Find negative keywords in your search terms report

Start with what the business actually sells

A search is irrelevant relative to an offer. Before opening a report, write down the products, services, delivery areas and customer types the campaign serves. Record exceptions too. A software business with a trial should treat free trial differently from free download, even though both contain free.

Distinguish business exclusions from campaign exclusions. A retailer can reject used products throughout its account while keeping running shoes in a shoe campaign and excluding them from a socks campaign. That distinction determines scope later. It also prevents a useful local exclusion from becoming an account wide mistake.

Keep a short protected vocabulary for product names, service names and valuable modifiers. Add converting search terms to the review context. A generic negative keywords list helps you recognize intent patterns, but every entry still needs a decision against the actual offer.

Review the search terms report in context

In Google Ads, open the relevant account or campaign and navigate to the Search terms report. The navigation can change, so use the interface search if needed. Choose a date range that contains enough recent activity and allows your usual conversion reporting delay to pass.

Include search term, campaign, ad group, matched keyword where available, impressions, clicks, cost, conversions and conversion value where relevant. Check which conversion actions feed the columns. A newsletter signup and a qualified sales opportunity should not carry the same weight in a decision about sales intent.

Keep campaign and ad group dimensions when exporting. The same query can be useful in one campaign and wrong in another. Combining those rows before checking context can disguise a profitable destination or make a routing problem look like a business wide exclusion.

The search terms report guide explains the difference between a query and the keyword that triggers it. Report visibility is also incomplete. Privacy thresholds and reporting rules mean that visible rows do not describe every paid search. Do not present visible flagged cost as a complete account waste total.

Sort expensive searches before judging marginal ones

Sort by cost, then inspect the highest spending searches individually. Read the full query and its destination. Look for requests the offer cannot satisfy, such as dental implants jobs in a patient acquisition campaign or payroll software course in a campaign selling payroll subscriptions.

Next, inspect converting queries separately. They reveal modifiers that generic lists often misclassify. Reviews can signal comparison before purchase. Cheap can indicate a budget conscious buyer. A conversion does not establish profitability, but it is a reason to investigate before proposing an exclusion.

Create a decision column with keep, exclude or investigate. Add a plain language reason, such as employment intent or unsupported product condition. A reason that merely says low performance is incomplete. State whether the issue is intent, economics, measurement or campaign placement.

Do not exclude relevant searches just because they have no conversions yet. Check tracking health, landing page fit, stock availability and conversion delay. If the query describes the right buyer, the next action may be a bid or page review rather than a negative keyword.

Group the reason rather than just the repeated word

Group unwanted queries by the request behind them. Employment, education, unsupported locations, used products and competitor support are useful starting categories. Split mixed groups before acting. Download can mean an unwanted document request or the correct installation step for paid software.

Use these queries to test decisions, not as a library to upload unchanged. Each classification depends on what the advertiser sells. Wrong location also needs care. Someone searching for a service in another area may be arranging it for a relative or researching a move.

For ambiguous terms, record the unresolved question and an owner. This keeps the approval queue useful. An account manager can ask whether refurbished stock is accepted without quietly turning every used product query into an account wide prohibition.

Find repeated cost patterns with n grams

An n gram is a consecutive sequence of words extracted from a query. Dental implants jobs contains the single words dental, implants and jobs, plus sequences such as implants jobs. Grouping repeated sequences exposes intent that is spread across many individually small rows.

Normalize capitalization and spacing for analysis, then sum clicks, cost, conversions and conversion value for each sequence within a relevant scope. Preserve the original query text. Do not merge plurals, spelling variants or reordered phrases into a proposed negative without checking native matching behavior.

For an implant service campaign, dental implants jobs, dental implants jobs remote and dental implants jobs salary share a sequence. Inspect all contributing rows before deciding that the sequence is consistently irrelevant. The commercial phrase dental implants also repeats, but repetition alone does not make it a candidate.

Build the aggregation from query rows, with a query contributing once to each distinct sequence it contains. If a phrase repeats within a query, avoid counting that row's cost twice for that phrase. Otherwise a text processing detail can make a weak pattern look expensive.

N gram totals overlap. A query can contribute to dental implants and implants jobs at the same time. Never add those group totals together as if they were separate spend. To estimate flagged historical cost, deduplicate the underlying query rows and keep their campaign context.

Compare short sequences with longer ones. The word manual may appear in searches for manual invoicing software and software instruction manuals. A longer sequence often expresses the unwanted intent more accurately. Phrase level evidence is particularly useful when a single word has several commercial meanings.

Set cost and conversion thresholds from account economics

Use thresholds to prioritize review, not to declare intent automatically. Define an acceptable acquisition cost from the business model and compare query cost with that benchmark. For ecommerce, consider contribution margin and conversion value rather than treating every recorded purchase as equally valuable.

A practical review rule flags a relevant query when its cost reaches the account's agreed review threshold without a mature conversion. You can set that first review threshold at the acceptable acquisition cost, then inspect intent and tracking before making a targeting decision. That flag means investigate. A clearly irrelevant request can justify exclusion earlier because the business cannot fulfill it, regardless of how little it has spent.

Separate the review window from the conversion window. Recent clicks can convert later, and imported offline outcomes can arrive after the initial lead. Recheck the period after those outcomes mature. Compare like with like instead of judging a recent query against older, fully attributed traffic.

Agree on these rules before reviewing the queue. Otherwise analysts can apply different standards to the same search. Record exceptions where seasonal demand, long sales cycles or unusual order values make the standard acquisition benchmark misleading.

Turn the finding into a precise exclusion

Choose exact when the complete query is unwanted but useful longer searches exist. Choose phrase when the same consecutive wording stays irrelevant with extra words before or after. In Google Ads, choose negative broad only when every component word signals the unwanted pattern in any order.

The negative keyword match types guide compares those rules. Google and Microsoft negatives do not expand to close variants, so separately review plurals and misspellings. Microsoft supports negative phrase and exact only. Amazon Sponsored Products uses negative phrase and exact with its own close variation behavior.

Apply the exclusion at the smallest scope that solves the problem. A category mismatch can belong in an ad group. A service mismatch can belong in a campaign. An account level exclusion requires an intent that is unwanted throughout the eligible account inventory, including future campaigns.

Check what the negative would block before publishing

Test each candidate against active keywords, advertised products and converting queries at the proposed destination. Check matching behavior, not just identical text. A negative phrase can block a longer converting query even when the proposed negative never appears as an active keyword by itself.

Also inspect existing exclusions and attached lists. A useful local decision can become redundant under a wider rule, while removing a local entry may leave the wider block in place. The campaign and account level guide shows how those layers interact.

Save the wording, match type, level, reason and approval with each change. Publish a coherent group that can be reviewed afterward. Check application status in the platform, then inspect subsequent reports. Historical spend remains historical spend, not proof of the amount a new exclusion saves.

Match review frequency to how quickly queries change

Review new or rapidly changing campaigns daily while their query patterns settle. Stable campaigns can use a regular weekly review with alerts for expensive new irrelevant searches between sessions. Use a longer lookback periodically to find recurring themes that daily checks fragment into small rows.

Repeat the conflict review after product launches, geographic expansion or changes to conversion goals. An exclusion that fits today's catalog can obstruct tomorrow's offer. Revisit old assumptions rather than treating a growing list as evidence that coverage always improves.