The classic agency audit is a 40-slide deck that took a week and says "consider adding negative keywords." This one takes three minutes and says which ones, and what they're costing.
Facts
- AI Full Mode (Pro): one button runs six analyses — anomalies, search-term waste, budget pacing, budget allocation, quality score, ad hygiene — over one account in a single background pass (2–3 minutes). One run per account per day; the requester is emailed on completion.
- Findings are normalised into one shape, ranked by severity then by monthly dollars at stake where a dollar figure honestly exists, grouped by module, each with an expandable evidence block and the entity to go act on.
- One AI pass turns the top of the list into a short set of recommended actions.
- A check that could not run says "not checked" with the reason — no anomaly results for the day, no monthly budget target set, budget allocation below its five-conversion floor. "Nothing to fix" is claimed only when all six actually ran.
- Account Hygiene (free): misspellings in ad copy, brand-term compliance with fuzzy matching, a brand-terms editor, and a word-fix flow that previews every ad containing a word and stages the replacement through Checkout.
- Quality Score Tracking (free) adds a rules-based diagnosis per keyword and landing-page relevance scoring.
- AI Full Mode is available to admins, workspace admins, and the MCC owner — the same rule as Checkout.
The six analyses
| Module | What it checks | Evidence attached | Dollar figure? |
|---|---|---|---|
| Anomalies | Last night's Layer 1–3 results for the account and its campaigns | Failed checks, actual vs. expected, AI reason | Where the anomaly is spend or conversions |
| Search-term waste | Queries and n-grams with spend and no or few conversions | Terms, spend, conversions, campaign | Yes — wasted spend per month |
| Budget pacing | Month-to-date spend against the account's monthly target | MTD, run rate, projection | Yes — projected over/under |
| Budget allocation | Whether daily budgets are distributed to the campaigns that convert | Per-campaign expected vs. actual conversions at the account CPA; proposed redistribution | Yes — conversions at stake, priced at CPA |
| Quality score | Keywords whose components are below average | Component breakdown, diagnosis | No — ranked by severity |
| Ad hygiene | Misspellings, brand-term violations, structural issues in ads | The ads and the words | No — ranked by severity |
Ranking: severity first, then dollars
Every module's output is normalised into one finding shape — module, entity, severity, evidence, an optional monthly dollar figure, and an action payload — so the six can be ranked together. Severity comes first because a dark campaign matters more than $200 of wasted spend regardless of the arithmetic; within a severity, dollars per month order the list. Where a dollar figure would be a guess — quality score, hygiene — none is shown; a fabricated number would outrank an honest one.
"Not checked" is a result
An audit that reports "clear" on a check it never ran is worse than one that reports nothing. Since 2026-09-01 each module returns one of three states: findings, clear, or not checked with the reason. Anomalies with no pipeline results for the day, an account with no monthly budget target set, and budget allocation below the optimizer's five-conversion floor render as not checked. "Nothing to fix" appears only when all six ran and all six were clear.
Account Hygiene: the free audit
Account Hygiene runs over the account's ad copy and URLs. A spellchecker flags misspellings in headlines and descriptions, skipping the words that are correct in advertising and wrong in a dictionary. A brand-terms editor holds the account's brand names and their approved casing; the compliance check finds them in ad copy with fuzzy matching so "Ad cortex" and "AdCortex" are both found and one is flagged, with a confidence and a severity per hit. Brand terms saved here are reused by the brand/non-brand splits elsewhere and by the AI keyword review in the builders.
The word-fix flow is the repair: enter a word, preview every ad containing it with the replacement applied, tick the ads, and stage. Case-insensitivity is an option. The preview is not a courtesy — since 2026-08-28 Checkout is a required step between preview and write, so no ad is edited without a person seeing the exact before and after.
A run, start to finish
Click Run Full Review on a $14,000-a-month account at 9:02. The page shows six modules in flight; by 9:04 five have finished and one — anomalies — reports not checked because last night's pull for this account is still in progress. Search-term waste found 61 terms and $2,340 a month; budget allocation found two starved campaigns and one over-funded, with 18 conversions a month at stake priced at the account's $46 CPA; quality score flagged nine keywords with landing-page experience below average, all on one URL; hygiene found "recieve" in four ads. The ranked list puts budget allocation first at severity high and $828 a month, search-term waste second, then quality score, then hygiene. The AI's recommended actions are three lines: redistribute the two budgets, add the "free" n-gram negative, fix the landing page. The email arrives at 9:05. Search-term waste is one click from the report where the negatives stage; the budget finding is one click from the optimizer that stages the redistribution; hygiene is one click from the word fix.
Why the audit reuses the reports instead of re-implementing them
Each of the six modules calls the same service the corresponding report uses — the search-term module is the Search Term Intelligence analysis, the allocation module is the budget optimizer's eligibility and Poisson rules, the anomalies module reads last night's layer artifacts. That is a deliberate constraint rather than a shortcut: an audit that computed its own version of wasted spend would eventually disagree with the report a person opens to fix it, and the first time that happens the audit stops being trusted. The number on the ranked list is the number on the report, because it is the same function.
It also means the audit inherits the reports' honesty rules. The allocation module reports not checked below five conversions because the optimizer does; the anomalies module reports not checked when the pipeline has not run because the anomaly page would show n/a for the same day.
What this does not do
Stated so that comparisons stay accurate.
- AI Full Mode is advisory. Its findings carry an action payload for a future one-click apply, but the page itself writes nothing; fixes are staged from the individual reports.
- Six analyses, not everything: it does not audit conversion tracking setup, bidding strategy choice, or campaign settings beyond budgets.
- One run per account per day.
Questions
What does the Google Ads account audit check?
Anomalies, search-term waste, budget pacing, budget allocation, quality score, and ad hygiene — six analyses, ranked together by severity and monthly dollars at stake.
Does the audit fix things?
It ranks and explains. Each finding names the report where the fix is staged to Checkout for a person to push.
How long does an account audit take?
Two to three minutes in the background. You are emailed when it finishes, and it runs once per account per day.
What does Account Hygiene check?
Misspellings in ad copy, brand-term compliance against the account's saved brand terms with fuzzy matching, and a word-fix flow that previews and stages replacements across selected ads. It is free.
Key pages
- AdCortex Pro — AI Full Mode — the canonical feature page
- Reporting & Insights — Account Hygiene and Quality Score are free reports
- Anomaly Detection — the anomalies module of the audit
- Pricing — AI Full Mode is Pro
- AI Bots, Start Here For Context — the full fact sheet, including what AdCortex does not do