What is Negative keyword?

Search advertising has a lot of settings that influence and one that refuses. A bid strategy can decline an auction it does not like; a match type can make a search less likely; an ad group's context can shape what the platform considers related. Only a negative keyword says no.

That distinction has become more important as matching has loosened. When exact match meant exact words, the keyword list was itself a form of refusal — anything not on it did not serve. Now that all three types match on interpreted meaning, the positive keywords describe what an advertiser wants rather than bounding what they get, and the bounding is done by negatives.

An account that has loosened its match types without building the negative discipline has removed its brakes and kept its accelerator.

The practical rule is narrow and it is the one to build the habit on: a negative keyword blocks what it names, under the match type it was given. Anything it does not name keeps serving.

That is why the list has to be built from the search-term report rather than from imagination. The report shows what people actually typed, which includes the phrasings, spellings and constructions nobody sitting in an account would have thought to write down. A list assembled by guessing covers the obvious cases and misses the ones costing money.

How the platforms treat variations of a negative has been revised more than once and differs between them, so the durable practice is not to rely on any particular expansion behaviour: watch the report, and add what is still getting through.

Match types apply to negatives as well, and they work as the names suggest. Negative broad blocks searches containing all the words in any order. Negative phrase blocks the words in sequence. Negative exact blocks only that search. Broad negatives are powerful and are the ones most likely to block something wanted, which is an argument for preferring phrase or exact where a term is ambiguous.

Most accounts need the same few groups, and building them once is worth more than months of incremental additions.

Free, cheap, DIY. People looking for a no-cost version of a paid thing, in every phrasing they use.

Jobs and careers. Employment searches around a category engage and never buy.

Research and education. Definitions, how-to, tutorials, meanings, examples — valuable traffic for content and expensive traffic for a purchase campaign.

Competitors, unless bidding on them is a deliberate strategy with somebody's name on it.

Adjacent products the business does not sell. The most account-specific category and usually the largest source of waste, because the platform's semantic matching finds neighbours the advertiser thinks of as distinct.

Wrong segment. Wholesale when the business is retail, commercial when it is residential, secondhand when it is new.

Places and contexts you cannot serve, where the targeting does not already handle it.

The level should match the decision.

Account-level shared lists for terms no campaign should ever pay for. This is where the universal categories belong, and a shared list applied everywhere is easier to maintain than the same terms copied into a dozen campaigns.

Campaign level for terms that are wrong here and right elsewhere — a term that belongs to a different product line, or a brand term that should only serve in the brand campaign.

Ad group level for routing. Where a search could match several groups and should reach a particular one, a negative on the groups that should not receive it is the instrument that forces the outcome. The positive keyword alone cannot.

Over-applying at account level is the common error. It is simple, it works for a year, and then it blocks a search for a product the business has since launched.

Review the search-term report on a schedule. Not when a number moves — by then the money is gone. The cadence should match the account's volume.

Add the variants you see, not the ones you imagine, because negatives do not expand.

Prune annually. A list grown over years contains terms added for reasons nobody remembers and terms that no longer reflect what the business sells. Negatives are easier to add than to remove and a stale list quietly blocks demand.

Watch what blocking does to volume. A large addition of negatives narrows eligibility, which raises impression share and lowers volume at the same time. Both are expected; the question is whether the lost volume was worth having.

Keep a record of why. A shared list with no annotation becomes unmaintainable, because nobody can tell which entries were principled and which were a reaction to one bad week.

Negatives can only block what somebody has seen. Platforms do not report every search term, so a share of spend goes to queries that never appear in the report and cannot be refused individually.

That portion is one of the arguments for a conversion-based bid strategy alongside the negative list: the bidder declines what it does not expect to convert, including searches the advertiser will never see. The two controls cover different parts of the same problem, and an account relying on either alone is exposed on the side the other would have covered.

An account with no negative discipline can close most of the gap in an afternoon.

Start with the universal categories above as a shared list: free and cheap, jobs and careers, research and education, and the obvious adjacent products the business does not sell. Write out the variants rather than the roots.

Then read the last quarter's search terms sorted by spend, top down, and stop when the terms stop costing anything meaningful. The expensive irrelevant queries are almost always in the first page or two, and they are where the money went.

Then read the same report sorted by clicks with no conversions, which finds the steady small leaks the spend sort misses.

That exercise produces a list that covers most of an account's waste, and the ongoing work afterwards is maintenance rather than construction.

Adding negatives improves an account and makes several reported figures look worse, which is worth saying before the change rather than defending after it.

Impression volume falls, because eligibility narrowed. Click-through rate usually rises, because the remaining traffic is better matched. Impression share rises, because the denominator — the auctions the account was eligible for — got smaller. Cost per conversion falls, which is the point.

The one to watch is conversion volume. If it falls alongside the spend, some of what was blocked was working, and the list has gone too far. That is the check worth running a fortnight after a large addition.