Opus Growth

Account audits

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

Account audits

Find wasted search terms and build a safe negative keyword list

A themed table of money-wasting search terms with ready-to-apply negatives and projected savings, applied only after you approve.

Find the search terms wasting money in my Google Ads account [account name] over the last 60 days.

Use the search terms report and the smart negative keyword suggestions. First read all my existing negatives (campaign level, ad group level and shared lists) so you do not suggest duplicates.

Flag a term as waste if:
- it spent more than [amount] with zero conversions, or
- it clearly has the wrong intent (jobs, free, DIY, second hand, places I do not serve, competitor names), regardless of spend.

Deliver a table with these columns: search term, campaign, cost, clicks, conversions, suggested negative, match type (exact or phrase), level (campaign, ad group or shared list), and a one-line reason. Group rows into themes and show the total spend these terms took in 60 days.

Rules: never block a term that converted at least once. Warn me if a negative would also block one of my converting keywords.

Then show a preview of exactly what you would add and wait for my approval before applying anything.

Account audits

Check conversion tracking before trusting any CPA or bidding decision

A pass/fail check of your conversion setup across Google Ads, GA4 and Tag Manager, with the fixes that stop Smart Bidding from optimising toward the wrong action.

Before I make any optimisation decisions, check whether conversion tracking in my Google Ads account [account name] can be trusted. Use the last 30 days.

Check and report pass/fail for each:
- primary conversion actions that recorded zero conversions, or are paused or removed
- the same lead counted twice (for example a Google Ads tag and an imported GA4 key event both set as primary for the same form)
- weak actions set as primary, such as page views, scrolls or button clicks
- lead actions counting "every" instead of "one"
- lookback windows and default values that make no sense for my business
- Google Ads conversions vs GA4 key events for the same action: flag any gap above 20%
- in Tag Manager, whether a conversion tag exists for each primary action and fires on the right trigger

Deliverable: a checklist table (check, status, evidence, fix), then a short verdict: can I trust my CPA today, yes or no.

If a primary/secondary setting should change, show me a preview of the exact change and wait for my approval. Do not change anything silently.

Account audits

Spot suspicious clicks and block fraudulent IPs after review

A scored list of suspicious IP addresses tied to campaign anomalies, with a blocking plan you approve before anything is excluded.

Account audits

Clean up weak ads, disapprovals and missing extensions

A per-ad-group checklist of ad strength, underperforming headlines, policy issues and extension gaps, plus draft replacement copy.

Audit the ads and extensions in my Google Ads account [account name]. Use the last 60 days.

For each active Search ad group, check:
- ad strength of every responsive search ad, and Google's action items for improving it
- headlines and descriptions with at least 1,000 impressions whose CTR is below half the ad group average
- headlines pinned without a clear reason (pinning limits combinations)
- disapproved or limited ads and the exact policy reason
- extensions: fewer than 4 sitelinks, no callouts, no structured snippets, or no call extension; also sitelinks that got impressions but zero clicks

Deliverables:
1. A checklist table per ad group: issue, evidence, fix.
2. For each weak headline or description, 2 replacement drafts that include the ad group's main keyword and a concrete benefit or offer from my business: [main offer or USP]. Respect the 30 and 90 character limits.
3. For each disapproval, what to change to get it approved.

Edit existing ads in place rather than creating new ones, so performance history is kept. Show me a preview of every edit and wait for my approval.

Account audits

Meta creative fatigue detector with refresh angles

A ranked list of fatigued Meta ads based on frequency, CTR decay and rising cost per result, with new creative angles to replace them.

Find ads showing creative fatigue in Meta ad account [ad account].

Data: pull daily ad-level data for the last 28 days: impressions, reach, frequency, link CTR, CPM and cost per result. For each ad, compare its last 7 days with its first 7 days of delivery inside the window. Ignore ads with fewer than 3,000 impressions in the last 7 days.

Flag an ad as fatigued only when all three are true:
- Frequency is above [frequency limit, e.g. 3.5] for prospecting, or above 6 for retargeting.
- Link CTR has dropped 25% or more.
- Cost per result has risen 20% or more.

For each flagged ad, read the creative (headline, primary text, format, CTA) so I can see which angles are wearing out.

Deliverable:
1. A ranked table with these columns: ad, ad set, frequency, CTR change %, cost per result change %, spend last 7 days, creative angle.
2. For each affected ad set, a note on whether a healthy ad is left to carry delivery.
3. Two new creative angles per ad set that clearly differ from the fatigued ones.

If you recommend pausing any ad, list the pauses as a preview and wait for my confirmation before calling any write tool.

Account audits

Meta under-delivery diagnosis: why ad sets are not spending

A ranked list of Meta campaigns and ad sets that are not spending their budget, with the likely cause, the evidence and a previewed fix for each.

Diagnose under-delivery in Meta ad account [ad account].

Scope: over the last 7 days, find active campaigns and ad sets that spent less than 60% of their budget or had at least one day with zero impressions.

For each one, check these causes in this order:
1. Ad and ad set status: disapproved or in-review ads, or all ads paused so no active ad is left.
2. Audience size and expected daily results from the delivery estimate.
3. A bid cap or cost cap set too low compared with the actual cost per result.
4. Schedule and end dates.
5. Retargeting audiences too small to serve.
6. Edits in the change history that coincide with the drop.

Deliverable: a ranked list with the biggest unspent budget first. Columns: entity, budget, actual spend, delivery %, most likely cause, evidence, fix. Label any cause you cannot prove with data as a hypothesis.

Some fixes require a change, such as raising a cost cap, broadening targeting or re-activating an ad. For those, prepare a preview of the exact edit and wait for my go-ahead before applying it.

Account audits

TikTok ad group ROAS audit: scale, hold or cut

Read-only ad group audit of the last 30 days, including Smart+, bucketed against your target ROAS with a likely cause for every weak ad group.

Audit TikTok ad account [account name] at ad group level for the last 30 days. Include Smart+ campaigns, because they do not appear in the standard campaign list. Pull them with list_tiktok_smart_plus at CAMPAIGN and ADGROUP level.

Get performance with tiktok_report_query: report_type BASIC, data_level AUCTION_ADGROUP, dimensions ['adgroup_id'], with spend, impressions, ctr and the purchase metrics (purchase count, purchase value, ROAS, cost per purchase). If Smart+ ad groups do not come back at ad group level, report them at campaign level (AUCTION_CAMPAIGN). Mark those rows "Smart+, campaign level" so they are not mixed up with standard ad groups.

For every ad group with spend, show: campaign, ad group, optimization goal and event (from list_tiktok_adgroups), spend, impressions, CTR, purchases, purchase value, ROAS and cost per purchase.

Then sort them into buckets using my target ROAS of [target ROAS]:
- Scale: ROAS at least 20% above target and spend above [minimum spend]
- Hold: ROAS within 20% of target
- Fix or cut: ROAS more than 20% below target, or spend above 2x the account's average cost per purchase with zero purchases. Calculate the average as total spend divided by total purchases.

Deliverable: one table sorted by spend. Add a total row with blended ROAS (total purchase value divided by total spend).

Then, for each "Fix or cut" ad group, write one line on the most likely cause. Base it on the settings returned by list_tiktok_adgroups (optimization goal and event, bid type and bid price, locations, ages, interests, audiences) and on the number of enabled ads from list_tiktok_ads. Possible causes: wrong optimization event, too narrow or too broad audience, bid cap, or too few active ads. If a setting is not visible in the data, say so and treat the cause as a hypothesis, not a finding.

Separately, list any ad group that optimizes for clicks, landing page views or add to cart instead of purchase.

This is a read-only audit. Do not change anything.

Account audits

Microsoft Ads search term waste audit with negative keyword plan

Finds 60 days of wasted search term spend on Bing, builds a de-duplicated negative keyword plan that protects converting terms and previews it before adding anything.

Audit search term waste in Microsoft Ads account [account name] for the last 60 days, across all Search campaigns.

Use microsoft_search_terms_report (days 60) for terms, matched keywords, match types, clicks, spend and conversions. The standard report may not include revenue. If it does not, pull microsoft_report_query with report_type SearchQueryPerformance, the same date range, and columns including CampaignName, AdGroupName, SearchQuery, Keyword, DeliveredMatchType, Clicks, Spend, Conversions and Revenue. If revenue is still unavailable, apply only the zero-conversion rule below and tell me ROAS could not be calculated.

Flag a search term as waste if:
- spend is above [spend threshold] with zero conversions, or
- it has at least 20 clicks and ROAS below 50% of my target ROAS of [target ROAS].

For each flagged term, show: campaign, ad group, search term, matched keyword and match type, clicks, spend, conversions, revenue and ROAS. Sort by spend.

Then build a negative keyword plan:
- group terms by shared root (competitor brands, "free", "used", "repair", job searches, wrong product types)
- suggest exact or phrase match and campaign or ad group level for each
- check the existing negatives first with list_microsoft_negative_keywords on each campaign and its ad groups, and skip anything already blocked
- check list_microsoft_keywords for every ad group (from list_microsoft_ad_groups), and never suggest a negative that would block one of my active keywords or a term that has converted

Also list search terms with 2+ conversions and ROAS above target that are not yet keywords.

Estimate monthly savings as half of the flagged 60-day spend for each negative. Label it as an estimate, since some of that traffic may shift to other terms.

Show the negative keyword changes as a preview with add_microsoft_negative_keywords (confirm=false), and wait for my approval before adding any of them.

Account audits

LinkedIn Ads spend justification audit for an e-commerce brand

Checks LinkedIn conversion tracking, ROAS and cost per conversion per ad set over 60 days and gives a keep, fix or pause verdict for each one.

We are an e-commerce brand, and LinkedIn has to earn its budget. Audit LinkedIn ad account [account name] for the last 60 days.

Use the current Campaign Manager names: "campaigns" are the top level and "ad sets" sit inside them. Report at ad set level and show the campaign each one belongs to.

Step 1, tracking: list every conversion rule with its type, whether it is enabled, its attribution windows, whether it carries a value, and which ad sets it is attached to. Flag running ad sets with no purchase or checkout conversion attached. Using a conversion-level breakdown of the last 60 days, flag conversion rules that recorded zero conversions. If the conversion list comes back truncated, say so.

Step 2, performance: for every ad set with spend, show campaign, objective, spend, impressions, CTR, average CPC, CPM, conversions, conversion value if LinkedIn returns it, ROAS and cost per conversion, plus the number of active ads. Where no conversion value is tracked, write "no value tracked" instead of a ROAS.

Step 3, verdict per ad set, using my target ROAS of [target ROAS]:
- Keep: ROAS at or above target
- Fix: tracking is missing, no conversion value is tracked, or ROAS is within 30% of target
- Pause candidate: spend above [spend threshold] with ROAS more than 30% below target, or zero conversions

Deliverable: one table with the verdict column, total LinkedIn spend and ROAS (based on ad sets with tracked value only, and say how much spend that covers), and one paragraph on whether LinkedIn should stay in the mix.

If an ad set needs a conversion attached, show it as a preview and wait for my approval. Do not detach any conversion and do not pause anything.

Account audits

Google Ads conversion action audit for lead-gen bidding

A table of every conversion action with its settings and data, plus a ranked fix list that protects Smart Bidding.

Audit the conversion setup of my Google Ads account [account name]. We are a local service business, so only form submissions and qualified calls should drive bidding. Use the last 30 days.

1. List every conversion action with its name, category, status, primary or secondary flag, counting type (every/one), click-through lookback window, default value and conversions recorded in the last 30 days.
2. Flag these issues: actions with zero recorded data, paused or removed actions that are still primary, page views or generic button clicks set as primary, lead actions counting "every" instead of "one", and pairs of actions that probably track the same event (double counting).
3. Check which conversion categories are primary in the account goals and any custom goals, and whether that matches what my campaigns actually optimize for.

Deliver one table, then a short fix list ranked by impact on Smart Bidding. Do not change anything yet. Show each proposed change, such as switching an action to secondary or changing counting to "one", as a preview and wait for my approval.

Account audits

Find duplicate GTM and GA4 tags on my WordPress site

A map of every place tracking loads on the site, with duplicates flagged and a single-source cleanup plan.

Account audits

Meta pixel and lead optimization event health check

Audits Meta pixels, custom conversions and ad set optimization events for a lead account, checks weekly result volume and reconciles instant form leads against reported leads.

Audit conversion tracking for my Meta ad account [account name]. We run lead campaigns for a local service business. Use the last 30 days.

1. List the pixels with their last fired time. Flag any pixel that has not fired in 24 hours.
2. List the custom conversions with their rules. Call out the ones Meta reports as never fired. For the rest, pull results per custom conversion for this period from the insights actions, and mark each one as recording results or silent.
3. For every active campaign and ad set, show the optimization goal and conversion event, spend, leads and cost per lead. Pull results in weekly buckets so the weekly volume is real, not an average. Flag:
- ad sets optimizing for link clicks or landing page views when we need leads,
- ad sets whose optimization event got fewer than [10] results in any of the last 4 weeks, which is too little for learning.
4. Find the instant forms on my Facebook Page [page name], download the leads submitted in this period and count them per campaign. Compare those counts with the leads reported in insights for the same campaigns. Flag any gap above 10%, and say which side is higher.

Deliver a findings table and a fix list ranked by impact. Switching an ad set's optimization event restarts learning, so note that cost for each such fix. Show every change, including a new custom conversion, as a preview and wait for my confirmation.

Account audits

Rewrite only the offending assets in disapproved RSAs

A before/after fix plan that replaces only the policy-violating headlines and descriptions, keeping each ad's performance history.

In campaign [campaign name], find every responsive search ad that is disapproved or limited by policy.

For each ad:
1. Read the full ad: headlines, descriptions, final URL and ad strength.
2. Identify the exact asset that triggered the policy. Typical causes are trademark use, excessive punctuation or capitalization, superlative claims without proof, and restricted health or financial wording.
3. Check the last 60 days of asset performance (impressions, CTR, conversions). If an offending asset is a top performer, say so and keep its meaning in the replacement.

Then propose a compliant replacement for the offending assets only. Stay within 30 characters for headlines and 90 for descriptions, and match our existing tone. Do not touch assets that are not causing the issue.

Deliverable: one before/after table per ad with these columns: offending text, policy reason, replacement, character count, why it should pass review.

Edit the existing ads in place instead of creating new ones, so performance history is kept. Show me a preview of every edit and wait for my approval before applying anything.

Account audits

Find and clear ChatGPT Ads delivery blockers

A ranked list of the blockers stopping ChatGPT Ads from serving, with evidence, fixes and previewed copy edits applied only after approval.

Our ChatGPT Ads campaign [campaign name] is getting few or no impressions. Find out what is blocking delivery.

1. Run a delivery diagnosis on the campaign.
2. Pull the full details of the campaign, its ad groups and its ads, including everything that decides delivery.
3. Render each ad and list the delivery blockers OpenAI reports for it, such as review status, rejected copy or a destination problem.
4. Pull the last 7 days of daily delivery, including zero-impression rows, so I can see when delivery stopped.

Deliverable: a table ranked by impact, with these columns: blocker, affected entity, evidence, fix, fix type. The fix type is one of: ad copy, destination, bid/budget, targeting, or setup.

For blockers caused by ad copy or the destination URL, draft a revised ad that addresses the reported reason. Show me the exact before/after changes and wait for my approval before updating any ad.

Finish with a 3-sentence status note I can forward to my manager: what was wrong, what we are changing, and when we expect delivery to resume.

Account audits

Weekly fraud scan: block the worst IPs with a clear score rule

A ranked table of suspicious IPs from the last 7 days and a previewed block list based on fraud score and click count.

Account audits

Audit Meta lead ads for fake and junk leads by ad set and placement

Downloads instant form leads, scores junk at lead level, computes junk rate and cost per clean lead by ad set and ad, sets it against placement delivery, and proposes placement changes only as approved previews.

My Meta lead ads in [campaign name] bring a lot of leads, but many never answer the phone or look fake. Audit the last 30 days.

1. Find the campaign, its ad sets and the instant forms on my Facebook Page, then download all leads from the last 30 days with the ad set and ad each came from.
2. Mark a lead as junk if it has any of these: a phone number in an invalid format or made of repeated digits, a duplicate phone or email, a gibberish or single-letter name, a disposable-looking email, or 3 or more leads with identical answers within 10 minutes.
3. Junk by source: table of ad set, ad, spend, leads, junk leads, junk rate, cost per clean lead.
4. Placement delivery: pull spend, impressions, clicks and leads by ad set broken down by publisher platform and position, including Audience Network. Leads are not tagged with a placement, so junk rate by placement cannot be measured directly. Instead, show for each ad set what share of its leads each placement delivered, and point out where a high junk rate coincides with one placement delivering most of the leads. Label this as an inference.

Decision rule: if an ad set has a junk rate above 40% with at least 10 leads, and a single placement (for example Audience Network) delivered more than half of its leads, propose removing that placement from the ad set. Show each change as a preview with the remaining placements and the expected effect on cost per clean lead, and wait for my approval.

Also read the questions on the form I use and tell me whether they make spam easy, with one concrete suggestion (for example a qualifying question or a higher-intent form with a review step, which would be a new form I can approve separately).

Account audits

Link a Google Ads performance drop to the change that caused it

A day-by-day timeline matching every account change to the metric movement after it, plus a ranked list of likely causes.

In Google Ads account [account], conversions for campaign [campaign] dropped starting around [date]. Find out which change caused it.

1. Pull daily cost, clicks, conversions, cost per conversion and conversion rate for this campaign, from 14 days before to 14 days after that date.
2. Pull the change history for the same window. For each change, include the date, user email, client type, resource, changed fields and old→new values.
3. Build a timeline table with these columns: date, who, what changed, old→new, and the movement in conversions and cost per conversion over the 3 days that followed.

Decision rule: treat a change as a suspect only if it happened 0 to 3 days before the drop and touched this campaign's budget, bid strategy or target, keywords, negatives, ads, targeting, schedule or conversion settings. Rank the suspects from most to least likely and explain the evidence for each.

If no change fits the rule, say so clearly and list the external explanations to check next, such as seasonality, auction pressure or tracking.

Do not change anything. For each suspect, propose the exact reversal as a separate list that I will approve.

Account audits

Weekly change governance check across all client Google Ads accounts

One table per client listing risky or unauthorized changes made in the last 7 days, including auto-applied Google recommendations.

Run our weekly change governance check across every connected Google Ads account.

1. List all connected accounts.
2. For each account, pull the change history for the last 7 days.
3. Flag a change if it meets any of these conditions:
   - It was made by an email outside [our agency domain].
   - It came from Google's automated systems or auto-applied recommendations, based on client type.
   - It changed a daily budget by more than 30%.
   - It changed a bid strategy or a target CPA/ROAS.
   - It removed keywords, ads or ad groups.
   - It edited, paused or removed a conversion action or goal.
   - It changed location or language targeting.
4. Where auto-applied changes appear, list the account's current recommendations so I can see what else Google may apply.

Output:
- One table per account with these columns: date, user, client type, resource, change, old→new, flag reason.
- A summary at the top ranking accounts by number of flagged changes.
- A separate line listing accounts with zero changes in 7 days, since those may be neglected.

This is read-only. Do not revert or dismiss anything.

Account audits

Organic traffic dropped after a site update: find the pages that broke

A table of pages that lost organic clicks, with their index status, canonical and meta settings, and a likely cause for each.

Organic traffic on [site] dropped after a site update around [date]. Find which pages broke and why.

1. In Search Console, compare page-level clicks, impressions and average position for the 28 days after that date with the 28 days before it.
2. Keep only pages that lost more than 30% of clicks and had at least 100 clicks in the earlier period.
3. Batch-inspect those URLs. For each one, return the index status, last crawl date, robots/indexing state and Google-chosen canonical.
4. On WordPress, find the matching posts and pages and their last modified dates. Read each page's meta title, meta description, canonical and focus keyword. List the active plugins and note any SEO or caching plugin.

Classify each page into one of these causes:
- NOINDEX or BLOCKED
- CANONICAL MISMATCH, where Google's canonical differs from the declared one
- TITLE/META CHANGED, where the page was modified after the date and its position fell
- NOT RECRAWLED
- RANKING LOSS ONLY, where the page is indexed and unchanged but its position dropped

Output a table with these columns: URL, clicks before and after, position before and after, index status, canonical, modified date, cause. Sort it by clicks lost.

Do not edit anything. List the proposed fixes per page, and I will approve them before you preview any change.

Account audits

Triage every keyword against my break-even ROAS

Scores every enabled Search keyword on 60-day ROAS against your break-even, sorts them into Scale, Keep, Watch and Cut, checks keyword-level impression share for winners and previews pauses and bid cuts for your approval.

My break-even ROAS is [break-even ROAS, e.g. 2.5] and my average order value is [AOV]. Review every enabled Search keyword in [account name] over the last 60 days.

For each keyword, pull cost, conversions and conversion value, plus its match type, ad group, campaign and the campaign's bidding strategy type. Only include keywords that are enabled in enabled ad groups and enabled campaigns. Calculate ROAS as conversion value divided by cost.

Sort each keyword into one of four buckets, using these rules:
- Scale: ROAS at least 1.5x break-even, with 3 or more conversions.
- Keep: ROAS at or above break-even.
- Watch: below break-even but fewer than 3 conversions, so the data is too thin to judge.
- Cut: cost above 2x AOV with zero conversions, or ROAS below 50% of break-even with 5 or more conversions.

Deliver two things:
1. A table with these columns: keyword, match type, campaign, bidding strategy, cost, conversions, conversion value, ROAS and bucket.
2. A summary showing the total spend and revenue in each bucket.

For Cut keywords, propose pausing them. Do not propose removing them.

For Watch keywords in campaigns on manual CPC, propose a max CPC cut of 20% from the keyword's current bid. If the campaign uses Smart Bidding or a portfolio strategy, do not touch its bids; say so instead.

For Scale keywords, pull keyword-level search impression share and the share lost to rank. List the ones losing impression share so I can review them. Lost to budget is only reported at campaign level, so add the campaign's lost-to-budget share next to each of them.

Show me a preview of every pause and bid change. Wait for my approval before applying anything.

Account audits

Diagnose Meta ad sets that are under-delivering or not spending

A ranked list of under-delivering ad sets with the most likely cause, the evidence and the fix for each.

Some of my Meta campaigns are barely spending, and I am losing leads because of it. Diagnose delivery in [ad account] for the last 7 days.

Flag every active ad set that either spent less than 60% of its budget or had zero impressions. For each flagged ad set, check these causes in order:
1. Effective status of the campaign, ad set and ads (paused parent, rejected or in-review ads, ended schedule).
2. Audience size and expected daily results from the delivery estimate, for this ad set's current targeting.
3. Bid cap or cost cap set below the actual cost per lead from the last 30 days.
4. Daily budget too low for the optimization event. Treat a budget below 2x [target CPL] as too low for lead optimization.
5. Recent edits in the change history that coincide with the drop.
6. Overlap with another ad set in the same campaign that targets the same area.

Deliver a ranked table:
- ad set
- spend vs budget
- most likely cause
- evidence (the actual numbers)
- recommended fix

Do not change anything. Show each fix (budget, bid, targeting or status) as a preview and wait for my approval.

Account audits

Find Meta placements and devices that waste lead budget

Breaks down each active lead ad set by platform, placement and device, flags segments with high CPL or zero leads, adds an ad set level lead quality check from instant form data, and previews any placement change before applying it.

I want to know where my Meta lead budget actually goes. Work in ad account [ad account] and cover every active ad set in a lead campaign.

1. Pull the last 30 days at ad set level with meta_insights_query. Run these breakdowns as separate pulls if Meta rejects the combination:
- publisher_platform + platform_position (Feed, Stories, Reels, Audience Network, Messenger and so on)
- device_platform
Request spend, impressions, inline_link_clicks, actions and cost_per_action_type. Use one lead action type consistently (instant form leads or the pixel Lead event, whichever these ad sets optimize for) and tell me which one you used.

2. For each segment, show:
- spend and its share of the ad set's spend
- impressions and link CTR
- leads and cost per lead

3. Flag a segment as waste if either is true:
- it takes more than 15% of the ad set's spend and its cost per lead is above 1.5x the ad set average
- it spent more than 2x [target CPL] with zero leads

4. Lead quality cannot be broken down by placement, so check it at ad set level. Find the Pages with list_meta_pages, list their instant forms with list_meta_lead_forms, and download the last 30 days of leads with download_meta_leads. Using the ad set each lead came from, count per ad set the leads with an invalid looking phone number (wrong length, missing digits, obvious fakes like 5555555555) and duplicates (same phone or email submitted more than once). Report counts and percentages only, not personal details.

Deliver:
- A table of flagged segments, sorted by wasted spend.
- A recommendation for each ad set: exclude specific placements, or keep Advantage+ placements. Use the ad set quality numbers to say whether a cheap ad set is really cheap or just collecting bad leads.

For each recommendation, warn me that narrowing placements can raise CPM and resets the learning phase. If I want a change, prepare it with update_meta_adset (publisher_platforms and the matching *_positions fields) using confirm=false, show me the preview for each ad set, and apply nothing until I approve.

Account audits

TikTok Ads account health check before a management update

A read-only 30-day health check of the TikTok account with a campaign table and the 5 biggest issues ranked by money at risk.

Account audits

TikTok creative fatigue audit: keep, refresh or pause each ad

Compares each TikTok ad's recent CTR, cost per conversion and frequency with its earlier performance, gives a keep, refresh or pause verdict, lists unused library videos as replacements and previews any pause before applying it.

Audit creative fatigue in TikTok account [advertiser account].

1. Pull ad-level daily data for the last 28 days with tiktok_report_query (dimensions ad_id and stat_time_day): spend, impressions, clicks, CTR, conversion, cost_per_conversion, video_watched_2s and video_watched_6s. Pull frequency separately at ad level for the first 14 days and for the last 7 days, without the day dimension, because daily frequency cannot be added up. If a metric is not returned, say so and continue without it.
2. Get ad names, ad groups and the creative each ad runs on (video_id, or tiktok_item_id for Spark ads) from list_tiktok_ads and list_tiktok_adgroups.
3. For each ad that spent at least [minimum spend], compare the last 7 days with the first 14 days.
4. Flag an ad as fatigued if either is true:
- CTR fell more than 25% and cost per conversion rose more than 20%.
- Frequency in the last 7 days is above 3.

Output a table with these columns: ad, ad group, video (video_id or Spark post), days live (days with spend in the 28-day window), CTR (first period vs. last period), cost per conversion (first period vs. last period), frequency, and verdict (keep / refresh / pause).

Then compare list_tiktok_videos with the video_ids used by all ads and list the library videos that are not used in any ad. These are replacement candidates. Note that Smart+ ads are not in the standard ad list, so a video used only in Smart+ may show up as unused.

If you recommend pausing any ads, prepare it with update_tiktok_ad_status (operation DISABLE, confirm=false), show me the exact list as a preview and wait for my approval before changing any status.

Account audits

LinkedIn conversion tracking audit: is every campaign measured?

Maps every LinkedIn conversion rule to the ad sets it is attached to, checks which rules actually recorded conversions, flags unmeasured spend, dead or duplicate rules and objective mismatches, and previews every attach or detach before applying it.

Check whether LinkedIn Ads account [account] measures what management will ask about. Remember that LinkedIn's API "campaigns" are the "ad sets" in Campaign Manager; report at ad set level and show the parent campaign (campaign group) next to each one.

1. List every conversion tracking rule with list_linkedin_conversions: type, enabled status, attribution windows, URL rules and the ad sets each rule is associated with. If the list is marked truncated, say so.
2. Pull the last 30 days from linkedin_campaign_analytics with pivot CAMPAIGN,CONVERSION to see how many conversions each rule recorded per ad set, and with pivot CAMPAIGN for spend, clicks, conversions and cost per conversion per ad set.
3. Get each ad set's objective and status from list_linkedin_campaigns, and group status from list_linkedin_campaign_groups (a paused group stops its ad sets even if they read ACTIVE).

Flag these issues:
- Active ad sets that spend but have no conversion attached.
- Enabled conversion rules with zero conversions in 30 days.
- Likely duplicate rules that could double count the same action (same type and overlapping URL rules attached to the same ad set). Tell me why you think they overlap.
- Site-wide rules (for example "All Pages") that inflate conversion counts.
- Ad sets whose attached conversion does not match their objective. For example, a lead ad set that only tracks page views.

Deliver a table (campaign, ad set, objective, attached conversions, spend, conversions, cost per conversion, issue) and a short fix list in priority order.

If an ad set needs a conversion attached or detached, prepare it with attach_linkedin_conversion or detach_linkedin_conversion using confirm=false and show me a preview of each one. Before any detach, check whether the ad set uses MAX_CONVERSION bidding and this is its last conversion, because that can stop delivery. Wait for my approval before applying anything.

Account audits

Microsoft Ads wasted spend audit from search terms

Total wasted spend on non-converting Bing search terms, grouped by theme, with a deduplicated negative keyword list to approve.

Account audits

Audit account structure and propose a cleaner campaign layout

A map of your current campaigns and ad groups with structural problems flagged and a proposed restructure plan.

Audit the structure of my Google Ads account [account name] using the last 90 days of data.

Start with an account snapshot, then map every active Search campaign, its ad groups, keyword count and number of active ads.

Flag these problems:
- ad groups with more than 20 keywords or keywords covering clearly different services
- ad groups with no active responsive search ad, or only one
- the same keyword in more than one ad group or campaign (show which one gets the traffic)
- broad match keywords in campaigns using manual CPC
- ad groups or campaigns that spent money with zero conversions in 90 days
- several services sharing one budget, where one service takes most of the spend

Deliverables:
1. Current structure as a simple outline, with problems marked next to each item.
2. A proposed structure: which campaigns and ad groups to create, merge or pause, and which keywords move where.
3. The 3 changes with the biggest expected impact, in order.

Do not change anything now. If I approve parts of the plan, show me a preview of each change and wait for my confirmation.

Account audits

Find wasted spend hidden in location, schedule and device settings

Audits every active Google Ads campaign for waste in location options, out-of-area traffic, cities, hours, devices and networks over 90 days. Delivers one table per area with a verdict and a savings-ordered list of setting changes, applied only after a preview.

Audit the settings of my Google Ads campaigns in [account name] for hidden waste. Use the last 90 days. My target CPA is [target CPA]. My business hours, when someone can answer the phone or reply to a lead, are [business hours, e.g. weekdays 09:00-18:00, Saturday 10:00-14:00].

Start with an account snapshot to list active campaigns, their type and bidding strategy. Then check each active campaign:

- Location option: read the current targeting and tell me whether it uses "Presence or interest" instead of "Presence". Show cost and conversions from users who were outside my targeted locations (use user location data via a GAQL query).
- Locations: which cities or regions spent more than [amount] with zero conversions, or had a CPA above 1.5x my target. If a campaign uses radius targets, list them too.
- Ad schedule: read the current schedule. Break down cost and conversions by day of week and hour. Flag hours and days with CPA above 1.5x target, and spend outside my business hours.
- Devices: device CPA vs account average.
- Networks: for Search campaigns, read whether Search Partners and Display expansion are on, and compare cost, conversions and CPA by network.

Judgement rules:
- Ignore segments with fewer than 20 clicks; they are not enough data.
- On Smart Bidding campaigns (Maximize Conversions, Target CPA, Target ROAS), bid adjustments for schedule and device are ignored except a -100% device exclusion. Only recommend "lower bid" on Manual CPC or Maximize Clicks campaigns; elsewhere the options are keep or exclude.
- Location bid adjustments are not available here, so location verdicts are keep or exclude.

Deliverable: one table per area (segment, cost, clicks, conversions, CPA, verdict: keep, lower bid, exclude) and a final list of recommended setting changes, ordered by estimated money saved. For each change name the exact action: switch to "Presence", add location exclusions, replace the ad schedule with specific blocks, set a device bid adjustment, turn Search Partners or Display expansion off, or edit a radius target.

Show me a preview of every setting change and wait for my approval before applying it. For schedule changes, show the current schedule next to the new one, because narrowing a schedule means replacing it.

Account audits

TikTok creative fatigue check with pause shortlist

Compares the last 7 days with the 21 days before at ad level, flags fatigued creatives, builds a safe pause shortlist and finds ad groups that need new creatives.

Check creative fatigue in TikTok ad account [account name]. Compare the last 7 days with the 21 days before (28 days in total).

Pull the data with tiktok_report_query, report_type BASIC, data_level AUCTION_AD, dimensions ['ad_id']. Run it once for each window using start_date and end_date, so frequency is calculated over the whole window and not summed by day. Request: spend, impressions, clicks, ctr, frequency, video_play_actions, video_watched_2s, video_watched_6s, and the purchase count and purchase value metrics. Calculate the 2-second and 6-second view rates as views divided by impressions.

For each active ad with at least [minimum spend] spend in the period, show: ad group, ad name, days live, spend, CTR (both windows), 2-second and 6-second view rates, cost per purchase (both windows), ROAS (both windows) and frequency where available. Take days live from the ad's creation date in list_tiktok_ads. If that date is not returned, use the first day with spend from a daily query (dimensions ['ad_id','stat_time_day']). Image ads have no view rates. Mark them "n/a".

Flag an ad as fatigued if at least two of these are true:
- CTR dropped more than 30%
- cost per purchase rose more than 25%
- ROAS dropped more than 25%
- frequency is above 3 in the last 7 days

If frequency is not returned for an ad, judge it on the other three signals, still requiring two, and note "frequency unavailable".

Deliverables:
1. A table of fatigued ads, worst ROAS first.
2. A pause shortlist: fatigued ads whose ROAS in the last 7 days is below [target ROAS], but never the last healthy ad in an ad group.
3. A list of ad groups left with fewer than 3 healthy ads, which need new creatives.

Show the pause list as a preview with update_tiktok_ad_status (operation DISABLE, confirm=false). Wait for my approval before pausing anything.

Account audits

GA4 lead measurement and Google Ads link checklist

Checks that a GA4 property tracks leads correctly, flags events that should be key events or fire too often, and confirms the Google Ads link and paid traffic attribution.

Check that GA4 property [property name] measures leads properly for a local service business and feeds Google Ads correctly. Use the last 28 days.

1. List the data streams with their measurement IDs and the site URL each one is bound to.
2. List the key events with their counting method. Then run a report split by event name, so every event in the property appears with its event count. Mark which ones are key events and which are not. Pay attention to lead-related events that are not key events yet, such as form_start, form_submit, generate_lead, and any phone or WhatsApp click events.
3. Find my Google Ads account [account name] and its 10-digit ID, then confirm whether the property is linked to it.
4. Show sessions and key events by session source/medium. I want to see whether paid traffic arrives as google / cpc rather than referral or (not set).
5. Pull sessions by page for my contact page [contact page path] and my thank-you page [thank-you page path, if any].

Over-firing rule: a form-based lead key event (form_submit, generate_lead or similar) should not fire more often than there are sessions on the page where the form sits. Flag any form key event whose count is higher than the sessions on [contact page path], and flag it as likely double counting if its count is more than 1.5 times those sessions. Phone and WhatsApp click events can fire from every page, so compare those against total sessions instead and flag them if they exceed 10% of all sessions.

Deliver:
- a checklist with OK or Missing for each item: data stream present, lead key event exists, counting method sensible, Ads link present, paid traffic tagged as google / cpc,
- the events that should become key events, with a reason for each and the counting method you recommend,
- the over-firing warnings from the rule above, with the numbers behind each one.

For any key event to create or any Ads link to set up, show me a preview and wait for my approval.

Account audits

Google Ads policy exposure report for management

A read-only report of every disapproved or limited Google Ads ad, with the spend and conversions at stake, a severity rating and a 5-line executive summary ready to share.

Build a policy exposure report for Google Ads account [account name] that I can share with management. Use the last 30 days. This is read-only, so do not change anything.

1. Pull the policy status of all ads, not just the problem ones, so you can tell whether an ad group still has an approved ad serving. List every ad that is Disapproved, Eligible (limited) or limited to areas of interest. For each one, show the campaign, ad group, policy topic, and the reason and evidence Google gives (triggering text, flagged URL, restricted countries or missing certification). Where Google gives no detail, say so instead of guessing.
2. Pull 30-day cost, conversions and cost per conversion at campaign level and at ad level. For affected Search campaigns, also show impression share lost to budget and to rank. Impression share lost to disapproved ads is not reported reliably here, so do not present one; use the count of blocked ads per campaign as the disapproval signal.
3. Assign severity:
   - High: the campaign drove more than [X, for example 20]% of account conversions in the period, or every ad in the ad group is disapproved.
   - Medium: some ads in the ad group are disapproved, but at least one approved ad is still serving.
   - Low: the ad is limited but still serving.

Deliverables:
- A table with these columns: ad, campaign, ad group, policy topic, status, 30-day spend, conversions, severity. Sort it High to Low, then by spend.
- A 5-line executive summary covering:
  - how many ads are affected,
  - how much spend sits in affected campaigns,
  - the single biggest risk,
  - one recommended next step for each High item.

Account audits

Trace a disapproval spike back to account changes

A timeline linking each new disapproval to the edit that preceded it (who, what, old to new), plus process recommendations for management.

Disapprovals increased in Google Ads account [account name] recently, and management wants to know why. Investigate in read-only mode.

1. List every ad that is currently disapproved or limited, with its policy topic.
2. Pull the account change history for the last 14 days, including user, timestamp, operation, changed fields, and old and new values.
3. Match each disapproved ad to any change made in the 72 hours before it, in the same ad, ad group or campaign. Edits to ad text, final URLs, assets and new ad creation matter most.
4. Flag disapprovals on ads that nobody edited. These usually point to a landing page change or a policy update rather than our own edit.

Deliverables:
- A dated timeline of the changes and disapprovals.
- A table with these columns: disapproved ad, policy topic, preceding change (user, date, field, old to new) or "no edit found", likely root cause.
- A summary of how many disapprovals came from our own edits versus outside causes.
- 3 process recommendations, such as a pre-publish policy check or rules for restricted wording.

Do not change anything in the account.

Account audits

TikTok compliance check: verification status and rejected ads

Checks a TikTok advertiser's verification and qualification under its Business Center, flags rejected or non-delivering ads, sizes the spend at risk and proposes ad fixes for approval.

Run a compliance check on our TikTok ad account [advertiser name] in Business Center [Business Center name].

1. Find the Business Center and confirm the advertiser sits under it.
2. Report its business verification, industry qualification and account standing.
3. List the campaigns, the active ad groups and the ads inside them. Also check for Smart+ campaigns, since they do not show up in the standard lists. Using the status and review fields the listing returns, flag every ad that is rejected or not delivering. If a rejection reason is returned, quote it. If not, give the likely reason and mark it as inferred from the status and the ad's text and landing page.
4. Pull the last 14 days of spend and conversions per campaign, and per ad for the flagged ads, so we can see which campaigns depend on the blocked ads.

Deliverables:
- A checklist with each item marked OK, Missing or At risk: business verification, industry qualification, account standing.
- A table of blocked ads with these columns: ad, ad group, status, reason (quoted or inferred), 14-day spend of its ad, 14-day spend of its campaign, suggested fix to the ad text or landing page.
- A separate list of actions that must be done in the TikTok interface, such as uploading qualification documents. These cannot be done from chat.

If an ad's text, landing page or call to action needs changing, show me the proposed update as a preview first and wait for my approval before applying it. On Spark ads built from an organic post, the text comes from the post, so only the landing page and call to action can change here; say so where it applies. Do not delete or pause anything.

Account audits

Set up click-fraud tracking and confirm it captures ad clicks

Installs the click-fraud tag with an approval step, avoids duplicate tags on WordPress, verifies the tag on live pages, then runs a follow-up check 48 hours later comparing captured clicks with Google Ads clicks.

I run my own Google Ads for [website URL] and want to start monitoring click fraud and invalid traffic. This has two parts: setup now, and a check after 48 hours of traffic.

PART 1: SETUP (now)
1. Find out where my site lives: an Opus Growth hosting server, a connected WordPress site, or neither.
2. If it is WordPress, list every place a Google tag or GTM container already fires (plugins, WPCode snippets, header/footer code) and warn me about any duplicates.
3. Generate the click-fraud tag bound to my domain, without installing it yet. Show me the exact snippet and where it will go:
   - Opus Growth hosting: it is written into every page automatically.
   - WordPress: it is added to the site header, shown first as a preview.
   - Anywhere else: I paste it into the <head> of every page myself.
4. Read the current tracking template of each enabled campaign and tell me whether this setup changes it or my final URLs. It should not; say so plainly either way.
5. Wait for my approval, then install. On WordPress, purge the page cache afterwards so the tag goes live.
6. Load my homepage and [main landing page URL] as a visitor would and confirm the fraud tag appears in the live HTML. If it does not, tell me why (cache, wrong page, tag blocked).

PART 2: CHECK (run this again in a new chat after at least 48 hours)
7. Run a fraud scan for the last 2 days. If it reports how many ad clicks it captured, compare that number with Google Ads clicks for the same 2 full days. If it captured less than 70%, give me the likely reasons (tag missing on some landing pages, consent banner blocking it, redirects dropping the click ID, cached pages without the tag).
8. List the top suspicious IPs it found, but do not block anything.

Deliver a short checklist with three columns: step, status (done / missing / problem / waiting for 48 hours), and what I need to do next.

Account audits

Find ad spend leaking to clicks from outside your service area

A breakdown of clicks and cost by users' physical location and a previewed fix for campaigns targeting "presence or interest".

My business only serves [service area, e.g. city or region]. Clicks from elsewhere are wasted money and often low-quality or invalid traffic. Check this for every active Google Ads campaign.

1. Read each campaign's location targeting and its location option (presence vs presence or interest).
2. For the last 30 days, query clicks, cost and conversions by the user's physical location (country and region), per campaign.
3. Table: campaign, location option, physical location, clicks, cost, conversions, share of campaign cost.

Flag any campaign where more than 5% of cost came from users physically outside my service area. Also flag foreign countries with 10 or more clicks and zero conversions, as these are a common sign of bot or click-farm traffic.

For each campaign set to "presence or interest", propose switching to "presence" (people in or regularly in my targeted locations). Show the change as a preview with the cost it would have saved over the last 30 days, and wait for my approval before applying it.

Account audits

TikTok Business Center change log mapped to spend and CPA swings

A per-advertiser view of who changed what across the Business Center this week and which changes preceded big spend or CPA moves.

Audit what changed across our TikTok Business Center [business center] in the last 7 days.

1. List the advertisers under this Business Center.
2. Pull the Business Center change log for the last 7 days. Include who made each change, when, on which advertiser and object, and what changed.
3. For each advertiser with changes, pull daily spend, conversions and CPA for the last 14 days at campaign level.

Flag these changes:
- Campaign or ad group status changes, including pauses, enables and deletions.
- Budget changes greater than 30%.
- Bid or optimization goal changes.
- Targeting edits.
- Deleted ads.
- Any change made by a user who is not on [team member list].

Then mark a flagged change as "impact" if, within 3 days, the affected campaign's daily spend moved more than 40% or its CPA moved more than 25% compared with its prior 7-day average.

Output:
- One table per advertiser with these columns: date, user, object, change, spend and CPA before vs after, impact yes/no.
- A top summary listing the advertisers with impact changes first.

Read only. Do not change anything.

Account audits

Performance Max asset group hygiene check against target ROAS

A per-asset-group table of ROAS, missing assets and signal gaps, plus drafted replacement text for the weakest group, applied only after approval.

Review every Performance Max campaign in [account name] over the last 30 days against my target ROAS of [target ROAS].

For each campaign and asset group, report cost, conversions, conversion value and ROAS. Then read each asset group's content and signals and check:
1. Asset counts versus the full set: 15 headlines, 5 long headlines, 5 descriptions, at least 15 images and at least 1 video. Note anything missing.
2. Whether an audience signal is attached and whether search themes are set.
3. Whether the final URL matches the products the asset group is meant to sell.
4. Asset groups that cover several unrelated product categories at once.

Also list Google's open recommendations for these campaigns, with their projected impact.

Deliver a table with these columns: asset group, campaign, ROAS versus target, missing assets, signal gaps, URL issue and priority.

For the asset group with the lowest ROAS that has meaningful spend, draft replacement text assets: headlines of 30 characters or fewer, long headlines of 90 or fewer, and descriptions of 90 or fewer. Base them on the products and offers on its landing page, and say which existing assets each draft replaces.

Show me a preview of the text asset changes. Wait for my approval before updating the asset group.

Account audits

Sync Microsoft Ads negatives with Google Ads and cut Bing waste

Wasteful Bing search terms plus the Google Ads negatives missing from Microsoft, merged into one previewed negative keyword batch.

My Microsoft Ads account [Microsoft account] was imported from Google Ads [Google account], but the negative keywords have drifted apart since then. Clean it up using the last 30 days of data.

1. Pull the Microsoft search terms report. Find terms that spent more than [min spend] with zero conversions.
2. List the current Microsoft negatives per campaign.
3. List the Google Ads negatives at every level: campaign, ad group and shared negative keyword lists.
4. Identify the Google negatives that are missing from the matching Microsoft campaigns. Match campaigns by name.

Deliver three sections:
- Section A: wasteful Microsoft search terms, with cost, clicks, matched keyword and match type.
- Section B: Google negatives missing in Microsoft, by campaign.
- Section C: one combined, de-duplicated list of proposed Microsoft negatives, with match type and level (campaign or ad group), plus the estimated monthly savings.

Never propose a negative that contains our brand or a product name we sell. Flag these instead.

Show me a preview of section C and wait for my approval before adding any negatives to Microsoft Ads.

Account audits

Meta learning phase audit and ad set consolidation plan

A table of every active ad set's learning risk and a consolidation plan so each one gets enough leads to exit learning.

Audit the learning phase in my Meta ad account [ad account] for the last 7 days.

For every active ad set, list:
- campaign and ad set
- optimization event and daily budget
- results and cost per result in the last 7 days
- learning status, if Meta returns it

If Meta does not return a learning status, estimate it. Fewer than 50 optimization events in 7 days means the ad set will likely stay in Learning Limited.

Then check the change history for the same 7 days. Flag edits that reset learning: budget changes above roughly 20%, targeting edits, optimization event changes and newly added ads.

Deliver:
1. A table sorted by spend.
2. A consolidation plan. Show which ad sets that target the same city and the same service should be merged, so each remaining ad set gets enough weekly leads at my target cost per lead of [target CPL]. Show the expected weekly results per merged ad set.

Do not change anything yet. For every pause, budget move or duplication you recommend, show the exact preview and wait for my approval.

Account audits

Meta creative fatigue scoreboard with refresh drafts

Scores every active Meta ad for fatigue in weekly buckets, labels it Keep, Refresh or Pause, and drafts new copy that replaces the creative on the same ad after your approval.

Run a creative fatigue check on all active ads in [ad account]. Look at the last 28 days in weekly buckets (7-day increments).

For each ad, pull these metrics per week:
- frequency and reach
- link click-through rate
- CPM
- leads and cost per lead (use the lead action types the account actually records, and tell me which one you counted)

Mark an ad as fatigued if its frequency in the latest week is above 3 and at least one of these is true:
- link CTR dropped 30% or more versus the ad's first week
- cost per lead rose 25% or more versus the ad's first week

If an ad has too few leads in a week to judge cost per lead, say so and judge it on CTR alone.

Deliver a ranked list, worst first, with one verdict per ad:
- Keep: healthy.
- Refresh: fatigued but still converting.
- Pause: fatigued and cost per lead above [target CPL].

For each Refresh ad:
1. Read the current creative (primary text, headline, image or video, button, link).
2. Write 2 new angles for the primary text and headline, aimed at local customers who need [service] (urgency, price clarity, reviews, same-day availability).
3. Show the current and new copy side by side, then show the creative replacement as a preview that keeps the existing image, button and link.

Apply new creative by replacing it on the existing ad, so the ad keeps its ID and learning history. Show me each replacement and each pause as a preview, and apply nothing until I approve. After an approved replacement is live, render a feed and story preview of the new creative so I can see how it looks.

Account audits

Meta audience saturation check and budget shift to Google

A verdict on whether your local Meta audience is saturated, compared with Google Ads and Business Profile call trends, plus a previewed budget shift.

My service area is small, and I think Meta has run out of new people to reach. Compare channels for the last 30 days versus the previous 30 days.

1. Meta saturation check
For each active ad set, pull reach, frequency, CPM, leads and cost per lead. Compare reach with the estimated audience size from the delivery estimate.

Call the ad set saturated when all three are true:
- 30-day reach is above 50% of the estimated audience
- frequency is above 4
- cost per lead rose more than 20%

2. Google Ads comparison
For my Search campaigns, pull conversions and cost per conversion. Run an impression share analysis and note lost impression share due to budget.

3. Business Profile
Pull calls, direction requests and website clicks for [location], so organic local demand is visible next to the paid channels.

Deliver:
- One table per channel.
- A clear verdict on saturation.
- A recommendation on reallocation. If Meta is saturated, propose moving up to [percent]% of the saturated ad sets' budget to Google Search campaigns that meet both conditions: budget-lost impression share above 15%, and cost per conversion below Meta's cost per lead.

Show both budget changes, Meta and Google, as previews. Apply nothing until I approve.

Account audits

Which Meta edits reset learning and raised my cost per lead

Read-only timeline that matches every learning-sensitive edit in the Meta activity history to the ad set's cost per lead before and after, and turns the costliest patterns into change rules for your team.

My cost per lead on Meta jumps every few weeks, and I suspect our own edits are the cause. Build an edit-impact timeline for [ad account] over the last 30 days. This task is read-only.

1. Pull the activity history and keep only edits that can reset learning:
- budget changes above 20%
- bid or bid strategy changes
- targeting or placement changes
- optimization event changes
- ads added, paused or replaced
- ad sets that were paused and then resumed
For each kept edit, record who made it and when. Use the old and new values where the history provides them. If a budget change has no old and new value, keep it, mark the size as unknown and do not guess the percentage.

2. List the ad sets and their ads so each edit maps to the right ad set, then pull daily spend, leads and cost per lead for each affected ad set.

3. For each edit, compare two windows:
- cost per lead in the 3 days before the edit
- cost per lead in days 1 to 7 after the edit
Count it as harmful if cost per lead rose more than 25% and did not recover within 7 days. If either window has fewer than [minimum leads] leads, or another edit hit the same ad set inside the window, mark the verdict as inconclusive instead of harmful.

Deliver:
- A chronological table: date, user, ad set, edit (with old and new values where known), before vs after cost per lead, verdict.
- The 3 most expensive edit patterns.
- A short set of change rules for my team, for example: batch edits on one day, change budgets in steps of 20% or less at most every 3 to 4 days, and avoid editing an ad set that has not yet exited learning (check its status in Ads Manager before editing, since the history does not show it).

Account audits

Diagnose low Quality Score keywords and the component behind each

A spend-ranked list of low Quality Score keywords showing when the score dropped, which component is weak and the specific fix.

Diagnose the low Quality Score keywords in my Google Ads account [account name]. Use the last 30 days.

1. List active keywords with a Quality Score of 5 or lower that spent at least [amount]. Sort by cost, highest first.
2. For each one, show the three components: expected CTR, ad relevance, landing page experience. Use the Quality Score history to tell me when the score dropped and which component caused it.
3. If ad relevance is below average, read the RSAs in that ad group and check whether the keyword or a close variant actually appears in the headlines.
4. If landing page experience is below average, run Core Web Vitals and a page audit on the final URL and report the main problems (speed, mobile, message mismatch).

Deliverable: a table with keyword, ad group, cost, QS now vs 90 days ago, weakest component, root cause and one concrete fix (move to its own ad group, rewrite headlines, change landing page, fix page speed).

This is a diagnosis only. Do not change anything in the account.

Account audits

Weekly Meta health triage across all client ad accounts

A red/amber/green triage table of every connected Meta ad account, with a short diagnosis and the next audit to run for each red account.

I manage several client Meta ad accounts. Run a weekly health triage across all of them.

1. List every connected Meta ad account.
2. For each account, compare the last 7 days with the previous 7 days on these metrics: spend, CPM, frequency, link CTR, results, cost per result.
3. Rate each account:
- Red: cost per result up more than 25%, OR spend down more than 30% with no budget change in the change history, OR frequency above 4 on prospecting campaigns.
- Amber: cost per result up 10-25%, OR CPM up more than 20%.
- Green: everything else.
4. For each red account, drill into the 3 ad sets that drive most of the change. Classify the likely problem as learning reset, creative fatigue, audience overlap or under-delivery, and cite the metric that points to it.

Deliverable:
- One triage table sorted red, amber, green, with columns: account, spend, cost per result change %, CPM change %, frequency, rating.
- For each red account: a two-line diagnosis and the specific audit to run next.

Read-only. Do not change anything in any account.

Account audits

Meta change history audit: which edits reset learning and what it cost

Reads 30 days of Meta change history, classifies edits that likely restarted learning, and compares each ad set's 7 days before and after, flagging overlapping edits. Delivers an edit table, a team summary and house rules. Read-only.

Audit the edits made in Meta ad account [ad account] over the last 30 days and measure their effect on learning and performance.

1. From the change history, list every significant change with who made it and when. Include:
- budget changes (flag any single step above 20%)
- bid or cost cap changes
- targeting edits
- optimization event changes
- new ads added to live ad sets
- ad sets paused for more than 7 days and then restarted (derive this from the status changes)
Meta does not report a learning reset directly. Mark "likely learning reset" from the edit type and size: optimization event changes, targeting edits, bid strategy changes, new ads, budget steps above 20% and restarts after a long pause count as likely; small budget steps and renames do not.

2. For each significant edit, pull the ad set's daily results and cost per result for the 7 days before and the 7 days after the edit, using the action that matches its optimization event.

3. Data quality checks before comparing:
- Confounded: another significant edit on the same ad set (or its campaign budget) inside either 7-day window. Mark it confounded and treat the comparison as indicative only.
- Incomplete: the edit is less than 7 days old, so the after window is short. Mark it and compare on the days available.
- Low volume: fewer than [minimum results] results in either window. Mark it and do not draw a conclusion.

Deliverable:
- A table with these columns: date, person, ad set, change (old value to new value), likely learning reset (yes / no), results before, results after, cost per result before, cost per result after, data quality flag.
- A summary by team member: number of likely resetting edits and their average change in cost per result, calculated on clean comparisons only. Show how many edits were left out as confounded, incomplete or low volume.
- 4 to 6 house rules for this account based on what the data shows, such as maximum budget step, batching edits into one change, minimum days between edits, and when to duplicate instead of editing.

This is read-only. Do not modify anything in the account.

Account audits

Vet TikTok's own recommendations against real ROAS

Checks every TikTok recommendation against 30 days of real ROAS, gives an accept, test or reject verdict and previews approved changes through the matching update tool.

Pull TikTok's current optimization recommendations for ad account [account name] with tiktok_recommendations, at ACCOUNT, CAMPAIGN and ADGROUP level. Then check each one against the last 30 days of data for the campaign or ad group it affects.

Match each recommendation to its entity by the campaign or ad group id it carries, using list_tiktok_campaigns and list_tiktok_adgroups. If a recommendation cannot be mapped to a specific entity, list it as "unmapped" and judge it on its own merit.

Get performance with tiktok_report_query (BASIC, AUCTION_CAMPAIGN or AUCTION_ADGROUP): spend, purchases, purchase value and ROAS.

For every recommendation, show: type (budget, bidding, targeting, creative, setup), affected entity, what TikTok suggests, that entity's spend, purchases, purchase value, ROAS and your verdict.

Decision rule, with my target ROAS of [target ROAS]:
- Accept: budget or bid increases only where ROAS is above target and there are at least 15 purchases in 30 days. Setup fixes (pixel, event, missing creatives) are accepted regardless.
- Test: the entity is within 15% of target, or has fewer than 15 purchases.
- Reject: budget increases or broader targeting on entities below target, and any change that would switch optimization away from purchase.

Deliverable: a table sorted by verdict, then the 3 recommendations with the highest expected impact on revenue.

There is no one-click "apply recommendation" tool, so for each recommendation I approve, name the tool you would use and show the exact change (old value, new value):
- campaign budget: update_tiktok_campaign_budget
- ad group budget: update_tiktok_adgroup_budget
- bid, locations, ages, interests, audiences or schedule: update_tiktok_adgroup
- pausing ads or ad groups: update_tiktok_ad_status or update_tiktok_adgroup_status

Budget changes are capped at +50% per change and +100% per 24 hours, so split larger increases into steps. Setup fixes such as installing a pixel or uploading new creatives cannot be applied through these tools. List them as manual tasks.

Run every change with confirm=false first and show me the preview. Wait for my confirmation before applying anything.

Account audits

Tag Manager container audit for lead conversion tags

A findings table covering live version, unpublished changes, mismatched conversion IDs and missing lead triggers in GTM.

Audit the Google Tag Manager container [GTM-XXXX] that handles lead tracking on my website.

1. Show the live version number and its publish date. Also show anything pending in the workspaces that was never published.
2. List all tags with their type, firing triggers and paused status. Highlight Google Ads conversion tags, GA4 event tags and Meta pixel tags.
3. For each Google Ads conversion tag, compare its Conversion ID and label with the send_to values of the conversion actions in my Google Ads account. Flag mismatches and any tags pointing to removed actions.
4. Check the triggers for lead events: form submit, thank-you page view, tel: link clicks and WhatsApp clicks. Flag lead events that have no tag, and any conversion tag that fires on All Pages.

Deliver a findings table with four columns: item, issue, risk, fix. For the fixes, prepare the edits in a new workspace and show me exactly what will change. Do not publish until I approve. Also tell me which version I can roll back to if the publish causes problems.

Account audits

Diagnose a sudden drop in conversions while clicks held steady

A timeline and ranked causes for a conversion drop, built from Ads data, change history, GTM versions and a live form test.

Conversions in Google Ads [account name] dropped sharply around [date], while clicks stayed stable. Find the cause.

1. Show daily clicks, conversions and conversion rate per conversion action, for the 14 days before and after that date.
2. Run a conversion tracking health check and list any conversion action that stopped recording.
3. List the Google Ads change history for the same window. Focus on edits to conversion actions, goals and campaigns.
4. Check the GTM version history. Was a new version published near that date, and what did it change? Also check for pending, unpublished edits.
5. Check where my WordPress site loads its tracking. Then list the contact forms and their recipient email.
6. Before running it, tell me which form you will test, because a test submission creates a fake lead. After I confirm, submit one test entry to the main contact form and report whether the notification email was sent.

Deliver a timeline of events and a ranked list of likely causes, each with its evidence. If the fix is a GTM rollback or a settings change, show it as a preview and wait for my approval.

Account audits

Fix rejected Meta ads without losing learning

Finds disapproved or limited Meta ads, reads Meta's stated reason and the creative, sizes the spend at stake and swaps in compliant creative in place after approval.

In Meta ad account [account name], find every ad in an active campaign whose status shows it was disapproved or has delivery issues. Use the issue details Meta returns for each ad (error code and message) as the stated reason.

For each ad:
1. Read the creative: primary text, headline, description, image or video, button, destination link, Page and Instagram account, and which Advantage+ creative enhancements are on. Note whether the ad was built from an existing Instagram post, because then only the button and message can change, not the copy.
2. Check the activity history for the last 14 days and note what changed on that ad or ad set shortly before the rejection.
3. Pull daily spend for the ad over the last 30 days and take the last day with spend as the day it stopped delivering. Then report spend, results and cost per result for the 14 days before that day, so I can see what is at stake. If the ad never spent, say so.
4. State the cause. Quote Meta's message where it gives one. Where it is generic, name the most likely policy cause and mark it as your inference. Common ones are personal attributes ("you" plus a condition), before/after imagery, unrealistic claims, a landing page that does not match the ad, and restricted product wording. Also say if an Advantage+ enhancement could have added the problem, such as generated text or touched up imagery.

Deliverable: a table with these columns: ad, campaign, spend at stake, Meta's reason, likely cause (stated or inferred), current text, proposed compliant text, enhancements to switch off.

Then, for each ad, show me the in-place creative replacement as a preview with current and proposed values side by side, and render a preview of the current ad next to it. After I approve, replace the creative in place so the ad keeps its ID, learning phase and history, switch off any enhancement we agreed on, and render a preview of the new creative. Remember that editing the creative sends the ad back to review. Do not delete any ad or ad set, and do not apply anything before my approval.

Account audits

Check whether invalid clicks are inflating Shopping and Search costs

Scans the last 14 days of ad clicks for fraud, estimates the cost of suspicious IPs, checks free IP exclusion slots and previews a block list per campaign for your approval.

I suspect invalid clicks are inflating costs in my Shopping and Search campaigns in [account name]. Check this using the last 14 days.

1. Run the fraud scan for the last 14 days. If the fraud tag is not set up yet, stop and walk me through the setup for my domain [domain] instead. Keep in mind the tag only sees clicks that land on tagged pages, so tell me whether my Shopping product pages are covered.
2. List IPs with a fraud score of 70 or higher, with these columns: score, reasons (repeat clicks, velocity, datacenter, Tor, bot user agent), click count and the campaigns they hit. Where the scan does not say which campaign an IP clicked, look up that IP's individual clicks to find out.
3. List my Shopping and Search campaigns and pull their daily performance for the same 14 days. Point out days where CTR spiked while conversion rate dropped, and check whether those days line up with the clicks from the suspicious IPs.
4. List my existing IP exclusions per campaign, and how many slots remain under Google's limit of 500 IPs per campaign.

Deliver:
- A table of IPs recommended for blocking, with the campaigns to block them on.
- The estimated cost of the clicks from those IPs: each IP's click count multiplied by the average CPC of the campaign it hit over the same period. Label it clearly as an estimate.
- A short verdict: is invalid traffic material here (more than 5% of spend) or just noise?

Leave out any IP the scan shows with a conversion. If the scan does not report conversions per IP, say so and mark the IPs I should double check against my orders before blocking. Never include our office IP [office IP].

Show me a preview of the block list per campaign. Wait for my approval before excluding anything.

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

Connect your accounts free