on All Pages that pushes a "cf7_success" dataLayer event when "wpcf7mailsent" fires, a custom event trigger on "cf7_success", and a Custom HTML tag on that trigger that sends a Lead event to my pixel
- turn on the Click URL built-in variable, then a Just Links click trigger for clicks where Click URL starts with tel:
- a Custom HTML tag that sends a Contact event to my pixel [pixel name] on that trigger
- only if the pixel base code is missing everywhere: a base code tag with my pixel id on All Pages. If the pixel already loads through a plugin, a snippet or the theme, do not add a second copy.
- if there is no All Pages trigger to use, include one in the draft
5. Optional: offer to send one test submission to my form to confirm the notification mail goes out. It sends a real email, so ask me first.

Present the gaps as a short checklist, then show the exact changes as a preview (every tool call with confirm=false first). Do not create the custom conversion, enable the built-in variable, create the GTM items, or publish the container until I approve each step. Before publishing, show me the GTM workspace status so I see exactly what goes live. After publishing, check the pixel's last fired time again and tell me to confirm the Lead and Contact events in Events Manager's Test Events.

Campaign setup

Pre-launch QA of a paused Meta lead campaign

Audits a paused Meta lead or messaging campaign before it goes live: objective, optimisation goal, locations, budget versus target CPL, delivery estimates, creative enhancements, lead forms, landing pages and placement previews. Every fix is shown as a preview first.

Run a pre-launch check on my paused Meta campaign [campaign name]. My target cost per lead is [target cost per lead]. My service area is [service area].

Check each item and mark it PASS, WARN or FAIL:
1. Objective and optimisation. The campaign objective is Leads (OUTCOME_LEADS), or Engagement for a messaging campaign, and every ad set optimises for LEAD_GENERATION, QUALITY_LEAD, OFFSITE_CONVERSIONS on a lead event, or CONVERSATIONS, not LINK_CLICKS or LANDING_PAGE_VIEWS. The objective of an existing campaign cannot be changed, so if it is wrong the fix is a new paused campaign: show it as a preview and list which ad sets and ads would need to be rebuilt inside it.
2. Locations and ages. Locations stay inside my service area and age ranges make sense for the offer. Compare locations, ages and custom audiences across the ad sets and flag pairs that target the same people, since they compete in the auction. This is a settings comparison, not a measured overlap, so mark it WARN rather than FAIL.
3. Budget. The daily budget per ad set (or the campaign budget divided by the number of ad sets, if budget sits at campaign level) is at least 3x my target cost per lead. Budgets come back in cents, so convert them before comparing. Run the delivery estimate on each ad set and flag any whose audience is very small or whose expected daily results are near zero.
4. Creative. For each ad, read the creative: headline, primary text, button, destination and which Advantage+ creative enhancements are on. Flag any enhancement that could rewrite my claims or prices, or add music or visual edits I did not ask for. Also report any delivery issue Meta already lists on the ad (effective status and issues), and check that the button matches the destination (for example a WhatsApp ad set needs a WhatsApp button).
5. Lead forms. List the Pages this ad account can advertise for, then the instant forms on the right Page, and check each form's status and questions (too many questions, missing phone, free-text where a choice list would qualify better). The form list does not return the privacy policy link or the thank-you screen, so mark those WARN and tell me to check both in Ads Manager. If the creative read shows which form an ad uses, use it; if not, ask me which form each ad uses.
6. Website ads. For each destination URL, read the live page and check that it returns 200, that the title and H1 match the ad's promise and that a phone number appears in the page text. Flag redirects, 404s and slow or empty responses.
7. Previews. Using the creative id from the creative read, give me preview links for every ad in mobile feed, Instagram feed and Instagram Story.

Output a table with the columns item, entity, status, finding and fix.

For every FAIL, show the exact change as a preview with confirm=false: a new paused campaign for a wrong objective, the ad set update for optimisation goal, locations, ages or budget, the campaign budget update if budget is at campaign level, and the creative update with the enhancements switched off. Budget increases above 50% are blocked in one step, so split larger raises into steps. If an ad was built from an existing Instagram post, do not rebuild its creative; flag it and tell me to switch the enhancement off in Ads Manager instead. Apply nothing and do not activate the campaign until I approve.

These prompts work best when your AI assistant is connected to your accounts with Opus Growth.

Connect your accounts free