Everything else in a Google Ads account is a plan; the search terms report is what actually happened. Keywords, match types, audiences, and bid strategies are instructions you gave the auction — the search terms report is the auction’s reply: the literal queries real people typed before your money was spent on their clicks. Which is why it is simultaneously the highest-yield audit surface in the platform and the most neglected one: accounts spending five figures a month routinely go quarters without anyone reading it, while broad match and Performance Max quietly stretch “emergency plumber dallas” into “how to become a plumber,” “plumber salary texas,” and “free plumbing course” — each click billed at commercial-intent prices.
The waste hides in plain sight because of how the report is usually read: sorted by spend, skimmed for the obviously absurd, a few negatives added, done. That method catches the comedy and misses the money. The expensive waste is structural — whole intent categories (informational, DIY, jobs, competitors’ customers looking for support), near-miss queries that look relevant but never convert, duplicate serving across campaigns bidding against each other, and the long tail of one-click terms that individually cost $4 and collectively cost thousands.
This guide is the audit method: how to pull the report so you actually see everything (including what Google now hides), the intent-bucket framework that finds waste categorically instead of anecdotally, the metrics that separate “no conversions yet” from “never will convert,” the negative keyword architecture that turns findings into a durable asset, the special handling PMax and broad match require, and the cadence that keeps the account clean in twenty minutes a week instead of a painful quarterly excavation.
The search terms report is what your money actually bought — audit it categorically, not anecdotally. Method: export a 90-day window (segment by campaign and match type), then bucket every meaningful term by intent: commercial-for-you, commercial-for-someone-else (jobs, DIY, competitors’ support, wrong geography), informational, and ambiguous. Waste concentrates in whole buckets, not individual weird queries. Judge terms on spend with zero conversions at statistically fair volume, cost per conversion vs your target, and click-through-but-bounce patterns — and check the long tail in aggregate: thousands of one-click terms are a budget line. Convert findings into shared negative lists by theme (jobs/careers, DIY/how-to, cheap/free, other-geo, competitor-support) applied account-wide, with exact/phrase precision to avoid blocking good traffic. Broad match and PMax need this discipline most and offer the least visibility — scheduled weekly triage (20 minutes) beats quarterly archaeology.
Pulling the Report So You Actually See It
Three mechanics before any analysis:
- Window and segmentation. Pull 90 days (long enough for the long tail to show its aggregate cost, short enough to reflect current matching behavior), and segment by campaign and by match type — the same term arriving via exact and via broad tells two different stories about your account structure.
- Know what’s hidden. Google omits low-volume terms from the report for privacy; a visible share of spend maps to “other search terms.” You cannot negative what you cannot see — but you can bound it: if the hidden share of spend is large and the visible terms are messy, the invisible ones are messier, and the structural fixes (match type discipline, tighter themes) matter more than term-by-term surgery.
- Include PMax’s view. Performance Max exposes search terms insights (and now supports campaign-level negatives) — less granular than Search reports, but auditable and increasingly actionable. Treat PMax’s search categories as a first-class part of the audit, not an afterthought; unreviewed PMax is where modern waste migrated when everyone learned to watch broad match.
The Intent-Bucket Framework
Reading term-by-term finds anecdotes; bucketing finds budget. Assign every term with meaningful spend to one of five buckets:
| Bucket | Examples (for a plumbing company) | Verdict |
|---|---|---|
| Commercial — for you | “emergency plumber near me,” “water heater replacement cost dallas” | Protect; consider promoting winners to exact-match keywords |
| Commercial — not for you | “plumber jobs dallas,” “plumbing supplies wholesale,” “[competitor] customer service” | Negative by theme — someone is buying, just not from businesses like yours |
| Informational / DIY | “how to fix leaking faucet,” “why is my water heater making noise” | Negative for search campaigns; content opportunity for SEO — route it there, don’t pay for it |
| Wrong scope | Out-of-area cities, services you don’t offer, residential terms in a commercial-only account | Negative + check location settings — geography leaks are usually a settings bug wearing a keyword costume |
| Ambiguous | “plumber,” “water heater dallas” | Judge on the numbers, not the vibes — the metrics section below |
The bucket exercise usually reveals that waste is not a hundred random weird queries — it is three to five categories the account never fenced off, each stoppable with one themed negative list. It also reveals the account’s gift to your SEO roadmap: the informational bucket is a pre-validated, spend-weighted content plan, which is exactly how paid data should feed the organic program (a core move in the channel sequencing playbook).
Filter the 90-day report to terms with exactly one click and no conversions, and sum their spend. In broad-match-heavy accounts this ‘orphan tail’ is routinely 10–20% of budget — invisible in any spend-sorted view because no single term looks worth acting on. You can’t negative a thousand one-off queries individually; you fix the tail structurally: tighter match strategy on the source keywords, better-themed ad groups, and pattern-based negatives (the recurring words inside the orphans — ‘free,’ ‘diy,’ ‘salary,’ other cities — are the levers). Track the orphan-tail percentage quarterly; it’s the single best health metric for matching discipline.
Judging Terms Fairly: The Numbers That Decide
Ambiguous terms — and impatient stakeholders — need decision rules that distinguish “hasn’t converted yet” from “never will”:
- Spend vs a statistically fair threshold. A term with 4 clicks and no conversions has told you nothing; a term that has spent 2–3× your target CPA with zero conversions has testified. Set the kill threshold as a multiple of target CPA, not a fixed click count — it self-adjusts across services with different economics.
- Conversion quality, not just presence. With CRM-fed conversion data, judge terms on qualified leads: some terms convert into form fills that sales never closes — the “cheap” and “free estimate” families are notorious. A term producing conversions that never become customers is waste with better camouflage.
- Engagement forensics for the borderline. High CTR + instant bounce on a relevant-looking term usually means intent mismatch (they wanted the DIY answer, your page sells the service). Session recordings on the landing page settle arguments the report can’t.
- Duplicate-serving check. The same term triggering multiple campaigns or ad groups means you’re bidding against yourself and splitting learning data — fix with structure and cross-campaign negatives, not by pausing the term.
From Findings to Asset: Negative Keyword Architecture
Ad-hoc negatives added where you happened to be standing decay into an unmaintainable sprawl. Build the findings into architecture instead:
- Shared, themed lists at the account level: Jobs & Careers; DIY & How-To; Cheap/Free/Discount-intent; Out-of-Area Geo; Competitor-Support; Irrelevant Verticals. Applied to all relevant campaigns, maintained in one place, portable to every new campaign on day one — the full construction method is in our negative keywords guide.
- Match types on negatives matter in reverse. Broad negatives block aggressively (and remember: negative keywords don’t expand to close variants the way positives do — cover plurals and key variants explicitly); phrase and exact negatives are the scalpel for surgical blocks that mustn’t catch good traffic. Default to phrase for themes, exact for specific toxic queries.
- Campaign-level negatives for traffic shaping: steering brand vs non-brand, or fencing service lines from each other, is a routing job for campaign-level negatives — distinct from the account-level waste-blocking lists.
- Protect the winners while you prune. Every audit should also promote: converting search terms that aren’t yet keywords become exact-match additions in the right ad group — the report is a two-way street, and the promotion side compounds just like the pruning side.
Audits create momentum, and momentum creates casualties: a broad negative ‘repair’ added to block DIY queries that also blocks ‘water heater repair near me,’ a ‘cost’ negative that kills your highest-intent pricing researchers, a geo negative for a city you actually serve half of. Symptoms appear as gradual impression and volume decline over the following weeks — rarely traced back to the audit that caused them. Discipline: preview the traffic each negative would have blocked (search the term list before committing), prefer phrase/exact over broad for anything sharing words with commercial queries, document every list change with a date and reason, and put a two-week volume check on the calendar after any large negative deployment. Pruning should cut waste’s share of spend — if total conversion volume dropped with it, you cut flesh.
The Cadence: Twenty Minutes a Week Beats Quarterly Archaeology
The audit above is the deep clean; the maintenance loop keeps it clean: weekly (15–20 min) — new terms since last review, sorted by spend, triaged into buckets; obvious negatives applied to the themed lists; converting terms flagged for promotion. Monthly — orphan-tail percentage, duplicate-serving check, PMax search categories review. Quarterly — full re-bucket of the 90-day window, negative-list hygiene (remove obsolete blocks, resolve conflicts), and the promotion pass. On broad-match and PMax-heavy accounts, weekly is not optional — those matching systems explore continuously, and exploration unreviewed is billing unreviewed. The whole loop is one recurring calendar block, and it is the highest-ROI recurring block in PPC management — which is why it’s the first thing checked in any serious account audit: the date of the last search-terms review predicts account health better than any single setting.
5 Common Search Terms Audit Mistakes
- Skimming for absurdities instead of bucketing by intent. The comedy queries are pennies; the intent categories are the budget.
- Killing terms on tiny samples. Four clicks and no conversion is noise; the fair threshold is a multiple of target CPA.
- Ignoring the hidden and PMax layers. The visible report is a sample; matching discipline and PMax insights govern the rest.
- Ad-hoc negatives instead of themed shared lists. Findings that don’t become architecture get re-discovered — and re-paid for — next quarter.
- Auditing only to prune. The report’s converting terms are your cheapest keyword research and your SEO content plan; harvest both directions.
Frequently Asked Questions
How often should I review the search terms report?
Weekly for triage, quarterly for the deep audit — with the weekly block being the one that actually protects the budget. Fifteen to twenty minutes reviewing new terms since the last pass (sorted by spend, bucketed by intent, negatives applied to the shared lists) prevents the waste categories from ever accumulating into a five-figure discovery. Broad-match-heavy and PMax-heavy accounts should treat weekly as mandatory: those systems explore new queries continuously by design, and each week unreviewed is a week of exploration billed at your CPCs. The quarterly deep pass does what the weekly can’t: aggregate views (orphan-tail percentage, duplicate serving, bucket-level spend shares), negative-list hygiene, and the promotion of proven terms to exact keywords. Accounts under ~$2k/month can stretch weekly to biweekly; accounts above $10k/month should consider the weekly block non-negotiable and named in someone’s job description.
A term looks relevant but has spent $400 with no conversions. Kill it or keep it?
Run it through three checks before the verdict. Statistical fairness: compare its spend to your target CPA — at a $150 target, $400 with zero conversions is 2.7× the threshold and legitimately suspect; at a $500 target it hasn’t finished testifying yet. Landing-page fit: high CTR with fast bounces on a plausible term usually means the page answers a different intent than the query — check whether a better-matched page (or a session-recording review) rescues it before the negative does. Conversion definition: if your tracked conversion is shallow (form fills) and the term’s traffic converts downstream via calls you’re not counting, the term is innocent and the measurement is guilty — the cross-check against CRM data settles it. If it survives none of these: negative it at phrase or exact scope (not by pausing a keyword that also matches good queries), note the date and reason, and let the budget flow to terms that converted. ‘Looks relevant’ is a hypothesis; 2–3× target CPA with zero results is a verdict.
Google hides a lot of my search terms now. How do I audit what I can't see?
Accept that term-level surgery only reaches the visible layer, and manage the hidden layer structurally. First, measure it: the gap between total spend and the spend attributed to visible terms tells you how big the blind spot is — if it’s large, that finding itself matters more than any individual negative. Second, infer from the visible sample: the hidden terms come from the same matching behavior, so a messy visible report implies a messier hidden one, and the fixes are upstream — tighter match-type strategy, more precisely themed ad groups, smarter bidding fed with quality conversion data (which makes Google’s own matching more selective), and robust themed negative lists that block whole patterns rather than individual strings. Third, use every visibility surface you do have: PMax search-terms insights, insights-page search trends, and analytics landing-page query data all triangulate the blind spot. The philosophical shift: in the hidden-terms era, negatives and structure are how you steer matching; the report is how you verify the steering worked.
Should I add every irrelevant search term I find as a negative keyword?
No — negative the patterns, not the specimens. A thousand individual bad queries usually reduce to a handful of themes (jobs, DIY, free/cheap, other cities, competitor support), and one phrase-match theme negative in a shared list blocks the entire family including the variants you haven’t seen yet — while a thousand exact negatives block precisely those thousand strings and nothing else, forever, as an unmaintainable pile. Reserve exact negatives for surgical cases: specific toxic queries that share words with good traffic and can’t be pattern-blocked safely. Also apply a materiality filter: a term with one click and $3 spend that will likely never recur isn’t worth a list entry — its category might be, which is the difference between auditing anecdotes and auditing intent. And re-check the list quarterly: negatives added for a service you now offer, or a city you’ve expanded into, are yesterday’s protection quietly blocking today’s revenue.
Do negative keywords work the same way in Performance Max?
The gap has narrowed but not closed. PMax now supports campaign-level negative keywords (a relatively recent and genuinely useful addition) alongside the longer-standing tools: account-level negatives apply to it, and brand exclusions control branded serving. What remains different: visibility is coarser — search-terms insights show categories and top terms rather than the full ledger a Search campaign provides — and PMax spends across many surfaces (Display, YouTube, Discover) that negatives don’t govern at all, so keyword hygiene addresses only its search-adjacent slice. Practical protocol: apply your themed shared lists at the account level so PMax inherits the waste-blocking, add campaign-level negatives for PMax-specific findings from its insights, review those insights on the same weekly cadence as Search terms, and use the other PMax controls (audience signals, asset-group theming, URL exclusions, and final-URL expansion settings) as the primary steering. For the fuller toolkit on constraining the campaign type, see our guide on reclaiming control over Performance Max.
When did anyone last read what your budget actually bought?
Our audit reconciles ninety days of search terms against your spend: the intent buckets, the orphan tail, the duplicate serving, and the themed negative lists that stop the waste durably — plus the converting terms your keyword list is missing.
Get a Search Terms Audit Explore Ads Management