Ad copy and creative: Prompt library

Ad copy and creative

Find weak RSA headlines and descriptions and replace them in place

A per-asset performance table with clear weak-asset flags and two tested-angle replacements for each, applied only after approval.

Ad copy and creative

Check message match between RSAs and WordPress landing pages

Reads every enabled RSA in a campaign, matches each final URL to its WordPress page (Elementor included) and rates whether the page repeats the ad's keyword, offer and CTA. Proposes the smallest fix and applies nothing without approval.

Check whether my ads and landing pages say the same thing. Google Ads account: [account]. WordPress site: [site].

Step 1: Ads
List the ad groups in campaign [campaign] and read every enabled RSA with its headlines, descriptions and final URL.

Step 2: Find each page
- Strip tracking parameters from each final URL and match it to a page or post on [site] by its slug.
- If a URL is not on [site], or matches more than one page, list it and ask me instead of guessing.

Step 3: Read each page
- First check how the page is built. On an Elementor page the visible headings and buttons live in the page design data, so read them there. On other pages, read the full content.
- Pull: the H1 heading, the first paragraph, the main offer or first button text, and the meta title from the SEO fields.
- You cannot see the rendered page. Treat the H1, the first paragraph and the first button in page order as the "first screen." Say so when the order is unclear, for example when a slider shortcode sits at the top.

Step 4: Rate each ad and page pair
Does the page repeat the ad's main keyword, offer and call to action on the first screen?

Show a table with these columns:
- ad group
- final URL
- page builder
- what the ad promises
- what the page says
- match (good, partial or missing)

Step 5: Fixes
For every partial or missing match, suggest one fix: either a headline edit in the ad or a heading edit on the page, whichever is the smaller change. Headlines must stay within 30 characters. Show the character count.

Preview the ad edits and the page edits separately:
- Ad edits: run each RSA change as a preview first, keeping every other headline, description and pin as it is.
- Page edits: the WordPress tools have no dry run, so show me the exact current text, the proposed text and the element or field you will change.

Do not apply anything until I confirm. After I approve:
- Apply the approved RSA edits. Note that edited ads go back through Google review.
- Apply the approved page edits, then purge the site cache (and regenerate Elementor CSS on Elementor pages).
- Confirm each change landed. Remind me that Elementor edits can be rolled back to the previous version and other edits are kept as WordPress revisions.

Ad copy and creative

Rewrite RSA headlines that caused disapprovals or limited serving

Lists disapproved and limited RSAs with Google's policy reason and evidence, rewrites only the assets Google flags, and edits the existing ad after a before and after preview. Flags issues that text changes cannot fix.

Some of my ads are disapproved or limited.

List every responsive search ad in [account] that is disapproved or "Eligible (limited)." For each one, show:
- campaign, ad group and ad ID
- the policy topic and its type
- the exact headline or description that triggered it, from Google's evidence
- the countries it is limited in, if Google reports that

Read each ad's full copy, including pins, so you can match the evidence to specific assets.

Explain each reason in plain language. Examples include trademark use, excessive capitalization, unproven claims like "best" or "No. 1," symbols or punctuation, and restricted categories such as health or finance.

Rewrite only the offending assets:
- Rewrite an asset only when Google's evidence points to it. If Google gives no detail for a topic, do not rewrite anything for that ad. Instead, list the assets you suspect and why, and ask me.
- Keep the same meaning but make it comply.
- Headlines can be up to 30 characters and descriptions up to 90. Show the character count for each.
- Edit the existing ad rather than creating a new one, so its history is kept. Leave every other asset and pin as it is.

Some problems cannot be fixed by rewriting text. Examples are a trademark that needs authorization, a missing certification, or an issue on the landing page such as a broken destination. For those, tell me what I need to do instead.

Show a before and after preview for each ad and wait for my approval before changing anything. Remind me that an edited ad goes back into review.

Ad copy and creative

Meta creative fatigue check with copy-only replacement creatives

A list of fatigued Meta ads by frequency, CTR and cost thresholds, with two replacement copy versions for each and a safe in-place swap after approval.

Run a creative fatigue check on Meta ad account [account]. Compare the last 28 days with the previous 28 days at ad level using spend, impressions, frequency, link CTR, CPM, results and cost per result.

Flag an ad as fatigued only when all three conditions hold:
- frequency is above 3.0
- link CTR is down at least 25% versus the prior period
- cost per result is up at least 20%

Skip ads that spent less than [minimum spend] in the last 28 days.

Deliver:
1. A table of fatigued ads with these columns: ad name, ad set, days running, frequency, CTR change, cost per result change.
2. For each fatigued ad, read the current creative (headline, primary text, description, CTA, destination link). Write 2 replacement versions that keep the same offer, image or video and link, and change the hook and the first line of the primary text.
3. For the versions I approve, prepare an in-place creative swap so each ad keeps its ID, learning and history.

Render a preview of every new creative first. Wait for my explicit confirmation before applying any change.

Ad copy and creative

Which messaging angles win on Meta: 90-day angle analysis

Pulls 90 days of ad-level Meta results, reads the creative of the ads that spent enough, tags each by angle, hook, format and CTA in a reviewable sheet, and ranks angles by spend-weighted cost per result. Ends with six new copy variations.

Analyze the last 90 days of Meta ads in [account] and tell me which messaging angles actually drive results. Count [result action, e.g. lead, purchase] as the result.

Data to pull:
- Ad-level spend, impressions, results, cost per result and link CTR for the last 90 days.
- For video ads, ThruPlays divided by impressions.
- Include only ads that spent at least [minimum spend].
- If more than [40] ads qualify, read creatives for the [40] with the highest spend and tell me how many you left out and how much spend they cover.
- Read the full creative of each included ad: primary text, headline, format and button.

Tag every ad on four dimensions:
- Angle: price/value, pain point, social proof, outcome, urgency or product feature.
- Hook type of the first line: question, statistic, bold claim or direct callout.
- Format: image, video or carousel.
- CTA.

Angle tagging is a judgement call. Show the tag sheet first: ad ID, ad name, first line of the primary text and the four tags. Mark tags you are unsure about with "?". Then continue with the analysis. If I correct any tags afterwards, recompute.

Deliver:
1. A table per angle with these columns: number of ads, total spend, spend-weighted cost per result, average link CTR.
2. The top 3 and bottom 3 angles, each with the single strongest ad as an example and one sentence on why it likely worked or failed. Flag any angle that rests on fewer than 3 ads as low confidence.
3. Six new copy variations built on the top 2 angles. Each one gets a primary text with the hook in the first 125 characters, a headline under 40 characters and a CTA. Each variation tests a different hook type.

This is analysis and drafting only. Do not create, pause or edit any ads.

Ad copy and creative

Turn converting Google RSA headlines into Meta ad variations

Ranks the RSA headlines and descriptions that convert in a chosen Google campaign, adapts the best into 4 Meta variations that avoid messages already running, creates them paused after approval, renders Feed and Reels previews and removes the ones you reject.

Reuse the Google Ads messages that already convert for [client] on Meta. Google Ads account: [Google Ads account], campaign: [Google campaign]. Meta ad account: [Meta ad account], ad set: [ad set].

1. Pull the RSA asset report for [Google campaign] for the last 90 days. Keep headlines and descriptions with at least [minimum impressions, e.g. 1,000] impressions. Rank them by conversion rate, with CTR as the tiebreaker, and show the top 10 in a table with impressions, clicks, conversions and cost per conversion. If fewer than 5 assets pass the threshold, tell me and suggest a lower one before going on.
2. Read the active ads in [ad set] and their creatives so you do not duplicate messages already running there. Note the Facebook Page, the image and the button they use. If the ad set runs video rather than image ads, stop and ask me how to proceed.
3. Write 4 Meta variations from the top assets. Each one needs primary text, headline and CTA button. Adapt, do not paste: a search headline answers a query someone typed, while Meta copy has to stop a scroll. Open with the problem or outcome, then bring in the proven phrase. Keep every claim identical to what the Google ad says. Add no new discounts or numbers.
4. Use [landing page URL] as the destination, the same Facebook Page, and the image the ad set already uses. Reuse its image hash if the creative returns one, otherwise upload that image once and reuse the hash for all 4.

Show me the 4 variations as a text preview first. Meta can only render a visual preview of a creative that exists, so after I approve the text, create the approved variations as new ads in [ad set], in paused status. Nothing spends while they are paused.

Then render Feed and Reels previews of each new ad and show me the links. For any version I reject after seeing the preview, preview its deletion and delete it only after I confirm. Leave the ones I keep paused. Do not change any existing ad.

Ad copy and creative

Find which TikTok ad texts actually drive ROAS

A read-only analysis of your TikTok ads. It ranks them by ROAS, compares the copy patterns of the best and worst performers, and turns the differences into copy rules backed by numbers.

In my TikTok ad account [advertiser account], analyse which ad texts drive ROAS. This is read-only, so do not change anything.

1. Check the pixels first. If no pixel has an active purchase event (CompletePayment or SHOPPING), stop and tell me, because purchase value and ROAS will be empty.
2. Pull ad-level data for the last 30 days: spend, impressions, CTR, purchases, purchase value and ROAS. If ROAS is not returned directly, calculate it as purchase value divided by spend.
3. Keep the ads that spent at least [minimum spend]. Read each one's current ad text and call to action. For Spark ads the text comes from the organic post caption, so mark those.

Deliver three things:
1) A table sorted by ROAS with these columns: ad name, ad text, CTA, spend, CTR, purchases, ROAS.
2) A pattern comparison between the top quartile and the bottom quartile by ROAS. If fewer than 8 ads qualify, compare the top half with the bottom half instead and say so. Cover:
   - opening words
   - whether a price or discount is stated
   - whether the product is named
   - social proof
   - urgency
   - length in characters
   - question versus statement
3) Five copy rules for new ads. Back each rule with the numbers behind it.

Flag any ad with fewer than 10 purchases as directional only and keep it out of the rules.

If Smart+ campaigns exist, list their ads in a separate table. Include their numbers only if the report returns them for those ad IDs; otherwise list them without metrics. Keep them out of the quartile comparison either way.

Ad copy and creative

Rank LinkedIn corporate gifting ads by messaging angle

Pull 90 days of creative-level LinkedIn performance, read the real copy (including post-based ads), group creatives by messaging angle, and get a scale/test/drop verdict plus an approval-gated pause list.

We use LinkedIn to sell corporate gifting and bulk orders. Work in LinkedIn ad account [ad account]. If you are not sure of the account id, list my connected LinkedIn ad accounts first.

1) Performance: run a LinkedIn performance report with pivot CREATIVE for the last 90 days. For each creative get spend, impressions, clicks, CTR (clicks divided by impressions), conversions and cost per conversion. Count website conversions, and also lead form leads if any campaign is a Lead Gen campaign, and tell me which conversion type you used. Also request conversion value.

2) Conversion value check: list the account's conversion tracking rules and check whether any of them carries a conversion value. If no value is tracked, or the report returns zero value for every creative, write "not tracked" in the value columns and do not estimate it.

3) Real copy: list all creatives in the account to get names, formats, headlines, descriptions and landing pages. For post-based ads (Sponsored Content, Document Ads, Thought Leader Ads) the copy is not in that list, so read the actual ad copy for those creatives by passing their ids in batches of up to 60. If the creative list is flagged as truncated, or copy cannot be retrieved for some ads, tell me which ones and leave them out of the angle analysis.

4) Classify each creative by its main angle from the copy. Use these categories: unit price or volume discount, logo customization, delivery speed, gifting occasion (employees, clients, events), minimum order or ease of ordering, and other. If a creative mixes angles, pick the dominant one and note the secondary one.

Deliver:
1) An angle summary table with these columns: angle, number of creatives, spend, CTR, conversions, cost per conversion, value per conversion (or "not tracked").
2) The best and the worst creative within each angle, with their copy quoted. Flag any angle where the sample is too small to judge (for example under 1,000 impressions or under 3 conversions in total).
3) Which angle to scale, which to test further, and which to drop, with one line of reasoning each.

Decision rule: propose pausing any creative that has spent more than 2x [target cost per conversion] with zero conversions. Show that pause list as a preview (confirm=false) with creative id, campaign, spend and copy, and wait for my approval. Only after I approve, set those creatives to PAUSED. Creatives that are under review cannot be paused, so list them separately instead of trying.

Ad copy and creative

Draft new LinkedIn text ads for wholesale and bulk orders

Confirm the target ad set is a Text Ad campaign, learn from the existing creatives' copy and 60-day results, draft 6 character-checked text ads, create them after approval, then propose pausing the weakest old ads with a second approval.

Work on the LinkedIn text ad set [campaign name] in account [ad account]. If you are not sure of the account id, list my connected LinkedIn ad accounts first.

1) Check the type: list the API-level campaigns in the account (these are "Ad sets" in Campaign Manager) and find [campaign name]. Confirm its campaign_type is TEXT_AD. If it is not (for example Sponsored Content), stop and tell me, because text ads can only be added to a Text Ad ad set. In that case list the TEXT_AD ad sets in this account so I can pick one, or offer to deliver the 6 drafts as copy only without creating anything.

2) Learn from what is there: list the creatives in that ad set and read their headlines, descriptions and landing pages. Then run a performance report with pivot CREATIVE for the last 60 days, narrowed to that ad set, and get impressions, clicks, CTR, conversions and cost per conversion per creative. Summarise in 3 bullets what the best performers have in common. If conversions are zero across the board, say so and judge on CTR instead.

3) Draft 6 new text ads that follow these rules:
- Headline of up to 25 characters and description of up to 75 characters. Count every character exactly, including spaces and punctuation.
- Each ad uses a different angle: volume pricing, custom branding, fast delivery, a low minimum order, a named buyer role (such as HR or procurement), and a proof point.
- No vague claims like "best quality". Use concrete numbers only if I have given them to you or they appear in the existing ads; otherwise leave a [placeholder] for me to fill in.
- Every ad points to [landing page URL].

Show a preview table with these columns: headline, character count, description, character count, angle.

4) After I approve the copy, run each ad as a preview first (confirm=false), then create it. Create the new ads as PAUSED unless I tell you to make them ACTIVE, and tell me how to switch them on.

5) Then propose pausing the existing creatives that have at least 1,000 impressions and the lowest CTR, at most 3 of them. Make sure the ad set keeps at least 2 active ads after the change, and flag it if it would not. Show that list with creative id, impressions, CTR and copy as a preview, and wait for a second approval before pausing anything. Creatives that are under review cannot be paused, so leave them out.

Ad copy and creative

Write proof-based chat cards from your Google reviews

Mines the last 12 months of 4 and 5 star Google reviews for recurring proof points, checks them against the services on the profile, and drafts 3 ChatGPT Ads chat cards built on real customer evidence.

Read the Google reviews for location [business location]. Page through the list with offset until you reach reviews older than 12 months, then filter by date and star rating yourself. Use only 4 and 5 star reviews that include text. Find the recurring proof points, such as arriving on time, a fair or upfront price, a clean job or a named service, and count how many reviews mention each one. Read the profile with gbp_location_details, check every service a review mentions against the profile's service list, and drop anything we do not actually offer.

Find ad group [ChatGPT Ads ad group] in campaign [ChatGPT Ads campaign]. If I did not give an adgrp_ id, list the ad groups with chatgpt_ads_insights (level ad_group, include_zero true) and confirm it with chatgpt_ads_details. Write 3 chat cards for it. Build each card on one proof point and frame it as an answer to someone asking for a reliable local provider. The rules:
- never quote or name a reviewer,
- never invent numbers,
- use the average rating only if it comes straight from the gbp_reviews data, and state the rating and review count exactly as returned.

Every card should push toward a call or a quote request, and the destination should be [landing page URL]. A chat card needs an image. Use [image URL] if I gave one. Otherwise pick a real job photo from the profile with gbp_list_photos, turn it into a public URL with upload_image (source_url), and tell me which photo you chose.

Show me a table with these columns: proof point, number of mentions, card title, card body, image, and why it should work in a chat context. Then show the create_chatgpt_ads_ad calls as a preview (confirm=false). The ads are created paused, so do not pass activate=true. Wait for my approval before creating anything.

Ad copy and creative

Prune losing chat cards and launch one challenger per winner

Checks that a ChatGPT Ads campaign measures leads, scores every chat card against your target cost per lead, pauses the clear losers and drafts one single-variable challenger per winner, all as a preview awaiting approval.

Start by checking conversion tracking for ChatGPT Ads campaign [ChatGPT Ads campaign]. Use chatgpt_ads_conversion_setup with check_recent_events=true to see the pixels, the event settings and which campaigns each setting is attached to, and chatgpt_ads_details on the campaign to confirm the attached conversion events. If no lead-type event (lead_created, appointment_scheduled, registration_completed or a custom lead event) is attached to this campaign, stop and tell me, because judging chat cards on CTR alone is misleading. An empty recent-events list only means nothing arrived in the last ~15 minutes, so mention it but do not treat it as broken tracking on its own.

If tracking is in place, pull the last 30 days with chatgpt_ads_insights: scope_id = the campaign id, level = ad, days = 30, conversions = true, include_zero = true, and no segment (conversions cannot be combined with a segment). For each ad show impressions, clicks, CTR, spend, conversions and cost per conversion (spend divided by conversions). Then read each ad with chatgpt_ads_details to get its ad group, current status, title, body, destination URL and image.

My target cost per lead is [target cost per lead]. Apply these rules only to ads with at least 1,000 impressions:
- Pause an ad if it has spent 2x the target with zero conversions.
- Pause an ad if its cost per lead is more than 1.5x the target with 3 or more conversions.
- Mark an ad as a winner if its cost per lead is at or below the target.
- Never leave an ad group with fewer than 2 active ads. If a pause would break this, keep the ad running and say so.

For each winner, write one challenger that changes only ONE element: the opening line of the body, the offer, or the call-to-action wording. Keep the title or body you are not testing, the destination URL and the image identical to the winner, so we learn what actually moved results. Reuse the winner's image if chatgpt_ads_details returns a usable public image URL. If it does not, use [challenger image URL]; if I attached the image file instead, turn it into a public URL with upload_image. An image is required for a chat card, so do not draft a challenger without one.

Show me a table with these columns: ad, ad group, metrics, verdict and reason. Then show the previews: chatgpt_ads_status with kind = ad, action = pause, confirm = false for each pause, and create_chatgpt_ads_ad with confirm = false for each challenger in the winner's ad group, created paused (no activate). Remind me that every new ad goes through OpenAI review before it can serve. Wait for my approval before calling anything with confirm = true.

Ad copy and creative

Build chat cards and context hints from real local search questions

Turns the problem and question searches behind your Business Profile and Search Console into customer intents, finds the ones your ChatGPT Ads context hints miss, and drafts new hints and chat cards for them.

Pull the searches people used to find Business Profile location [business location] over the last 3 months, including low-volume terms. Google reports these with a delay of about 2 months, and small terms only as "fewer than N", so treat those as estimates. Also pull the Search Console queries for property [Search Console property] over the last 90 days. Keep only problem and question style queries, for example: "how much", "near me", "emergency", "same day", "broken", "leaking", "not working".

Group these queries into 4 to 6 customer intents. Then list the ad groups in ChatGPT Ads campaign [ChatGPT Ads campaign] with chatgpt_ads_insights (scoped to the campaign, level ad_group, include_zero true), and read each one's context hints with chatgpt_ads_details. Page with hints_offset if a group has more hints than one call returns.

Show me a table with these columns: intent, 3 example queries, combined impressions (Business Profile and Search Console listed separately), and whether an existing hint covers the intent (yes/no, naming the ad group).

For each intent that is not covered and has meaningful volume:
- draft context hints as short natural sentences, written in the language our customers search in, that describe the conversation in which our ad should appear, and say which ad group they belong in,
- write one chat card (title and body) that answers the question directly, names the service area and asks for a call or a quote, with [landing page URL] as the destination.

A chat card needs an image. Use [image URL] if I gave one, otherwise pick a fitting real job photo with gbp_list_photos and turn it into a public URL with upload_image (source_url).

Do not promise prices or response times unless they already appear in the Business Profile data from gbp_location_details.

Do not change bids or budgets. Add hints with hints_mode add, so existing hints stay. Show the update_chatgpt_ads_ad_group and create_chatgpt_ads_ad calls as a preview (confirm=false). New ads are created paused. Wait for my approval.

Ad copy and creative

Replace stock or missing chat card images with your real job photos

Finds the ChatGPT Ads cards with weak images, shortlists real team and job photos from your Business Profile and WordPress media library, and swaps only the image after your approval.

List every ad in ChatGPT Ads campaign [ChatGPT Ads campaign] with chatgpt_ads_insights (scoped to the campaign, level ad, include_zero true). For each ad, read the creative with chatgpt_ads_details and render it with chatgpt_ads_ad_preview. Flag each card that has no image, a generic stock-style image, or an image that crops badly in the card.

Then collect real photos from two sources: Business Profile location [business location] with gbp_list_photos, and the media library of WordPress site [WordPress site] with wordpress_media. Use file names, alt text, Business Profile categories (TEAM, AT_WORK, EXTERIOR and so on) and dimensions to shortlist photos that show our team, our vehicle, our tools or a finished job. Open a WordPress candidate with media_id to get its real size, and use return_image to look at it when you can. Skip these:
- low-resolution files,
- photos with text overlays,
- collages,
- anything that shows a customer's face or a house number.

If you cannot actually see a photo, do not assume it is clean. Mark it "needs my visual check" in the table, so I can look at it before I approve.

For each flagged card, pick the photo that best fits the service the card promotes. Explain each choice in one line. Show me a table with these columns: card, current image problem, chosen photo (source and URL), size, visual check (done by you / needs my check), and reason.

Convert the chosen photos into public image URLs with upload_image (source_url). Then prepare an update_chatgpt_ads_ad call for each card that changes the image only. Do not touch the title, body or destination. Remind me that a new image starts a new OpenAI review and the ad stops serving until it passes. Show the updates as a preview (confirm=false) and wait for my approval. After the changes are applied, render the cards again and report their review status so I can check them.

Ad copy and creative

Competitor messaging matrix with a management summary

Pulls competitors' live ads from the public Google Ads Transparency Center and, if a Meta account is connected, the Meta Ad Library, maps their angles against your enabled Google ads and hands you a matrix, table stakes, open angles and a five-bullet summary.

Use competitor_ads with platform = google and region = [region, e.g. TR] to pull the live ads these competitors are running now: [competitor domain 1], [competitor domain 2], [competitor domain 3]. Then try platform = meta for the same competitors, searching by brand or Facebook Page name rather than domain: [competitor brand 1], [competitor brand 2], [competitor brand 3]. Meta results are optional: they need a connected Meta account and the Ad Library matches page names, not domains. If Meta is not connected or returns nothing, say so once and continue with Google only.

Classify every ad that comes back on these points:
- Main angle: price, speed, quality, trust, guarantee or urgency.
- The specific offer.
- The proof used: numbers, reviews or awards.
- The CTA.
- The format, if the library result shows it.

Then read our own ads. Use list_accounts to find the customer id for [our Google Ads account], list_entities with entity = ads to find the enabled ads and their ad groups, and get_ad_details by ad_group_id to read the full headlines and descriptions. Mark which of these angles our ads already use.

Deliver:
1) A matrix table with competitors as rows and angles as columns, showing the ad count in each cell, plus a row for us. Note per competitor whether the counts come from Google, Meta or both.
2) The messages that appear in at least two competitors. These are the category's table stakes.
3) Angles that no competitor uses and that we could credibly own.
4) A five-bullet summary I can send to management as written.

The libraries only expose what is publicly live, so if a library returns nothing for a competitor, say so and do not fill the gap with guesses. This is a read-only analysis, so make no changes.

Ad copy and creative

Refresh Microsoft Ads RSAs with competitor and search term insight

Find the weak Microsoft Ads RSAs in one campaign, compare their CTR with a real Google Ads benchmark, check them against converting queries and competitor claims, and preview rewritten copy that keeps each ad's history.

Management wants to know why our Microsoft Ads CTR trails our Google Ads CTR. Work in Microsoft campaign [Microsoft campaign name].

1. Find the campaign with list_microsoft_campaigns, its ad groups with list_microsoft_ad_groups, and read every ad in those ad groups with list_microsoft_ads. Keep the ad_group_id and ad_id for each ad.

2. Run microsoft_entity_performance_report with level=ad for the last 60 days and keep only the rows for this campaign. For each ad give impressions, clicks, CTR, conversions and cost per conversion. Calculate the campaign average CTR from those rows. Mark an ad as weak if it has at least 1,000 impressions and a CTR below that average.

3. Get the Google benchmark. Run performance_report with level=campaign and days=60 on the Google Ads account and pull the row for [comparable Google campaign name]. Calculate its CTR from clicks and impressions and put it next to the Microsoft campaign CTR. If no comparable Google campaign exists, use the Google account total and say so.

4. Run microsoft_search_terms_report for the same 60 days. Keep the queries from this campaign if the rows show the campaign, sort by conversions and take the top 30. Check whether our headlines use the wording people actually type in those queries, and list the phrases that appear in converting queries but nowhere in our copy.

5. Run competitor_ads with platform=google for [competitor domain 1] and [competitor domain 2]. These are their Google Ads Transparency Center ads; Microsoft has no public ad library this tool can read, so treat them as a proxy for their Bing messaging, not as confirmed Bing copy. List the claims, offers and proof points they lead with.

Deliver:
1) A table of our weak ads with impressions, CTR, conversions, the query wording they miss and the competitor claims we lack.
2) For each weak ad, replacement headlines (30 characters or fewer) and descriptions (90 characters or fewer), keeping Microsoft's limits of 3 to 15 headlines and 2 to 4 descriptions per ad.
3) Two sentences for management explaining the CTR gap, using the real Microsoft and Google CTR figures. Be clear about what the data shows and what is an assumption (for example, that competitors run the same copy on Bing).

Show each change as an update_microsoft_responsive_search_ad preview with confirm=false, so the ads are edited in place and keep their history instead of being rebuilt. Wait for my approval before applying anything with confirm=true.

Ad copy and creative

Write PMax headlines from the words customers actually search

Turns real Search Console queries for a landing page into themed PMax headlines and long headlines, proposes which weak assets to replace, and applies only approved lines.

My Performance Max headlines sound like me, not like my customers. Fix that using real search language.

1. Find my property with gsc_list_sites. With gsc_performance, pull queries for the last 90 days filtered to the page [landing page URL], sorted by clicks. Exclude brand queries, meaning anything containing [brand terms], and keep the top 50 that remain.
2. Group them into 4 to 6 themes, such as price, location, urgency, comparison or a specific product. For each theme, show total clicks and impressions.
3. Find my Google Ads account with list_accounts, look up the asset group [asset group name] with google_ads_query, and read its current text assets with get_pmax_assets. Also run a google_ads_query on asset_group_asset for that asset group to get each text asset's performance_label. If that field comes back empty, say so and rank weak assets by your own judgement of how generic or repetitive they are instead.
4. Write new copy for each theme:
   - 2 headlines (30 characters max) and 1 long headline (90 characters max)
   - Reuse the customers' exact wording where it reads naturally.
   - Do not overlap with existing assets.
   - Include character counts.
   - No keyword stuffing, no all caps, no unverifiable claims.
5. Suggest which existing assets these should replace. Prioritise any asset labelled "Low", respect the limits of 15 headlines and 5 long headlines, and never drop below 3 headlines or 1 long headline.

Show the result as a proposed add/remove list, with the resource_name of each asset to remove. Make no changes until I confirm. Then run update_pmax_assets with confirm=false for only the approved lines, show me the preview, and apply it with confirm=true only after I say go. Finally, re-read the asset group and show the updated text assets.

Ad copy and creative

Check that my PMax landing page matches the assets and loads fast

Compares a Performance Max asset group's copy with its landing page, checks mobile speed and conversion tags, adds asset group performance, and gives a keep, fix or change verdict. Any final URL or path change is previewed and waits for approval.

Check whether the landing page behind my Performance Max asset group [asset group name] in campaign [campaign name] supports the ads or works against them.

1. Find the campaign and asset group IDs. Read the asset group's final URL, all of its headlines, long headlines and descriptions, and its ad strength. Read its display paths (path1, path2) with a GAQL query on asset_group if they are not in the asset group read.
2. Pull the asset group's performance for the last 30 days with a GAQL query on asset_group: cost, conversions, cost per conversion and conversion rate (conversions divided by clicks). Show the campaign's figures next to it for context.
3. Read the landing page as a visitor sees it: title, H1 and heading structure, plus the page text.
4. Run an SEO and speed audit on that URL on mobile, and pull its Core Web Vitals on mobile. Note which measurement tags the audit finds on the page (Google Ads conversion, GTM, GA4). Also run a conversion tracking health check on the account so a tag that is present but not recording shows up.
5. Score the match between the ads and the page:
   - Does every offer, price or promise in the assets appear on the page? Treat the title, H1 and the first block of text as the top of the page and say whether each claim appears there or only further down.
   - Is the main call to action on the page the same as the one the ads lead to?
   List each mismatch with the exact asset text and what the page says instead.
6. Flag these problems: mobile LCP over 2.5 seconds, CLS over 0.1, INP over 200 ms, no Google Ads conversion or GTM tag found on the page, or a primary conversion action recording zero.

Then give a verdict using this rule:
- Keep the page if it matches and passes the speed and tag checks.
- Fix the page if it matches but is slow or missing tags. List the top 3 fixes from the audit.
- Change the final URL if the page does not match and one of these candidate pages fits better: [candidate URLs, or leave empty]. Run the same match and speed checks on each candidate before recommending it. If I gave no candidates, tell me what the right page should contain instead of guessing a URL.

If you recommend a different final URL or new display paths, show the exact before and after as a preview. Do not apply it until I approve.

Ad copy and creative

Client onboarding: asset coverage matrix for every Search campaign

A read-only matrix showing which campaigns lack sitelinks, callouts or snippets, ranked by spend.

Ad copy and creative

Find and fix broken or redirected sitelink destinations

Fetches every sitelink URL in enabled campaigns, follows redirects hop by hop, checks index status and Google's canonical in Search Console, flags broken or mismatched links and edits them in place after a preview.

In [client account], list the enabled campaigns, then read the extension assets for all campaigns in one call and keep the sitelinks of the enabled ones: campaign, link text, description lines, final URL and asset id. Deduplicate URLs so each one is checked only once.

Fetch each URL live with the page outline and record:
- HTTP status
- if the status is 3xx, the Location header. Fetch that target too and keep following until you reach a non-redirect status or 5 hops. Record the final URL and the hop count.
- the page title of the final page

Then confirm the Search Console property [property] is in my verified list and inspect the final URLs there, up to 20 per call. Note whether each one is indexed and which canonical Google chose. A noindex page or a canonical that points elsewhere usually means the sitelink lands on a page the site no longer treats as the main version.

Flag any of these issues:
- 4xx or 5xx status
- redirect that ends on the homepage
- redirect chain longer than one hop
- not indexed or noindex
- Google's canonical differs from the final URL
- a page title that clearly does not match the sitelink text

For each problem sitelink, suggest the correct live URL on the same site. Look for it in the Search Console pages report for the last 90 days, and fetch it to confirm it returns 200 with a matching title before you suggest it. If the topic changed, adjust the link text (max 25 characters) and the description lines (max 35 characters each, both or neither) and show the character counts. If no suitable page exists, say so and recommend removing the sitelink instead.

Show the edits as a preview and apply them only after I confirm. Edit the sitelinks in place rather than recreating them.

Ad copy and creative

Write differentiating callouts from competitor ads and customer reviews

Reads competitors' live ads and the client's Google reviews, finds proof points competitors do not claim, writes 8 compliant callouts and swaps out only the weaker existing ones after a preview.

Write callouts for [campaign name] in [client account] that say something competitors do not.

1. Fetch the live Google ads of [competitor domains] in [region, e.g. TR] and list the claims they repeat, such as free shipping, 24/7 support or years in business, with how many of the competitors use each one. If no ads come back for a domain, say so rather than guessing what it claims.
2. Find the client's Business Profile location [business name or location]. Read its Google reviews and keep the ones from the last 12 months by review date, paging until you reach older ones. Extract the specific points customers praise again and again, with a mention count and one short quote for each. Count a mention only when the review clearly says it.
3. Read the campaign's current callouts and their performance over the last 90 days: impressions, clicks and CTR (clicks divided by impressions). Mark any callout with fewer than [minimum impressions, e.g. 500] impressions as too little data to judge.

Propose 8 new callouts, each 25 characters max with the count shown. Each one must:
- be backed by at least 3 review mentions in the last 12 months or a verifiable fact I confirm: [facts, e.g. founding year, warranty length, delivery time]
- not duplicate a claim that competitors repeat
- avoid unproven superlatives such as "best" or "#1", which can be disapproved

Output a table with these columns: callout | character count | evidence | used by competitors (yes/no) | replaces (which existing callout).

Recommend replacing only the existing callouts in the bottom half by CTR, and keep any callout with too little data. Show the additions and removals as a preview and wait for my approval. After I approve, add the new callouts first, then remove the replaced ones, so the campaign is never left without callouts.

Ad copy and creative

Rank RSA headlines by conversion value and replace the weakest ones

A per-ad-group ranking of RSA headlines and descriptions by value per cost, with replacement copy for the losers previewed before any edit.

In Google Ads account [account], pull the RSA asset report for campaign [campaign name] for the last 60 days.

For every headline and description with at least 1,000 impressions, build a table per ad group with these columns: asset text, pinned position, impressions, CTR, conversions, conversion value, cost, and conversion value / cost. Asset-level numbers come from combinations, so treat them as directional, not causal.

Decision rule: mark an asset "replace" only if both of these are true:
- its value/cost is below 50% of the ad group average, and
- its CTR is below the ad group median.

Never mark the last remaining brand headline or an asset that has more than 10% of the ad group's conversions.

For each "replace" asset, write 2 alternatives in the same role. A price or offer headline stays an offer headline, and a product-type headline stays a product-type headline. Headlines must be 30 characters or fewer and descriptions 90 or fewer. Show the character count for each.

Do not change pinning without asking me.

Finally, show a preview of the update_rsa edits ad by ad, keeping the same ads so their history is preserved. Wait for my approval before applying anything.

Ad copy and creative

Lift Poor and Average Ad Strength without deleting any RSA

A per-ad fix list that brings weak RSAs to Good or Excellent Ad Strength while protecting the headlines that already sell.

List every enabled responsive search ad in [campaign name] whose Ad Strength is Poor or Average. Include Google's own action items for each one.

For each ad, diagnose the cause:
- number of headlines (target 15) and descriptions (target 4)
- near-duplicate headlines, meaning the same idea reworded
- over-pinning
- missing angles that matter for e-commerce: product category, price or offer, free shipping or returns, delivery speed, brand, social proof

Protection rule: do not remove or rewrite any asset that drove 2 or more conversions in the last 30 days.

Fill the gaps with new headlines and descriptions written in our tone. Use only claims that appear on our site or that I confirm. Keep headlines to 30 characters and descriptions to 90, and show the character counts.

Output one table per ad with three columns: current problem, proposed change, expected Ad Strength effect.

Then preview all update_rsa edits so the existing ads keep their history. Do not create new ads. Wait for my approval before applying.

Ad copy and creative

Find gaps in competitor ad claims and write differentiated RSA copy

Maps every claim two competitors make in their live search ads against your own RSAs, flags where you stay silent or undersell a stronger offer, and writes verified, differentiated headlines and descriptions to swap in.

Competitors: [competitor domain 1] and [competitor domain 2]. Our account: [customer ID], campaign [campaign name], product categories [our product categories], market [location] / [language].

1. Fetch each competitor's live Google ads from the Ads Transparency Center.
2. Check the SEO history first so we do not pay twice, then pull each competitor's paid keywords and ad copy for our market (this uses SEO add-on credits). Keep the keywords that overlap with our product categories. If a domain returns no ads in this scan, say so and work with what was found.
3. Extract every claim they make: discount percentages, free shipping thresholds, delivery time, returns window, installment options, warranty, assortment size and price anchors. Note that public ad libraries do not show every ad, so the matrix reflects only what was found.
4. List the campaign's ad groups and read our RSAs. Build a matrix: rows are the claims, columns are us, competitor 1, competitor 2. Highlight two kinds of gap:
- where we say nothing and they do
- where we may have a stronger offer but do not say it

Before writing copy, list the claims you plan to use about us and ask me to confirm each one and give the real figures, because I will not run unverified claims.

After I confirm, write 6 headlines (30 characters or fewer) and 2 descriptions (90 or fewer) that lead with our real advantages. Do not name competitors.

Run the RSA asset report for the campaign and, for each ad you want to change, name exactly which low-conversion-value headlines and descriptions each new asset replaces (one new asset per removed asset, staying within 15 headlines and 4 descriptions per ad). Show one update_rsa preview per ad with confirm=false and apply only after my approval.

Ad copy and creative

Fix disapproved or limited RSA assets while keeping the offer

The exact headlines and descriptions behind each policy issue, with compliant rewrites that keep the selling point, previewed before editing.

In Google Ads account [account], list every responsive search ad that is disapproved or approved (limited). Include the policy topic and Google's reason for each.

For each one, identify the specific headline or description that most likely triggers the issue. Common causes:
- excessive capitalization or punctuation
- unverifiable superlatives such as "cheapest" or "best"
- trademarked terms
- symbols used as words
- phone numbers in text
- price or discount claims that do not match the landing page

Return a table with these columns: ad, ad group, policy topic, offending asset, why it violates, and that asset's conversions over the last 30 days.

Rewrite each offending asset so it is compliant but keeps the same selling point. For example, turn "cheapest" into a concrete price or a price-match statement, if I confirm one exists. Keep the 30 and 90 character limits.

Prioritise the ads with the highest spend over the last 30 days.

Show the update_rsa preview for each ad. Wait for my approval, then apply the edits and tell me to recheck the policy status in 24 hours.

Ad copy and creative

Audit RSA claims against the live landing page before they cost ROAS

Checks every factual claim in your 20 highest-spend RSAs against what the live landing page actually says, flags broken URLs and overpromises by severity, and drafts corrected copy for approval.

In Google Ads account [customer ID], run an ad-level performance report for the last 30 days and take the 20 responsive search ads with the highest cost. Read each ad's headlines, descriptions and final URL.

Open every unique final URL as a visitor would, record the HTTP status, and read the page text. Then check each factual claim in the ad copy against the page:
- price or "from" price
- discount percentage
- free shipping threshold
- delivery time
- returns window
- installment offer
- stock or "new season" claims

Some prices and offers are loaded by JavaScript and will not appear in the served HTML. When a claim is missing from the page text, mark it "not found in served HTML" rather than mismatch, and list those for me to check in a browser.

Return a table with these columns: ad, final URL, claim in ad, what the page says, status (match / mismatch / not found in served HTML), severity.

Severity is high if the ad promises more than the page delivers or the URL does not return 200. It is medium if the page has a better offer the ad does not mention, and low otherwise.

For high and medium rows, write corrected headlines or descriptions within the 30 and 90 character limits. For URLs that do not return 200, do not edit the ad; tell me which ad is affected and what to fix.

Show the update_rsa preview (confirm=false) for each ad you would change and wait for my approval before changing anything.

Ad copy and creative

Find fatigued Meta lead ads and draft replacement creatives

A table of tired Meta ads, judged on frequency, CTR and cost per lead trends, with two ready-to-review replacement variations for each.

Check my Meta ad account [ad account] for creative fatigue on lead and call campaigns.

1. Pull ad-level data for the last 60 days, split into days 1-30 and days 31-60: spend, impressions, frequency, link CTR, leads, cost per lead.
2. Mark an ad as fatigued if frequency is above 3.5 AND link CTR fell 25% or more versus the first half, OR if cost per lead rose 30% or more on at least 10 leads. Mark an ad as "too early" if it has fewer than 1,000 impressions in the recent half.
3. Give me a table: ad name, ad set, both periods side by side, % change, verdict (fatigued / watch / healthy / too early).
4. For each fatigued ad, read the current creative and write 2 replacements. Keep the same offer and service, but change the hook (first 125 characters of primary text), the headline (max 40 characters) and the visual direction (describe the image or video in one line).
5. Render previews where possible.

Do not change anything yet. List the creative swaps you would make with update_meta_ad_creative, so each ad keeps its ID and history, and wait for my approval on each one.

Ad copy and creative

Rewrite Meta ads that bring cheap but unqualified leads

Cost per qualified lead by ad, based on real form answers, plus pre-qualifying copy for the ads that attract the wrong people.

My Meta lead ads bring volume, but too many leads are outside my service area or not ready to book. I want to judge creatives on qualified leads, not raw cost per lead.

1. Download instant form leads from the last 60 days with the ad, ad set and campaign each came from.
2. Count a lead as qualified if it meets [qualifying rule, e.g. postcode inside service area and wants the job within 2 weeks]. Also flag obvious junk (fake names, invalid phone numbers).
3. Pull spend per ad for the same period.
4. Table per ad: spend, leads, qualified leads, qualified %, cost per lead, cost per qualified lead. Only judge ads with at least 15 leads.
5. Read the creatives of the 3 ads with the lowest qualified %, compare them with the best one and name the copy traits behind the difference (vague offer, "free" wording, no area, no starting price).
6. Rewrite those 3 ads to pre-qualify: state the service area, a starting price or typical range, and who the service is for. Accept a slightly higher cost per lead.

Show previews, then list the creative swaps with update_meta_ad_creative and wait for my approval before applying any of them.

Ad copy and creative

Find which LinkedIn message angles win, with a summary for management

A per-ad and per-angle performance table of your LinkedIn ad copy, the top and bottom ads explained, and a 5-line summary for management.

Ad copy and creative

Rewrite weak LinkedIn Text Ads with new angle-based variants

Underperforming LinkedIn Text Ads identified, four new character-checked variants drafted, and a change list to create and pause ads after approval.

Review the Text Ads in LinkedIn campaign [campaign name] for the last 60 days. List every creative with its headline, description, status, impressions, clicks, CTR and conversions.

Mark an active ad as an underperformer if both of these are true:
- It has more than 10,000 impressions.
- Its CTR is at least 30% below the campaign average.

Then write 4 new Text Ad variants for [target audience, e.g. HR directors at mid-size companies]:
- Headline: maximum 25 characters.
- Description: maximum 75 characters.
- Each variant uses a different angle: cost saving, speed, risk reduction, proof.
- No claims I cannot back up. Use only facts from my current ads.

Every variant points to [landing page URL]. Show the variants in a table with character counts and the angle used.

Once I approve the copy, prepare two steps: create the new ads, then pause the underperformers. Show me the complete list of changes and wait for my confirmation before doing anything. Keep at least 2 ads active in the campaign at all times.

Ad copy and creative

Audit TikTok ad text and landing page message match

Reads every active TikTok ad and its live landing page, flags broken URLs, missing offers, CTA mismatches and missing UTMs, and prepares in-place ad fixes for approval.

Check message match between my TikTok ads and their landing pages in ad account [advertiser account]. Cover every active ad with spend in the last 14 days.

For each ad, read its ad text, CTA and landing page URL. Then read the live landing page: get the HTTP status, page title and H1 from the page outline, and the opening copy of the page text as the closest stand-in for the first screen.

Some pages render their content with JavaScript or block automated readers. If a page comes back empty, blocked or with no readable copy, mark it "could not read" and do not judge its offer or CTA fit; I will check those by hand.

Flag these issues:
- The URL is broken or redirects (status other than 200).
- The offer, price or discount in the ad is not on the page.
- The CTA does not fit the page, for example "Shop now" leading to a blog post.
- UTM parameters are missing.

Deliver a table: ad, ad text, CTA, URL, HTTP status, page H1, issue, and severity (high = losing money now, medium, low). Sort it by severity, then by spend.

For each ad-side issue, propose a fix: new ad text (maximum 100 characters), a different CTA or a corrected URL. On Spark Ads (built on an organic post) only the CTA and landing page can be changed; for a text issue there, tell me to edit the post instead. Prepare the fixes as in-place ad updates with before/after previews (confirm=false), and wait for my approval before applying any of them. List page-side issues separately; I will fix those on the website myself.

Ad copy and creative

Raise Ad Strength from Poor or Average with fact-based copy

Every low Ad Strength RSA gets a duplicate check and the missing headlines and descriptions drafted from your real business facts.

Ad copy and creative

Turn converting search terms into new RSA headlines

Converting search terms grouped by theme, with headlines written in the customer's own wording and new RSAs drafted where coverage is missing.

Ad copy and creative

Copy your best Google Ads headlines into Microsoft Ads RSAs

Proven Google headlines and descriptions are swapped into weaker Microsoft Ads RSAs in place, with a before and after preview.

My Google Ads campaigns have far more data than my Microsoft Ads (Bing) campaigns. I want to use the Google winners on Bing.

Step 1: Pick the Google winners.
- Use the last 90 days of RSA asset performance in Google campaign [Google campaign].
- Pick the top 5 headlines and top 2 descriptions by conversions, counting only assets with at least 500 impressions.
- If total conversions are under 5, rank by CTR instead and tell me you did.

Step 2: Review the Microsoft side.
- List the RSAs in Microsoft campaign [Microsoft campaign] with their current headlines and descriptions.
- Show each ad's last 90 days of clicks, CTR and conversions.

Step 3: Propose the swaps.
- For each Microsoft ad, recommend which headlines and descriptions to replace with the Google winners.
- Keep the Microsoft ad's own top performers.
- Respect the limits of 30 characters per headline and 90 per description.

Edit the existing Microsoft ads in place so their history is not reset. Show the before and after copy for each ad as a preview and wait for my approval.

Ad copy and creative

Write Meta ad copy from real Google review language

Recurring themes and objections from the client's Google reviews, turned into five Meta ad variations grounded in authentic customer language.

Use real customer language from the Google reviews of [business location] to write Meta ad copy for [Meta account].

1. Pull the reviews from the last 12 months.
   - From 4 and 5 star reviews, extract the outcomes customers describe and the specific details they praise (speed, staff, price, results).
   - From 1 to 3 star reviews, extract the objections and doubts people had.
2. Read the current active Meta ads so the new copy is clearly different from what is running.

Deliver:
1. A table of the top 8 themes with these columns: theme, number of reviews mentioning it, one short verbatim fragment.
2. Five ad variations, each built around one theme. Each variation gets primary text, a headline under 40 characters and a CTA.
   - At least 2 variations should answer an objection.
   - At least 2 should lead with a customer outcome.

Rules:
- Quote reviewers only as short anonymous fragments, never by name.
- Do not invent numbers, ratings or results that the reviews do not support.

For the 2 variations I choose, render previews for Feed and Stories using the image of the current top ad. Do not publish or change anything.

Ad copy and creative

Adapt winning TikTok ads into Meta Reels and Stories ads

Shortlists the TikTok ads that held attention best, writes Meta-ready copy and production notes, moves the approved videos into Meta and builds paused Reels and Stories ads with previews.

Find the TikTok ads for [client] that earned the most attention in the last 60 days and adapt them for Meta Reels and Stories.

1. Pull TikTok ad-level data for the last 60 days: spend, impressions, clicks, CTR, video plays, 2-second and 6-second views, average watch time, conversions and cost per conversion.
   - Calculate the 2-second and 6-second view rates by dividing those views by impressions.
   - Calculate the account median for the 6-second view rate and for CTR across ads that spent at least [minimum spend].
   - Shortlist the ads that spent at least [minimum spend] and sit above the median on both.
2. For each shortlisted ad, read its ad text, CTA and creative source.
   - If it runs on an uploaded video, match the video_id in the video library and note the duration and preview link.
   - If it is a Spark ad built from an organic post, mark it as such. Its video file is not in the library.

Deliver:
1. A shortlist table with the metrics above, the median benchmarks and the creative source.
2. A Meta adaptation for each ad: primary text, headline, CTA and production notes. The notes should cover:
   - TikTok-native elements to remove or replace, such as platform UI references and trending sounds that may not be licensed for Meta
   - captions for sound-off viewing
   - text kept inside the Reels and Stories safe zones

For the videos I approve:
- Get each video into the Meta ad account. Try the TikTok library preview link as the video URL first. If Meta cannot fetch it, or the ad is a Spark ad, give me a one-time drop link so I can upload the original file myself.
- Use the Page assigned to this Meta ad account and ad set [ad set].
- Show me each ad as a preview first (confirm=false). Create it as PAUSED only after I confirm.
- Once an ad exists, render its Instagram Reels and Instagram Story previews so I can check framing and safe zones. Nothing gets switched on.

Ad copy and creative

Write a TikTok DM welcome chat that answers pre-purchase questions

A greeting and tappable question-and-answer pairs for TikTok Direct Message ads, built to move chats toward checkout.

Ad copy and creative

Adapt your best Meta ad copy into TikTok-native ad texts

Finds your highest-ROAS Meta ads, rewrites their copy in a casual creator voice for TikTok, and swaps it into your weakest editable TikTok ads only after you approve a preview.

In Meta ad account [Meta ad account], use meta_insights_query at ad level for the last 60 days. Request these fields: ad_id, ad_name, spend, actions, action_values and purchase_roas. Count purchases from the purchase action in actions and keep only ads with at least 15 purchases. Rank the remaining ads by purchase ROAS. If purchase_roas is empty for an ad, calculate ROAS as purchase value from action_values divided by spend. Take the top 5. For each one, read the primary text, headline and CTA with get_meta_ad_creative. If fewer than 5 ads reach 15 purchases, tell me how many qualified and work with those.

Rewrite each one for TikTok:
- Ad text of up to 100 characters, with no emojis or hashtags.
- Spoken and casual in tone, the way a creator would say it. Not a polished brand voice.
- Keep the exact offer and the core benefit that made the Meta ad work. Drop anything that relies on a long caption.
- Add a suggested on-screen hook for the first 2 seconds of the video, as text only. This is a suggestion for the next edit of the video. It is not applied to the ad.

Show a side-by-side table with these columns: Meta ad, Meta ROAS, purchases, original primary text, TikTok text, character count, hook.

Next, in TikTok ad account [TikTok advertiser ID], list the ads in ad group [TikTok ad group] with list_tiktok_ads. Then pull their last 30 days with tiktok_report_query at data_level AUCTION_AD with dimension ad_id. Request spend, conversion and cost_per_conversion, plus purchase value or ROAS metrics if the account tracks them (for example complete_payment_roas). If no purchase value is reported, say so and rank the ads by cost per conversion instead of ROAS.

Important: Spark ads are built from an organic TikTok post. You can spot them in list_tiktok_ads because they carry a tiktok_item_id. Their text comes from the post, so update_tiktok_ad cannot change it. Mark Spark ads in the list and leave them out of the rewrite. Pick the 3 weakest non-Spark ads and propose which new text each should get. If fewer than 3 non-Spark ads exist, tell me.

Show the changes as a preview first, using update_tiktok_ad with confirm=false and only ad_text filled in. Each ad keeps its video, identity and history. Apply with confirm=true only after I approve.

Ad copy and creative

Turn your best Google Ads headlines into ChatGPT Ads chat cards

Picks your proven RSA headlines and descriptions by conversions and turns them into 3 conversational ChatGPT Ads chat cards with an image and landing page, previewed before anything is created.

In Google Ads, find the ID of campaign [Google Ads campaign] with list_entities if you do not already have it. Then run rsa_asset_performance for that campaign with days=90 and min_impressions=500, so headlines and descriptions under 500 impressions are left out. If any rows under 500 impressions still come back, drop them yourself. Rank the rest by conversions, then by conversion rate. Use get_ad_details on the campaign's ad groups to read the final URL the top assets point to.

Run chatgpt_ads_details on ad group [ChatGPT Ads ad group] to confirm it exists, which campaign it belongs to and what its context hints say. The cards should fit the conversations those hints describe.

Take the top 5 headlines and top 3 descriptions and use them as raw material for ChatGPT Ads chat cards in that ad group. A chat card is read inside a conversation, so the copy should answer a question like "who can fix this near me today?" and should not read like a list of keywords. Write 3 card variants, each with a title and a body. Give each variant one clear angle: speed, price transparency or trust. Each variant needs a call-to-action that asks for a call or a quote and names the service area [city]. Only use claims that already appear in the source assets. Keep the title short and the body to one or two sentences. If the preview or the API rejects a field for length, shorten it and show me the new version.

A chat card cannot be created without an image. Use [image URL] as the card image. If I attach a photo instead, turn it into a public URL with upload_image first. If I give you neither, stop and ask me for one before you build the preview. Use [landing page URL] as target_url, or the final URL from the source RSA if I leave that blank.

Show me a table with these columns: variant, angle, title, body, source asset(s), and the Google conversions of those assets. Then run create_chatgpt_ads_ad with confirm=false for each variant and show me the dry runs. The ads are created paused. Creating an ad sends its title, body, image and destination to OpenAI review, so wait for my approval before you call it with confirm=true.

Ad copy and creative

Audit how your chat cards render and fix delivery blockers

Lists every ad in a ChatGPT Ads campaign, renders each card, reads OpenAI's delivery blockers, diagnoses zero-impression ads and applies approved copy, destination and image fixes.

List every ad in ChatGPT Ads campaign [ChatGPT Ads campaign] with chatgpt_ads_insights: scope it to the campaign, level ad, last 7 days, include_zero true. This gives every ad id together with its 7-day impressions. For each ad, read the full creative with chatgpt_ads_details, render it with chatgpt_ads_ad_preview and note the serving issues and review status OpenAI reports. If any ad has zero impressions, run diagnose_chatgpt_ads_delivery for the campaign with check_landing_pages true.

Build one table with these columns: ad, status, review status, impressions in the last 7 days, reported blockers, rendering issues, and the proposed fix. Rendering issues include truncated text, a missing or poorly cropped image, an unclear call-to-action, or a card that does not say which service or area we cover.

Sort the fixes into three groups:
1) Copy fixes: draft the replacement title and body.
2) Destination fixes: propose the correct URL, using the landing page findings from the diagnosis.
3) Image fixes: say what kind of image is needed. If I give you a photo ([image URL or attached photo]), turn it into a public URL with upload_image and include it in the update. Otherwise leave the image as it is.

If the diagnosis points to bids, budget, targeting, schedule or account spend limits, list those separately and do not change them.

Before I approve, remind me that editing the title, body, destination or image starts a new OpenAI review, so the ad stops serving until that review passes. Show the update_chatgpt_ads_ad changes as a preview (confirm=false) with the current and proposed values side by side, and wait for my approval. After the changes are applied, render the previews again and report the new review status, so I can confirm that the cards look right.

Ad copy and creative

Check message match between chat cards and their landing pages

You get a 1-5 message-match score for every chat card and its destination, with proposed copy or URL fixes.

For every active chat card in ChatGPT Ads campaign [ChatGPT Ads campaign], read the card text and its destination URL.

Then open each destination the way a visitor would and check the following:
- the HTTP status code and any redirects,
- the H1 heading,
- whether the service and area named in the card appear on the page,
- whether the phone number is a clickable tel: link,
- whether a quote or contact form exists on the page.

Run a page audit on each unique destination and note its mobile speed problems.

Score the message match from 1 to 5. A 5 means that someone who clicked the card finds the exact service, the same offer and an easy way to call within seconds.

Show me a table with these columns: card, destination, score, gaps found, and the recommended fix. For each card scoring 3 or lower, propose either a better page that already exists on [website domain] or a rewrite of the card so it matches the page as it is now. Only change the ads. Do not edit the website. List any site issues as tasks for me instead.

Show the update_chatgpt_ads_ad changes as a preview and wait for my approval.

Ad copy and creative

Adapt your best Meta lead ad angles into ChatGPT Ads chat cards

You get chat cards rewritten from your lowest cost-per-lead Meta ads, adapted from feed style to conversation style.

In Meta ad account [Meta ad account], pull ad-level results for the last 60 days from our lead and call campaigns. Keep only ads with at least 10 leads. Rank them by cost per lead and take the top 3.

Read the creative of those 3 ads: primary text, headline, description, button and destination. For each ad, name the angle, the offer and the objection it removes.

Then adapt each winning angle into one ChatGPT Ads chat card for ad group [ChatGPT Ads ad group]. A Meta ad interrupts a feed, while a chat card answers someone who is already asking. So:
- drop scroll-stopping hooks, emojis and urgency gimmicks,
- lead with the answer,
- keep the offer exactly as it appears in the Meta ad,
- end with a call or quote request.

Point every card to [landing page URL]. Also flag anything that does not carry over to ChatGPT Ads, such as instant forms or WhatsApp buttons.

Show me a table with these columns: Meta ad, cost per lead, angle, chat card text, and what changed and why. Then show the create_chatgpt_ads_ad calls as a preview. Wait for my approval before creating anything.

Ad copy and creative

Meta creative concepts from competitors' long-running ads

Analyzes competitors' live Meta ads and your top-spending ads side by side, then writes three new creative concepts built on angles that competitors keep running. Nothing is created or published.

Use competitor_ads (platform=meta) to pull the live Meta ads for [competitor 1] and [competitor 2]. For each ad, note:
- The format: single image, video or carousel.
- The hook in the first line of the primary text.
- The headline and the CTA button.
- How long the ad has been running, if the library data shows a start date.

Treat ads that have been running for 30 days or more as likely winners. If no start dates come back, say so and judge by repetition instead (the same angle across several ads).

Then find the account ID for [Meta ad account] with list_meta_ad_accounts. Use meta_campaign_insights at level=ad with date_preset=last_30d to rank our ads by spend, and check with list_meta_ads that the top 5 are still active. For each, show CTR (clicks / impressions), cost per result (spend / the main conversion action in the data), and frequency (impressions / reach). Read their creatives with get_meta_ad_creative.

Deliver:
1) A side-by-side table comparing competitor ads with ours: format, hook, headline, CTA, and the angle behind it.
2) Three new creative concepts for us. Each one should test an angle that competitors' long-running ads suggest works and that our current top ads do not use. For each concept, give the hook, primary text (125 characters or fewer), headline (40 characters or fewer), CTA, format, and which of our current ads it would challenge. Show character counts.
3) Pick the strongest concept and explain why.

For a visual baseline, if get_meta_ad_creative returns a creative_id for the ad the strongest concept would challenge, render that existing creative with preview_meta_ad in MOBILE_FEED_STANDARD and INSTAGRAM_STORY, and describe what the new concept would change in each placement. Do not create, duplicate or publish any ad or creative.

Ad copy and creative

Match competitor offers with callouts and promotion assets

A list of the offers competitors promote in their ads, compared with our extension assets, and new callouts staged for approval.

Use competitor_ads to pull the live Google ads for [competitor domain 1] and [competitor domain 2]. List every offer they promote, such as discounts, free shipping, installments, free trials, guarantees, fast delivery and price-match. Note how often each offer appears.

Then use list_extension_assets to read the extension assets linked to [campaign name]. Pull 30-day performance for those assets with extension_asset_performance, showing impressions, clicks, CTR and conversions.

Deliver a table with three columns: competitor offers, whether we have an equivalent asset, and how that asset performs. Flag any of our callouts or sitelinks that have at least 500 impressions and zero clicks.

Then propose:
- Up to 4 new callouts, 25 characters or fewer each.
- At most 1 promotion asset.

Only propose offers that we actually run. List them separately so I can confirm each one is valid and still current.

After I confirm, show a preview of the exact manage_extension_assets change. Wait for my approval before adding anything.

Ad copy and creative

Competitor ad-to-landing page teardown vs our page

A side-by-side teardown of where competitor ads send traffic versus our landing page, with five prioritized fixes.

Ad copy and creative

Audit a PMax asset group for missing and weak assets

Read-only audit of every asset group in a Performance Max campaign: counts against Google's limits, image sizes, low-rated assets, duplicate copy and the top 5 fixes to produce.

Audit the assets in my Performance Max campaign [campaign name]. This is a read-only review, so do not change anything.

1. Find my Google Ads account with list_accounts, then find the campaign's asset groups.
2. For each asset group, read its assets with get_pmax_assets. Then run a google_ads_query on asset_group_asset for that campaign to get each asset's field_type, status, performance_label, text, and for images the full-size width and height in pixels. Business name and logo can sit on the campaign instead of the asset group, so also check campaign_asset for BUSINESS_NAME and LOGO / LANDSCAPE_LOGO if they are not in the asset group.
3. Build a table for each asset group with one row per asset type and three columns: current count, Google maximum, and gap. Cover these types:
- Headlines (max 15, 30 characters each)
- Long headlines (max 5, 90 characters)
- Descriptions (max 5, 90 characters, at least one 60 characters or less)
- Landscape 1.91:1, square 1:1 and portrait 4:5 images (max 20 images in total). Use the pixel sizes to confirm each image's real ratio, and flag any image below 1200x628 landscape, 300x300 square or 480x600 portrait.
- Square and landscape logos
- Videos (max 5)
- Business name
4. List every asset whose performance label is "Low", and every asset still labelled "Learning". The data does not say how long an asset has been learning, so just list them and say so if the label field comes back empty. For each one, show the text, or for images and videos the file name, size and anything the data says about what it shows.
5. Flag these specific problems:
- No portrait image
- Fewer than 5 headlines that are meaningfully different from each other
- Duplicate or near-duplicate text across headlines, long headlines and descriptions
- Any headline that just repeats the business name
- No description of 60 characters or less
6. Finish with the top 5 fixes ranked by your judgement of likely impact, and include the asset group's ad strength in your reasoning. Mark each fix as text, image or video so I know what I need to produce.

Ad copy and creative

Replace "Low" rated PMax headlines and descriptions

Rewritten replacements for every low-performing PMax text asset, previewed side by side before anything is swapped.

In my Performance Max campaign [campaign name], find every headline, long headline and description that Google rates "Low" over the last 30 days. Skip assets added in the last 14 days, because they have not had time to learn.

For each low asset, give me:
- The current text
- Why it is likely weak, for example generic, duplicated, no benefit, or no number or offer
- One replacement that stays within the character limit

My business sells [main product or service], and my strongest selling point is [key differentiator]. Use the "Best" rated assets in the same asset group as a style reference. Do not copy them.

Rules for the replacements:
- Mix angles across the set: price or offer, proof, speed, location, and a direct call to action.
- Do not make claims I cannot back up. No "best", "number 1" or "guaranteed".

Show everything as a remove/add table per asset group. Do not apply anything. Wait for my approval, then apply only the rows I confirm with update_pmax_assets. Keep at least the minimum number of headlines and descriptions live at every step.

Ad copy and creative

Mirror Google sitelinks and callouts to Microsoft Ads without duplicates

A Google vs Microsoft asset diff and a previewed plan to add only the missing sitelinks and callouts in Bing.

Our client [client account] runs the same Search campaign on Google and Microsoft. Bring the Microsoft side up to the Google standard.

Step 1. Read the sitelinks and callouts linked to Google campaign [Google campaign name]. Then read the extensions already attached to Microsoft campaign [Microsoft campaign name].

Rules:
- Treat a Microsoft extension as a match if its text or URL is the same or nearly the same. Never add a duplicate, because duplicates are a common cause of disapprovals.
- If a Google sitelink points to a page that does not fit this Microsoft campaign's targeting, mark it "review" instead of "add".
- Microsoft tooling here supports sitelinks and callouts only. List the Google structured snippets separately as a manual task.

Step 3. Show exactly what will be added to the Microsoft campaign as a preview. Wait for my approval before adding anything.

Ad copy and creative

Build structured snippets per ad group from the client's WordPress site

Pulls real services, categories and products from the client's WordPress site and turns them into structured snippets at campaign or ad group level, skipping anything already linked, with a preview before anything is added.

Create structured snippets for [campaign name] in Google Ads account [customer ID], using the client's WordPress site [site URL] as the only source of values.

1. Confirm the site is connected, then check whether WooCommerce is active. List the published pages and the post categories. If WooCommerce is active, also list the product categories (taxonomy product_cat) and the products.
2. List the campaign's ad groups. Then list the structured snippets already linked to the campaign, and query the ad group assets of type STRUCTURED_SNIPPET for those ad groups so you also see snippets attached at ad group level.

For each ad group theme, pick the most fitting header from Google's predefined structured snippet header list, such as Services, Types, Brands, Models, Courses or Destinations. Each snippet needs:
- 4 to 10 values
- each value 25 characters max
- every value matching something actually offered on the site, with no invented services or brands

Where ad group themes differ clearly, attach the snippet at ad group level. Otherwise use one campaign-level snippet. Skip any header and values combination that already exists at the same level.

Output, grouped by ad group: header | values | source page or product for each value | level (campaign / ad group).

Then run manage_extension_assets with confirm=false for each snippet (campaign_id for campaign level, ad_group_id for ad group level) and show me the previews. Add nothing until I approve.

Ad copy and creative

Turn high-ROAS search terms into new RSA headlines

New headlines that mirror the exact queries producing the most revenue, mapped to the ad groups where they matched.

Ad copy and creative

Borrow high-CTR organic messaging for paid RSA headlines

Finds the organic queries where your pages beat the site's own CTR benchmark for their position, reads the meta titles that earn those clicks, and proposes RSA headline swaps for the ads sending traffic to the same pages.

Use Search Console property [GSC property], WordPress site [site URL] and Google Ads account [customer ID].

1. Build the benchmark: pull site-wide query data for the last 90 days (dimension query, sorted by impressions, a large limit). Group queries into position bands (1 to 2, 2 to 3, 3 to 4, 4 to 5) and compute the impression-weighted CTR for each band. That is the site-wide CTR at a similar position.
2. Pull page + query data for the last 90 days for pages under [category URL path]. Keep the queries that meet all three conditions:
- average position of 5 or better
- 200 or more impressions
- CTR at least 1.5 times the benchmark for its position band

3. For the pages behind those queries (top 10 by impressions if there are more), find each page's WordPress ID and read its current meta title and meta description.
4. In campaign [campaign name], list the ad groups and read the RSAs with get_ad_details. Keep the ads whose final URL matches one of those pages.
5. Run the RSA asset report for the campaign over the last 30 days and, for the matching ads, rank the current headlines by conversion value (treat assets with very few impressions as not yet judged).

Compare the angles organic searchers respond to, such as price, assortment, a specific attribute or a guarantee, with the angles in those RSA headlines.

Deliver:
- a table with these columns: page, top organic query, organic CTR, band benchmark CTR, meta title, angle missing from ads
- up to 5 new headlines of 30 characters or fewer, each replacing a named weak headline in a named ad

Show the update_rsa preview (confirm=false, using remove_headlines and add_headlines) and wait for my approval. Do not change anything on the WordPress site.