Negative Keyword Match Types in Google Ads: Which to Use and Why
How broad, phrase, and exact negatives actually block traffic, and the decision rule for picking one without accidentally muting good queries.
The short version
- Negatives never expand: no synonyms, no plurals, no misspellings, so add every variant explicitly.
- Phrase is the safe default; exact carves out one query next to a converter; broad requires all words present and is narrower than people expect.
- A one-word broad or phrase negative can silently block good traffic; check the search terms report for that word first.
- Cutting heuristic (mine, not Google's): 3x target CPA spent with zero conversions and off-intent, and wait for ~100 clicks before trusting a zero.
- Use shared exclusion lists for universal junk and watch for negatives starving the wrong campaign.
Google Ads negative keywords come in three match types (broad, phrase, and exact), and they behave differently from positive keywords in one crucial way: negatives never expand. A negative broad match blocks only queries containing all your words in any order, a negative phrase blocks queries containing your exact word sequence, and a negative exact blocks only the query that matches word for word. For most accounts, negative phrase match is the safest default, with negative exact reserved for surgically cutting a single converting-adjacent term.
How the three negative match types actually block traffic
The single most important fact: negatives ignore synonyms, close variants, plurals, and misspellings. A positive broad match keyword for "running shoes" can trigger on "jogging sneakers." A negative broad match for "running shoes" will not block "jogging sneakers," because negatives match only the literal words you type. This asymmetry is deliberate. Google wants to make it easy to show ads and hard to accidentally suppress them, so the matching layer treats negatives as a strict, literal filter.
Negative broad match (no punctuation, e.g. free trial) blocks a search only when the query contains every word in your negative, in any order, with any words in between. So the negative free trial blocks "trial free download" and "start a free 30 day trial" but does not block a search for just "trial" or just "free." All words must be present.
Negative phrase match (in quotes, e.g. "free trial") blocks a search only when the query contains your exact word sequence, contiguously. It blocks "want a free trial now" but not "free 30 day trial," because the words free and trial are no longer adjacent and in order.
Negative exact match (in brackets, e.g. [free trial]) blocks a search only when the query is exactly those words in exactly that order with nothing added. It blocks "free trial" and nothing else. Not "free trial software," not "trial free."
The best match type for negatives (and when the default is wrong)
For blocking a theme or an unwanted concept, negative phrase match is the right default. It is broad enough to catch the natural variations people type around your bad term but narrow enough that it will not swallow queries where those words happen to co-occur for a good reason. If you want to stop showing for job-seeker traffic, the phrase negatives "jobs" and "careers" and "salary" will cover the realistic phrasings without you enumerating every one.
Use negative exact when you are cutting one specific search term that sits next to a converter. Classic example: you sell software and you convert on "crm software" but never on "crm software free." A negative phrase "crm software" would block your winner too, because your winner contains that sequence. You need [crm software free] as a negative exact so you carve out only the freeloader query and keep the buyer.
Use negative broad sparingly and mostly for brand-safety or category exclusions where any co-occurrence is disqualifying. If you never want to serve on anything containing both "used" and your product, negative broad used gets it done regardless of word order. The trap: because broad requires all words present, a two-word or three-word negative broad can be surprisingly narrow. People assume broad is the widest net; for negatives it can be the narrowest, because more required words means fewer matching queries.
The traps that silently kill good traffic
Trap one: a broad negative single word is a sledgehammer. Adding the single negative "free" as broad or phrase blocks every query containing free, including "tax-free retirement software" or "free shipping calculator" if those are relevant to you. Before you add a one-word negative, run it through Campaigns > Insights and reports > Search terms and search that word to see what converting queries contain it.
Trap two: negatives do not handle close variants, so plurals and misspellings leak through. If you block "invoice" as phrase, the query "invoices" still serves. You must add both singular and plural, and the common misspellings, as separate negatives. This is the opposite of positive keywords, where close-variant matching handles those for you.
Trap three: order matters for phrase and exact but not for broad. "trial free" is not blocked by phrase "free trial." If word order varies in real queries, either add both orderings as phrase negatives or use a broad negative, which ignores order.
Trap four: the 10-word limit. A single negative keyword can contain at most 10 words. Long junk queries have to be blocked by a shorter phrase inside them, which means you accept some collateral. Pick the shortest distinctive phrase that appears in the junk but not in your good terms.
A decision rule for pulling the trigger
Here is the workflow to run today. Open Campaigns > Insights and reports > Search terms, set your date range to the last 30 to 90 days, and sort by cost descending. Add columns for conversions and cost per conversion. Now you are looking for spend with nothing to show for it.
My cutting heuristic (this is a judgment call, not a Google rule): if a search term has spent roughly 3x your target cost per acquisition with zero conversions, and it is genuinely off-intent, add it as a negative. The 3x is mine, chosen because at 1x you are cutting terms that were one conversion away from being fine, and past 3x you are just burning money waiting. Adjust it up if your conversions are rare and lumpy, down if you have high volume and fast feedback.
Before trusting any zero-conversion verdict, wait for roughly 100 clicks on that term. Below that, a single missed conversion swings the read, and the cost of waiting a few more days is usually smaller than the cost of cutting a term that was about to convert. The 100 is a practitioner threshold, not a statistical guarantee; treat it as the point where you stop being fooled by small samples.
For the match type of the negative itself: block the theme with phrase, block the exact freeloader query with exact, and reserve broad for all-words-must-be-present category exclusions. When in doubt, phrase is the safe middle.
Building and maintaining the list without drowning
Group negatives into shared negative keyword lists under Tools > Shared library > Exclusion lists, then apply the same list across campaigns. A universal junk list (jobs, careers, free, wikipedia, youtube, reddit if those are off-intent for you) applied account-wide saves you from re-adding the same terms per campaign. Keep campaign-specific negatives at the campaign level so you do not cross-block between product lines.
Watch for the cross-campaign conflict: a negative applied at the account or shared-list level can block a term you want a different campaign to serve. If a campaign suddenly loses impressions, check Campaigns > Insights and reports and look at your negative lists before blaming bids or budget. Google will not warn you that a negative is starving a campaign.
Free negative keyword tools and generators can seed a starter list of obvious junk (competitor names, off-topic modifiers, common freebie-seeker words), which is useful before you have your own search-term history. QueryCut reads your actual search terms report and proposes negatives with the match type already chosen, which removes the manual pass, but the mechanism above is what makes any list safe: literal matching, no close variants, order matters for phrase and exact, all-words for broad. Understand that and you can build the list by hand in an afternoon.
Questions people also ask
What is the best match type for negative keywords?
Negative phrase match is the best default because it catches natural variations of a bad term without over-blocking. Use negative exact when you need to cut one specific query that sits next to a converting term, and use negative broad only for category exclusions where all your words must be present regardless of order.
Do negative keywords block plurals and misspellings?
No. Unlike positive keywords, negatives do not use close-variant matching, so "invoice" as a negative will not block "invoices" or "invocie." You have to add the singular, plural, and common misspellings as separate negative keywords.
Why does my negative broad keyword not block a search that clearly contains one of the words?
Negative broad match requires every word in your negative to be present in the query, in any order. If your negative is two words and the search contains only one of them, it will not be blocked. More required words makes a broad negative narrower, not wider.
How much should a search term spend before I add it as a negative?
As a practitioner rule of thumb, wait until a term has spent roughly 3x your target cost per acquisition with zero conversions and is genuinely off-intent, and until it has accumulated around 100 clicks so a single missed conversion does not fool you. These are judgment thresholds, not Google rules; loosen them for low-volume accounts and tighten them for high-volume ones.
Can a negative keyword block a keyword I want to run?
Yes. A negative applied at the account level or in a shared exclusion list can block terms in a campaign where you actually want to serve. If a campaign loses impressions unexpectedly, review your negative lists in Campaigns > Insights and reports before adjusting bids or budgets.