What makes a 2026 search-terms audit different from an older one?

The old audit assumed the search terms report was a near-complete ledger and that match types held. Neither is true now. AI Max and broad match expand one keyword into a long tail of queries, most of which fall below the reporting threshold and never appear as rows, so the report itemises a shrinking share of spend. And matching runs on inferred intent rather than the literal string, so a query’s match type no longer tells you what triggered it. A current audit therefore starts by measuring how much of the account you can even see, checks which AI features are silently enabled, and reads query relevance and cost instead of trusting keyword-level attribution. It is a hygiene audit of where the money actually goes, not a checklist of keyword settings.

What should I check first in a search-hygiene audit?

Start with visibility, because it determines whether the rest of the audit is even valid. Compute the visible-spend share per campaign: sum the cost of every row in the search terms report and divide by the campaign total. If you can see 70%+ of spend, the report is a usable to-do list and term-by-term mining will pay. If it is under 30%, most of the money is in queries you cannot read, and no amount of single-query negating will reach it — the fix is upstream in match types and structure. Doing this first stops you spending an hour mining a report that is structurally blind, which is the single most common wasted effort in a 2026 audit.

Which AI Max settings should the audit confirm are on or off?

Confirm three expansion settings per campaign: search-term matching (AI Max’s query expansion), text customization (formerly Automatically Created Assets), and final URL expansion. The September 2026 auto-upgrade switched search-term matching and text customization on by default for campaigns using campaign-level broad match or Automatically Created Assets, so an account can be running AI Max expansion without anyone having chosen it. Final URL expansion matters for hygiene because it can send clicks to pages you did not target, which shows up as off-theme queries converting oddly. Note which of the three is enabled on each campaign before you read the search terms — the settings explain the queries.

What thresholds flag wasted spend during the audit?

Use a few practitioner rules of thumb. If more than 20 to 25 percent of the queries in a 30-day search terms pull are off-theme, match types and negatives need serious work. If visible spend is under 30 percent of campaign cost, the report is too blind to mine and the fix is structural. For an individual query, weigh accumulated cost against relevance rather than waiting for a fixed click count: an expensive off-theme query is waste on the first pound, while a cheap borderline one can wait. These are calibration points, not Google metrics — the value is that they convert ‘this looks bad’ into a specific action on a specific campaign.

Which old audit checks no longer matter?

Drop the checks that assumed keyword-level control you no longer have. Auditing exact-match keywords for tight coverage is close to moot when AI Max treats every keyword as broad and expands past it anyway. Counting negatives, or chasing a ‘complete’ negative list term by term, is theatre on a low-visibility campaign where the wasteful queries never appear as rows. And judging a campaign purely on its self-reported conversion column misleads, because AI Max borrows credit from keywords it displaced. Replace all three with the modern equivalents: measure visible share, negate on cost and relevance and patterns, and judge performance on incremental account-level results rather than a campaign’s own attribution.

The 2026 Search-Hygiene Audit, in Order

A search-terms audit run the way it was in 2022 now gives you false confidence. You tick through exact-match coverage, count the negatives, confirm the conversion column looks healthy, and conclude the account is clean — while AI Max quietly expands into queries none of those checks can see. Search Engine Land put the shift bluntly: Google Search Ads in 2026 require a different kind of audit. The mechanics that made the old checklist meaningful — a near-complete search terms report and match types that held — are both gone.

This is the audit rebuilt for that reality, in the order the steps actually depend on each other. It is a hygiene audit — where the money goes, what it goes to, and whether you can even see it — not a full account review of bids and creative. Each step has a specific tab, a filter, and a threshold so you finish with actions rather than impressions. It pulls together the single-metric pieces already in the library — the visibility ratio and the post-migration negative audit — into one sequence you can run end to end in an afternoon.

Step 1: Measure what you can see, before anything else

Start with visibility, because it decides whether every later step is valid. For each campaign you manage, open the search terms report over a 30-day window, sum the cost column across every visible row, and divide by the campaign’s total cost for the same window. That percentage is the share of spend you can actually itemise and negate. Do it per campaign, never account-wide — a healthy brand campaign at 80% will mask a broad-match prospecting campaign at 20% if you blend them, and the blend hides the exact campaign you most need to fix.

The number routes the rest of the audit. Above roughly 70%, the report is a usable to-do list and term-by-term mining in later steps will pay. Between 30% and 70%, mine but pair it with upstream fixes. Below 30%, stop — the report is structurally blind, the wasteful queries live in the anonymous “Other search terms” bucket, and any hour spent adding single-query negatives is aimed at the fraction you can see while the expensive majority stays invisible. Running this step first is what stops the most common wasted effort in a modern audit: diligently mining a report that cannot show you the problem. The full method is in the visibility ratio audit.

Step 2: Confirm which AI features are silently enabled

Before you read a single query, find out what is generating them. Three expansion settings change what shows up in the search terms report, and after the September 2026 auto-upgrade an account can be running them without anyone having chosen to. Check each campaign for search-term matching (AI Max’s query expansion), text customization (formerly Automatically Created Assets), and final URL expansion. The auto-upgrade switched search-term matching and text customization on by default for campaigns that used campaign-level broad match or Automatically Created Assets, so “we never turned on AI Max” is not evidence it is off.

These settings are the explanation for the queries you are about to audit, which is why they come before the query read and not after. Search-term matching is why off-theme queries appear at all; final URL expansion is why a query can convert against a page you never pointed it at, showing up as an odd relevance-to-conversion mismatch later in the audit. Note the state of all three per campaign. If a campaign was auto-upgraded and you have not reviewed it since, treat its search terms as unaudited regardless of when you last looked — the matching changed underneath the old review. The scope of what moved is covered in the AI Max auto-upgrade scope check.

Step 3: Read queries for relevance and cost, not attribution

Now read the report — sorted by cost descending, judging each query on relevance, not on the conversions it claims. In a 2026 account the conversion column is the least trustworthy figure, because AI Max treats every keyword as broad and takes credit for conversions your existing exact and phrase terms would have won. Cost, by contrast, is money that irreversibly left the account, and the query string tells you whether that money went to your business or to something tangentially related. The audit action is simple and reliable: expensive plus off-theme equals negate, whatever the conversion column says.

Apply a threshold to keep the read objective. If more than 20 to 25 percent of the queries in your 30-day pull are off-theme, the campaign’s match types and negative coverage need serious work, not a touch-up. Interpret the query strings with the knowledge that some are Google’s read of intent rather than the literal search, which changes how you negate them — the detail is in how AI-interpreted terms change negation. For deciding when a borderline query has earned a negative rather than a wait, the cost-versus-relevance rule beats a fixed click count, as laid out in how many clicks before you negate.

Step 4: Convert findings into patterns, not one-off negatives

On a low-visibility campaign — which, after step 1, you now know you have — one-off negatives lose to patterns. The wasteful queries are a long tail of near-unique strings, most of which you will never see individually, so a negative against a single query blocks one row while a hundred siblings keep spending. The audit move is to look across the off-theme queries you did surface for the recurring fragment they share — a product you do not sell, an intent word like “free” or “jobs,” a wrong geography — and negate the fragment. That single negative reaches the invisible siblings a per-query negative cannot.

This is where an n-gram analysis earns its place in the audit: it ranks the fragments by the cost sitting behind them, so you negate the word draining the most budget first instead of guessing. Group the findings into a small number of pattern-level negatives rather than a sprawling list of exact strings — the reasoning for grouping over enumerating is in pattern grouping versus single negatives. The output of this step is a handful of high-leverage negatives, each blocking an intent class across every future phrasing, seen or hidden.

Step 5: Fix visibility upstream where mining cannot reach

For any campaign that failed step 1 — visible spend under 30% — the real fix is structural, because you cannot negate your way out of blindness. The first lever is match type: broad-match spend producing one-off junk is exactly the spend disappearing into the anonymous bucket, so moving it toward phrase or exact concentrates budget on fewer, higher-volume queries that clear the reporting threshold and reappear where you can act on them. Tighter matching buys back the ability to see what you are paying for, which is the entire game on a blind campaign. The trade-offs are in the match-type decision for 2026.

The second lever is structure: segmenting a sprawling campaign so each ad group chases a tighter intent raises per-query volume and pulls more terms into view. For Performance Max, the equivalent is splitting by channel and applying brand exclusions to surface the Search-side queries, covered in the Performance Max search terms report. Each structural change is also a visibility change, so re-run step 1 afterwards to confirm the ratio actually moved — the audit is a loop, not a line, and the visibility ratio is both the diagnosis and the scoreboard.

Step 6: Retire the checks that no longer earn their time

Finish by deleting dead weight from your checklist, because time spent on obsolete checks is time not spent on the five steps above. Auditing exact-match keywords for airtight coverage is close to moot when AI Max expands past them regardless of the match type you set. Chasing a “complete” negative list term by term is theatre on a low-visibility campaign where the wasteful queries never appear as rows to negate. And signing off a campaign because its own conversion column looks healthy misleads, since that column borrows credit from the keywords AI Max displaced — the reconciliation problem detailed in AI Max attribution gaps.

Replace each retired check with its modern equivalent, and the audit stays honest as the platform keeps automating. Instead of verifying keyword settings, measure the visible share of spend. Instead of counting negatives, negate on cost, relevance, and patterns. Instead of trusting a campaign’s self-reported conversions, judge it on incremental account-level results. Run in that order — visibility, settings, relevance read, patterns, structural fix, then prune — and a modern search-hygiene audit produces a short list of specific actions rather than a false clean bill of health. The pillar routine in reading the Google Ads search terms report is the weekly habit this quarterly audit keeps honest.