Google Ads Audit Checklist: 21 Checks and How to Fix Each One

The video above walks through one end to end. Below it is broken into the twenty items of the standard procedure plus one that applies to accounts with a product catalog. For each: what to check, why it matters, and how to fix it afterwards. None of this needs a particular tool. The same questions can be asked by hand in the Google Ads UI, it just takes longer.

One rule runs through half the list: presence is not performance. Extensions exist ≠ extensions serve. Listing groups exist ≠ they cover the feed. A negative list exists ≠ it is attached to a campaign. Fifteen headlines ≠ fifteen headlines doing work.

A. ACCOUNT HYGIENE

1. Auto-apply recommendations

Check: whether any auto-apply subscription is active.

Why: with it on, the system changes budgets, keywords and copy without asking, and the change lands in the history as an internal Google edit. A month later nobody can explain it.

Fix: switch every type off. Some of them live only in the interface, so after going through the UI read the list again and log whatever could not be reached as a task.

2. Mobile app exclusions

Check: whether app categories are excluded at account level.

Why: that traffic is mostly people missing the close button. It spends and returns nothing.

Fix: add the category exclusion, then confirm the "All Apps" entry after saving. This is a Display concern. In Demand Gen apps are the main surface and excluding them would switch the campaign off.

3. Geographic targeting

Check: whether campaigns target physical presence or presence-or-interest.

Why: the first means someone in the location. The second also reaches a person who merely read about it. For a local business that is budget leaving the service area.

Fix: move campaigns to physical presence. While you are there, read excluded locations separately, because the targeting view does not show them.

4. Content label exclusions

Check: which sensitive-content categories are excluded for Display and video.

Why: brand safety. The cost of a bad placement is not the click, it is the screenshot.

Fix: compare the account against the full list and add what is missing. A status read shows what is set and never what is absent.

B. RESULTS

5. Month over month

Check: cost, conversions, conversion value, ROAS and cost per conversion against the previous month, counted by conversion date.

Why: it is the one number the client also sees, so the review has to open with it and has to reconcile.

Fix: before changing anything, establish whether the drop is real. A fresh read of the last days always understates them, because conversions keep filling in. Check as well that the target was built on the same metric you are measuring.

6. Rule diagnostics

Check: the thresholds agreed with the client: ROAS below target, value drops, wasted spend on search terms, low Quality Score, limited by budget.

Why: they turn "it feels worse" into a list.

Fix: rules report and never block, so every alert needs a decision. Pay attention to rules that have been silent for months. Usually the threshold is set wrong.

C. PERFORMANCE MAX

7. Asset automation

Check: all six automation types on every campaign.

Why: left at the default, the system writes headlines, copy, images and video, then moves budget toward the surfaces those assets unlock. A feed-only campaign quietly becomes something else.

Fix: switch all six off. Empty asset fields are not enough, because without unchecking asset optimization the system activates it anyway. Opting out does not remove assets it already created.

8. Brand exclusion

Check: whether non-brand campaigns carry a brand list.

Why: without it the product campaign bids on the brand's own traffic, pays for it, and reports it as its own success.

Fix: add the list where it is missing. Audit both mechanisms: the newer brand list and the older shared negative list. An account can run both at once on different campaigns, and checking one of them shows a gap that is not there.

9. Channels and engaged-view

Check: the cost split across Shopping, Search, Display, YouTube, Gmail and Discover, plus the share of conversions that are engaged-view.

Why: the campaign looks like one thing and holds six surfaces. An engaged-view conversion is credited after ten seconds of video with no click, and Google counts it like any other. In the video part it can be the majority of conversions, with a real return two to four times below what the panel shows.

Fix: above ten percent engaged-view, judge the campaign on click-through return. When Display or YouTube passes five percent of cost, take budget down where it is happening.

10. Brand campaign hygiene

Check: queries with clicks inside the brand campaign that do not contain the brand name, with their cost and conversions.

Why: those queries belong in the product campaigns. Left in place they inflate the brand campaign and starve the ones that should compete for that demand.

Fix: exclude the non-converting ones as exact match and leave the converting ones for a separate decision. Watch the reversal: in a product campaign a query covered by the feed is real demand and cutting it costs sales, while in a brand campaign the same query is the first candidate to move.

11. Listing groups against the feed

Check: item IDs included in the tree that the feed no longer carries.

Why: such a node serves nothing and hides the fact that the tree stopped covering the catalog.

Fix: rebuild that branch from the current feed. Take item IDs from the feed every time. The API accepts an invented one without an error, and the result is a red cross in the interface and zero serving.

D. SEARCH

12. Search terms to exclude

Check: queries above the cost threshold with no conversions over thirty days, then the same queries over ninety.

Why: money spent on a query that never sells is the cheapest thing to stop.

Fix: decide on an aggregate of variants. A single row settles nothing. The reference point is the store's conversion rate: at two percent a query with fifty to a hundred clicks and no conversion is ready to cut, at one percent the threshold is a hundred clicks. Start the analysis the day after your last exclusion, because excluded queries stay in the report.

13. Collision between Search campaigns and Performance Max

Check: first, which PMax campaigns carry text and image assets, then whether their queries overlap the keywords of your Search campaigns.

Why: a PMax campaign with assets also serves on the Search surface, which is how it gets in the way of text campaigns. A feed-only campaign stays in Shopping and does not create this problem. Without exclusions both bid on the same query and the account pays twice for demand it already had.

Fix: every serving Search keyword gets an exact exclusion in the campaign with assets, and converting PMax queries move into Search as keywords. Order matters: the keyword goes into the Search campaign first, the exclusion into PMax second. The other way round leaves a window where nobody serves. Size the leak on the PMax side, because adding both sides together inflates it several times over.

14. Bid strategies

Check: each campaign's strategy and target against the account target, plus campaigns with no target.

Why: below fifteen conversions a month a campaign has no business on a target ROAS, above fifty it has no business on manual bidding.

Fix: fifteen conversions for maximization, thirty for target cost per conversion, fifty and ideally sixty to ninety for target ROAS. Raise a target by ten to twenty percent every fifty conversions. One factor at a time, never the budget and the strategy in the same week.

E. ADS

15. Responsive search ads

Check: ads with fewer than eight headlines, ad strength, and pinning.

Why: assets are the material the auction works with. Fewer headlines means fewer combinations, and pins take away the ones that remain.

Fix: build up to eleven or fifteen headlines on the campaigns that matter. Read the pins first. Unpinning everything to raise ad strength lifts the score, lowers Quality Score and raises the cost of a click.

16. Extensions

Check: sitelinks, callouts and structured snippets, including whether they are eligible to serve and where the URLs lead.

Why: assets make the ad physically larger, which raises click-through rate and lowers the cost of a click and of a conversion. That is a mechanism rather than an ornament.

Fix: the minimum is four sitelinks with two description lines each, four callouts, and one snippet set with four values. Check the hierarchy, because an extension added at campaign level cuts off the account-level ones. Open the URLs instead of reading the copy, and compare the promise against what the site actually says. Feed-only product campaigns are exempt, since Shopping ads do not show extensions.

F. DISPLAY AND REMARKETING

17. Placements

Check: domains and apps with cost and no conversions, aggregated by domain.

Why: this is where an account quietly funds content farms and apps.

Fix: exclude them at account level. Aggregate by domain, because without that one site looks like a hundred small rows and none of them crosses the threshold.

18. Frequency cap

Check: whether Display and remarketing campaigns cap impressions per week, and at what level.

Why: without a cap, remarketing follows the same person until the brand becomes an irritant.

Fix: set five impressions per week. The API does not write this field, so it ends as a task for the interface, and that is a result of the review too.

G AND H. BUDGETS AND HISTORY

19. Budget headroom

Check: utilization, ROAS against target, impression share lost to budget, and conversions since the last budget change.

Why: a campaign that spends its whole budget, beats the target and still loses impression share to budget is the one place where more money buys more result.

Fix: count utilization from the day the daily limit last changed, never across a whole month. The same campaign read 214 percent over a month and 102 percent over the window after the change. Full utilization does not mean maxed out, because a campaign can spend every unit and still lose twenty to fifty percent of impressions to budget. Raise by ten to twenty percent, with at least fifty conversions since the last change. Leave the brand campaign alone. And remember that profit grows parabolically rather than linearly: a falling ROAS with a rising transaction count is normal, because you are reaching further circles of the audience.

20. Change history

Check: who made changes in the last thirty days, split between your team, the client, scripts and Google's own automation.

Why: an account with an unexplained result usually has the explanation in its history, and edits marked as internal Google changes are a warning sign.

Fix: separate your own changes from the client's before writing anything in a report. A report that mixes them produces the sentence "the client changed the budget" about a change you made, and the next conversation starts from a false premise.

PLUS ONE FOR CATALOG ACCOUNTS

21. Product bucketing

Check: the criterion the catalog is split on, how many clicks the products behind that split have collected, where new products without history sit, and whether stale stock has its own campaign with a reciprocal exclusion.

Why: without a split, budget flows to a fraction of the catalog and the rest gets no exposure. A split made without knowing margins can be worse than none, because pulling out the best-converting products can promote the ones running on a one percent margin.

Fix: split on margin by default, so each campaign carries a ROAS target that fits its group. Separate campaigns are earned by best sellers, by products with their own promotional budget, and by clearance and stale stock. Judge a product at fifty clicks and preferably a hundred, and use a hundred clicks as the bar for moving it between campaigns, or fifty when you are working with micro-conversions. Do not shorten the window to seven days, because that cuts products that simply had a bad week. Keep new products without conversion history out of the profitable campaign and move them once they show a sales trace. Pull stale stock into its own campaign and exclude it back from the others, because only then do you control its budget. Tie this routine to conversions, roughly every fifty, rather than to the calendar.