What is Exact match?

Exact match makes a keyword eligible for searches that mean the same thing as it.

That is a different promise from the one the name makes, and the gap has widened in stages. The setting originally required the keyword's exact words in exact order. Then it accepted close variants: plurals, misspellings, accents, abbreviations, stemmings, and function words that do not change the meaning. It now matches on interpreted intent, which admits searches phrased entirely differently.

It remains the tightest of the three types, and the comparison is what makes it useful. Against phrase, which admits anything containing the meaning, and broad, which admits anything related, exact admits only equivalents. The set is small and coherent even if it is no longer literal.

Predictable intent. Every search that triggers an exact keyword means substantially the same thing, which means one ad can speak to all of them and one landing page can answer all of them.

Relevance. Because query, ad and page can be aligned tightly, exact-match keywords support the strongest relevance signals available, which is part of what rank rewards and therefore part of what an account pays.

Attribution you can reason about. Performance attached to an exact keyword describes a narrow, stable set of searches, so a figure attached to it means something over time. The same figure attached to a broad keyword describes a shifting population.

A clean home for known value. Terms the account has established are worth a particular amount are best held where the set cannot drift far.

Certainty. An exact-match keyword no longer tells you which searches will trigger it, only that they will be equivalent in the platform's judgement — and that judgement is not the advertiser's.

The failures this produces are specific rather than general. A same-meaning variant can be a different product in the advertiser's catalogue even though it reads as equivalent linguistically. It can be a free or open-source version of a paid thing. It can be a competitor's brand where the platform reads the category rather than the name. In each case the platform's semantic judgement is defensible and the commercial consequence is not.

Only negatives fix these, and they have to be found in the search-term report rather than anticipated.

The old practice of building accounts around exact match for control no longer achieves what it was for.

An account with thousands of tightly exact-matched keywords, each in its own group, was buying certainty at the cost of clerical effort. Now it buys a narrower set of same-meaning variants at the same clerical cost, and it fragments conversion data across many entities that each need volume for the bidding to work.

The modern shape is usually fewer, better groups: exact keywords where value is known and the ad should be specific, looser types where discovery matters, and negatives maintained continuously. The unit of control has moved from the match type to the negative list and the bid strategy.

Keep the ad and page specific. The reason to use exact match is alignment, and an exact keyword pointed at a generic page gives up most of the benefit.

Read the search terms anyway. Advertisers skip this for exact-match campaigns on the assumption that there is nothing to see. There is: the variants the platform admitted, some of which will not be wanted.

Harvest and promote. Search terms arriving through broader keywords that convert consistently belong as exact keywords with their own ads. This is how an account's structure improves over time, and it is the main use of the broader types.

Do not duplicate across types. Running the same keyword as exact, phrase and broad in parallel groups splits data and achieves nothing the platform's own arbitration does not already do.

Watch for collisions. Where a search could match several keywords in the account, the platform selects one, and an exact keyword matching the search identically is generally preferred. A keyword showing little volume may be losing that internal selection rather than losing auctions, which is a different problem with a different fix.

Exact match is worth using and not worth over-using.

It earns its place on the terms an account knows are valuable, where a specific ad and a specific page will do better than a generic one, and where a stable figure attached to a stable set of searches supports a decision.

It does not earn its place as the account's primary control mechanism, because it stopped being one. The controls are negatives, the conversion definition and the bid strategy, and an account still relying on match type for precision is protecting itself with a setting whose meaning changed underneath it.

If exact match has one unambiguous home, it is an advertiser's own name.

The traffic is valuable, the intent is specific, the volume is predictable and the cost is low. A brand campaign built on exact-match keywords with a tightly matched ad and the right landing page does exactly what the setting is good at, and the same-meaning expansion is mostly benign because the ways of expressing a brand name are limited.

Two cautions even here. Same-meaning matching can pull in a competitor's name where the platform reads the category rather than the brand, which is worth checking and blocking. And the category term that contains the brand — the generic product somebody is looking for alongside the name — is a different proposition and belongs elsewhere.

Because the set of triggering searches is narrow and stable, the figures attached to an exact keyword are unusually interpretable, and that is worth using.

A cost per conversion attached to an exact keyword describes substantially the same searches month to month, so a trend in it is a real trend rather than a change in what was being bought. The same cannot be said of a broad keyword, whose figures describe a population that shifts as the model learns.

This makes exact-match keywords the right place to hold the account's reference points: the terms whose economics are known, whose performance is tracked over time, and against which the rest of the account's discovery can be judged.