A search-query review should end with a decision attached to each useful query: keep it, change the message, route it to a better page, or exclude it. A spreadsheet of “irrelevant terms” is not enough; a broad negative can block profitable demand along with waste.
Read a query in its campaign context
For each term, inspect the actual search phrase, match type, campaign and ad group, landing page, spend, and the strongest downstream evidence available. Separate three questions: did the phrase match the intended task; did the ad set the right expectation; and did the landing page let that person complete the task? Do not judge intent from a single word such as “cheap” or “near me.”
| Query pattern | First check | Likely action |
|---|---|---|
| “free … template” on a paid-service campaign | Is a free resource part of the offer? | Route to it if yes; otherwise test a narrow exclusion |
| Service + city outside coverage | Is coverage genuinely unavailable? | Exclude that geography or phrase only after confirming service limits |
| Product query with high clicks and no orders | Does the page show stock, delivery, price, and payment clearly? | Fix the broken handoff before excluding the whole product theme |
Illustrative examples only; the action depends on account and order data.
Protect the learning
Review one change at a time and annotate the date. A query with no conversion may have low volume, a long consideration cycle, or a tracking gap. Check qualified leads, delivered orders, cancellations, or another real outcome when available; platform conversions are signals, not proof of business value. For COD, compare placed orders with delivered and retained value before changing bids.
Create a query decision log with the phrase, evidence window, decision, expected effect, owner, and revisit date. Use exact or phrase exclusions only when the unwanted intent is clear; prefer campaign-level control when the issue applies across ad groups. Recheck the next reporting period to catch overblocking.
Classify before adding negatives
Review each query with its match behavior, campaign goal, landing page, and downstream quality. Classify it as relevant and proven, relevant but unproven, ambiguous, or clearly excluded. Add a negative only when the exclusion is supported and its scope will not block a useful query elsewhere. For ambiguous terms, inspect actual search context or test a controlled route before excluding. Record the decision, match type, level, and reason so another operator can reverse it. This avoids turning a one-off irrelevant click into a broad account-level suppression.
Optional AI-assisted work brief — Search query disposition sheet
Classify each query against campaign goal, match behavior, destination, and downstream quality as proven relevant, unproven relevant, ambiguous, or excluded. For every negative keyword, state match type, level, evidence, and collateral risk. Escalate ambiguous terms rather than applying broad exclusions.
Worked scenario — One irrelevant query, narrow action
A query about free software appears in a paid service campaign. Check whether it repeats, which match behavior triggered it, and whether another campaign legitimately needs the term. If exclusion is justified, set the narrowest useful scope and document why. Do not block the root term across the account based on one search.
AI disclosure: AI tools assisted with drafting. Examples are illustrative and do not describe client account results.