The Best Match Type for Negative Keywords (And When to Break the Rule)

Default to phrase match negatives for safety, use exact for surgical exclusions, and reserve broad for whole themes you never want to touch.

August 24, 2026 · QueryCut

For most negative keywords, phrase match is the best default: it blocks any search containing your term in order without accidentally suppressing traffic you want. Use negative exact for a single term you must exclude but whose words appear in good queries, and use negative broad only for whole themes (like a competitor's brand or an unrelated industry) where every word appearing anywhere means the query is junk. The reason is that negative match types behave differently from positive ones, and getting the difference wrong either lets waste through or silently kills good traffic.

How negative match types actually work (and where they differ from positive)

Negative keywords use the same three match type labels as positive keywords, but the matching logic is stricter and, critically, they do not use synonyms, close variants, or intent expansion. A positive broad match keyword for "running shoes" can trigger on "jogging sneakers." A negative broad match keyword for "running shoes" will not. Negatives only block on the literal words you type. That single fact explains most of the mistakes people make.

Negative phrase match blocks any search that contains your exact words in the exact order, with anything before or after. Negative "free trial" blocks "software free trial," "best free trial offer," and "free trial" itself, but not "trial free" (wrong order) or "free demo" (different word). Negative exact match blocks only searches that are the term and nothing else: negative exact "free" blocks the single-word query "free" and nothing containing free plus another word. Negative broad match blocks any search that contains all your words in any order, but only those literal words: negative broad "free trial" blocks "trial for free" and "free 30 day trial" but not "free."

The other place negatives diverge from positives: negative match types ignore close variants entirely. Positive keywords match plurals, misspellings, and stem variations. Negatives do not. Negative exact "job" will not block "jobs." This is the single most common reason a negative "does not work": the searcher typed a plural or a typo and your literal negative never fired.

The decision rule: which type for which situation

Start every negative as phrase match unless you have a specific reason to change it. Phrase match is the balance point: broad enough to catch the variations that matter (word order aside), narrow enough that you will not nuke queries that merely share a word. If you are cleaning a search terms report and you see "cheap plumbing course" and you sell plumbing services not courses, add phrase "course" and you block every course query in one line.

Drop to exact match when the same word lives in both good and bad queries and you can only exclude one specific phrasing. If "free" appears in your winners ("free shipping software") but the bare query "free" wastes money, negative phrase "free" would kill your winners too. Negative exact "free" blocks only the standalone junk query and leaves the rest alone. Exact is your scalpel.

Go broad only for whole themes where any co-occurrence of the words signals a bad match. Competitor brand names, unrelated product categories, and job seekers are the classic cases. If you never want traffic mentioning a competitor, negative broad on the two-word brand catches every rearrangement and modifier. The tradeoff: broad negatives are the easiest to over-block, so keep the word count small (one or two words) and only use words that never appear in a query you would want.

• Default: phrase match. Blocks the term in order, anywhere in the query.
• Scalpel: exact match. Blocks only the standalone query, protects winners that share a word.
• Theme kill: broad match. Blocks all your words in any order, reserved for competitor names and off-topic categories.
• Remember: negatives never match plurals, typos, or synonyms. Add each variant yourself.

The failure modes that cost you traffic

The expensive mistake is a broad or phrase negative that is one word too aggressive. Say you sell enterprise CRM software and add phrase negative "small" to filter out "small business crm." You have now blocked "small team crm," "smallest crm setup," and any legitimate query where "small" appears next to your product for a buyer you actually want. The word looked like a qualifier for a bad segment, but it also appears in good ones. Before adding any phrase or broad negative, mentally run three good queries through it and confirm none get blocked.

The opposite failure is the negative that does nothing because you chose exact when you meant phrase. Adding negative exact "free download" and expecting it to stop "free download of accounting software" is a common miss. Exact only blocks the query that is precisely "free download" with no other words. If waste keeps appearing after you add a negative, the first thing to check is whether the offending search term has extra words that your exact negative cannot reach.

The silent failure is the plural or misspelling gap. You add negative "internship" and job-hunt traffic keeps costing you money because everyone searches "internships." Google will not extend your negative to the plural the way it would for a positive keyword. You have to enumerate: internship, internships, intern, interns. This is tedious, which is exactly why theme-level negative lists exist and why account managers with large search terms reports lean on tooling to generate the variants rather than typing each one.

Length limits, lists, and the practical mechanics

A single negative keyword can be up to 10 words and 80 characters. That is a hard cap, so a long phrase negative like an entire competitor slogan may need trimming. In practice you almost never want a 10-word negative anyway, because the more words you require, the more specific the query has to be to trigger it, and long-tail junk rarely matches word for word. Short negatives catch more.

Negative keyword lists (Tools > Shared library > Negative keyword lists) let you build one list and apply it across many campaigns. The match type is set per keyword inside the list, so a single list can mix phrase-match theme blocks with a few exact-match scalpels. This is the right home for evergreen junk: "free," "jobs," "salary," "reddit," "login," competitor names. Campaign-level and ad-group-level negatives are for exclusions specific to that scope.

Watch the precedence rule that trips people up: an ad-group negative can conflict with a campaign positive keyword, and negatives always win. If you add negative phrase "shoes" at the ad group level and also bid on "running shoes" there, the negative blocks your own keyword. When traffic to a keyword drops to zero unexpectedly, check for a negative you added somewhere up the hierarchy that swallows it.

A workflow you can run this afternoon

Open Campaigns > Insights and reports > Search terms and sort by cost descending. Scan the top spenders for queries that will never convert for you. For each, decide the smallest unit that identifies the junk: is it a single throwaway word ("free," "jobs"), a two-word theme, or a specific phrase? That decision picks your match type. Single junk words that also appear in good queries go exact. Junk words that only ever appear in bad queries go phrase or broad.

Then handle variants. For any word you block, add its plural and its obvious misspellings as separate negatives, because the system will not do it for you. Group the evergreen ones into a shared negative keyword list so you never have to rebuild them per campaign. If your search terms report is large enough that manually finding variants and picking match types is a slog, that is the point where a generator that scans the report and proposes negatives with match types (QueryCut does this) saves the tedious part. The judgment about which to apply is still yours.

After you apply negatives, do not assume they worked. Come back to the search terms report in a week and confirm the blocked queries stopped appearing. If they did not, the usual culprit is match type (exact when you needed phrase) or a variant you missed. This verification loop is the difference between a negative list that quietly rots and one that keeps the account clean.

Questions people also ask