Negative keywords do not behave like the positive keywords you are used to, and the single difference that costs the most money is this: negatives do not match close variants. Add shoe as a negative and your ad can still show for shoes, shoe’s, or the misspelling shooes, because Google deliberately does not expand negatives to plurals, misspellings or synonyms the way it expands positive keywords. Google’s own guidance is explicit that you need to add synonyms and singular or plural versions if you want to exclude them. The block you thought you applied only ever covered one exact spelling.
That asymmetry is the source of a whole category of leaked spend. Practitioners assume a negative works like a positive — catch the root, catch the variants — and it does not. This post covers how each of the three negative match types actually behaves, why the plurals-and-variants gap exists, how to set the match type deliberately, and how to avoid the opposite failure of over-blocking. It pairs with the search terms report walkthrough, which is where you find the variants you then have to add by hand.
Why negatives ignore close variants on purpose
Positive keywords match close variants so a single keyword can reach the plurals, misspellings and reorderings of what you bid on without you listing every form. Negative keywords do the opposite: they match only the exact words you specify, precisely so that one negative cannot silently block a wide swathe of traffic you meant to keep. If negatives expanded to variants, adding free as a negative might also catch freelance or freedom, and you would lose converting queries with no obvious cause. The narrow behaviour is a safety feature, not a bug.
The cost of that safety is manual work. Because the system will not infer that shoe and shoes are the same concept to you, you have to enumerate the forms that matter: the singular and the plural, the common misspellings your report shows real spend on, the abbreviations, and the genuine synonyms. You do not need every theoretical variant — you need the ones actually appearing in your search terms report with cost against them. That is why the report is the input to good negative lists: it tells you which variants are real in your account rather than which ones are possible in principle.
Negative broad match: all words, any order
A negative broad-match keyword blocks a search only when that search contains every word in your negative term, in any order, with any additional words allowed. The negative broad term running shoes blocks shoes for running and cheap running shoes online because both contain running and shoes, but it does not block a search for running alone, because not all the words are present. The critical rule is that all words must appear; missing any one means the negative does not fire.
This makes negative broad the most aggressive type and the easiest to misuse. A short two-word broad negative can catch long, high-intent queries you never intended to exclude, because the only requirement is that both words appear somewhere in the search. Reserve negative broad for sets of words that are genuinely irrelevant in any arrangement — a clear off-topic pairing that never converts — and be cautious about single common words, which can block far more than the query you were looking at when you added them.
Negative phrase and exact: order and isolation
Negative phrase match blocks searches that contain your full term in the order you wrote it, with extra words allowed before or after. The negative phrase "free shoes" blocks free shoes near me and where to get free shoes, but not shoes for free, because the word order does not match. Phrase is usually the safest default for multi-word waste: it excludes the junk phrase wherever it appears in sequence without over-reaching into unrelated queries that happen to share one word, which is exactly the over-blocking risk that negative broad carries.
Negative exact match blocks only searches for your exact term with no extra words at all. The negative exact [jobs] blocks the standalone query jobs while still letting your ad serve on jobs software or project management jobs board. Use exact when a word is wasteful only in isolation but valuable inside longer, qualified queries — a common pattern for words like jobs, free, login or a bare brand term. Remember that even negative exact ignores plurals and variants, so [job] and [jobs] are two separate exclusions you may both need.
Set the match type with punctuation, deliberately
You control a negative keyword’s match type the same way you control a positive one: punctuation. A bare term such as tennis shoes is negative broad, quotes such as "tennis shoes" make it phrase, and square brackets such as [tennis shoes] make it exact. The trap is bulk-pasting. When you paste a plain column of words with no punctuation — the way most people build a first negative list — every one goes in as negative broad, the most aggressive type, and a list meant to trim the edges can start blocking converting queries instead.
So treat list-building as a match-type decision per line, not a copy-paste. For a scraped list of obvious junk words, broad may be fine. For anything containing a word that also appears in real converting queries, reach for phrase or exact so the negative excludes the wasteful pattern without collateral damage. Applied at the campaign or account level, these lists are the main containment tool once broad match and AI Max expansion widen the matching surface — which makes getting the match type right on each line more consequential, not less.
The opposite failure: over-blocking your own traffic
Getting match types wrong does not only leak spend by under-blocking — it can also starve the account by over-blocking, and this failure is harder to spot. Negatives leave no row in the search terms report: a query your negative excluded simply never appears, so the traffic you lost is invisible unless you go looking for it. A pile of loosely-scoped negative broad terms can quietly suppress conversion volume, and the symptom — impressions and clicks drifting down for no obvious reason — rarely points back at the negative list on its own.
Guard against it with three habits. Keep negatives as tight as the job requires, preferring phrase or exact for any multi-word term rather than defaulting everything to broad. Maintain a written log of every negative — the term, its match type, the level it lives at, the date, and why — so that when volume dips you can reconstruct what each term is blocking instead of guessing. And review the list periodically against current queries, because a negative that made sense against last quarter’s waste can block this quarter’s opportunity. Precise negatives and a clean log are what let you block waste aggressively without quietly cutting the traffic that pays. For the wider audit, see the wasted spend the top-cost view hides.