Negative keywords

Updated November 12, 2026

What negative keywords are

A negative keyword is an exclusion: a word or phrase that blocks your ads from matching searches containing it. If you sell paid software and add 'free' as a negative, searches for 'free crm software' stop triggering your ads and stop spending your money.

Negatives exist because positive keywords no longer mean what they say. Broad match, close variants and Performance Max all let Google match your ads to queries you never chose, which is often useful and reliably messy. The negative list is the fence you build around that expansion: the platform explores, the fence keeps the exploration on your property.

Negatives are also the one Google Ads lever that is pure savings when used correctly: blocking a query that spends and never converts costs nothing and reclaims the spend. The craft is in 'used correctly', which is what the rest of this page covers.

How negative match types work

Negatives have match types, and they behave differently from positive match types in one critical way: negatives do not expand. A negative does not block synonyms, misspellings or plural variants on its own; it blocks exactly what its match type says.

The non-expansion rule is why negative lists grow: you often need 'free', 'freebie' and 'freeware' separately. It is also the safety property, since a negative never blocks more than its letters say, so a precisely written negative cannot silently over-block the way people fear.

Where negatives live: levels and shared lists

Placement level decides the blast radius, and the right level is the narrowest one that covers the problem:

Ad group level blocks the query for that ad group only. Use it for traffic shaping, like keeping 'running shoes' searches out of the trail-shoes ad group so the better ad serves. Campaign level blocks across the campaign; use it for terms wrong for the whole campaign. Shared negative lists are account-level lists applied to many campaigns at once: the right home for universal junk, job seekers, freebie hunters, DIY researchers, and competitor terms you have decided not to pay for. Build the shared lists once and every new campaign inherits the fence on day one.

Performance Max is the special case: it has no keywords, and exclusion options have historically been narrower, with account-level negatives and brand exclusions as the practical levers. Keeping PMax off your own brand name is the single most common PMax exclusion, because PMax will otherwise book brand demand it did not create; see account structure for the cannibalization picture.

How the AdsGuru routine builds proposals

Weekly, AdsGuru reads the search terms report, including PMax search themes where available, and looks for queries that spent without converting, plus queries whose intent is visibly wrong regardless of spend, like job-seeking phrases on a product campaign.

Matching queries are clustered by intent rather than listed raw: one proposal might be 'jobs and careers family: 14 queries, $63 spend, 0 conversions', with every underlying query and its own stats attached. Each cluster comes with a proposed negative at a proposed level: ad group for precision cases, campaign or shared list for universal junk, and a proposed match type, phrase negatives for phrases, exact negatives where a single ambiguous word needs surgical blocking.

Two hard safety rules. A query with conversions in the window is never proposed as a negative, whatever its cost looks like. And proposals prefer precision over reach: a phrase negative on the actual junk phrase, rather than a broad single-word negative that might fence off good traffic. You approve, edit or reject each cluster; approved negatives apply immediately and are logged. On autopilot, the same rules hold inside the same caps.

Add negatives manually: the weekly sweep

The manual version of the routine, for doing it yourself or auditing what AdsGuru proposes:

1. In Google Ads, open Campaigns, then Insights and reports, then Search terms. Set the date range to the last 7 days.
2. Sort by cost, descending. Read every query above a spend floor you care about (for a small account, everything above a few dollars).
3. Mark queries with meaningful spend and zero conversions, plus anything with visibly wrong intent: free, jobs, DIY, courses, other brands you do not want.
4. For each, decide the level: this ad group only, this campaign, or the account-wide shared list.
5. Choose the match type: phrase negative for junk phrases, exact negative for single ambiguous words you cannot broadly ban.
6. Add them (select the term, Add as negative keyword, pick the level), then recheck in two weeks for new variants; junk mutates.

Done properly this takes a focused half hour weekly on a mid-size account, which is exactly why it gets skipped, and why it automates well: the judgment per query is real but repetitive. The automation answer page covers the tooling angle.

Negative keyword mistakes

• Negating a converter. The costliest mistake: a query that converts occasionally gets blocked for having a bad week. Always check the longer window before negating, and never negate on cost alone. This is the mistake AdsGuru hard-blocks.
• Broad single-word negatives. A negative broad 'cheap' blocks 'cheap' in every query, including 'best cheap crm for startups', which might be your best buyer. Prefer phrase negatives on real junk phrases.
• Negative-positive conflicts. A negative at the campaign level can block your own keyword in an ad group below it; Google flags conflicts, but only if someone reads the flags. The nightly audit checks for these.
• One giant cleanup, then nothing. Query mix shifts with seasons, products and Google's matching behavior; negatives are a weekly diet, never a one-time purge.
• Forgetting new campaigns. A new campaign without the shared lists attached relearns every lesson at full price. Shared lists exist so this cannot happen; attach them by default.
• Over-fencing early. On a new campaign, aggressive negatives before data arrives can block discovery of queries you did not know were good. Fence from evidence, lightly at first.

Each of these costs real money in a different direction, which is why the routine's safety rules exist: never negate converters, never fence wider than the evidence, show the blast radius before approval.

Do negatives apply to Meta ads?

No; negative keywords are a search concept, and Meta ads do not target searches. The nearest Meta equivalents are exclusions of a different kind: excluded audiences (blocking existing customers from prospecting campaigns), excluded placements, and brand safety controls. AdsGuru handles those through its Meta routines rather than through this one.

The spirit transfers, though: in both ecosystems the platform spends by default into everything it can match, and the account owner's job is maintaining the fences. On Google the fence is the negative list; on Meta it is exclusion audiences and the creative and budget discipline that keeps spend pointed at the right people.

Common questions

Related