The match-type decision most accounts are still making is the one from 2019: broad for reach, exact for control, phrase as the safe middle. That map no longer matches the territory. In 2026 all three match types match on meaning rather than on the literal words, and exact match has expanded to include close variants, reordered words, implied words, and paraphrases Google judges equivalent. Phrase match now fires on synonyms and same-meaning variations too. The three types have converged toward each other, and the old mental model quietly stopped protecting you.
That means the choice is no longer "how literally do I want this keyword read." It is "how wide an intent net can my account actually absorb this week." The answer depends on your conversion data, your bidding, and — above all — whether you have the search-term review cadence to catch what a wider net drags in. This post lays out what changed, why phrase match lost its job, and a decision framework you can apply to your own keywords without guessing.
What actually changed: all three types expanded
Every match type in 2026 matches on intent, not on the string. Broad match uses machine learning to serve on searches loosely related to the keyword. Phrase match now shows for search terms that carry the meaning of your phrase, including synonyms, misspellings, singular and plural forms, and reordered words. Exact match, the type whose entire promise was literalness, now matches same-meaning queries, implied words, and paraphrases. The distinguishing feature used to be how strictly each type read the query; that feature has largely evaporated.
What is left is a difference of degree, not of kind. The three types now describe how wide the intent net is — broad casts the widest, exact the narrowest — but all of them cast a net rather than a straight line. This is why a clean keyword build no longer protects you the way it once did: the matching happens at serve time on Google's reading of intent, and no match type opts out of that. The practical consequence is that the search terms report is now the only reliable record of what your keywords actually bought, regardless of which type you chose.
Why phrase match lost its job
Phrase match used to be the compromise: more reach than exact, more control than broad. That role depended on phrase match reading the query fairly literally — the words, in order, with room around them. Once phrase expanded to synonyms, reordered words, and same-meaning variants, its search-term output started to look like a slightly narrower version of broad match rather than a wider version of exact. The middle position it occupied collapsed because the thing that defined the middle — literal-ish reading — no longer exists. Search Engine Land has documented phrase match losing ground to broad match as the two converge.
This does not mean phrase match is useless — it means it is no longer a safe default. If you pick phrase expecting the controlled middle and then never open the report, you get much of broad match's query variety with none of broad match's scale advantages. The honest framing for 2026 is that the real decision is broad versus exact: how much intent variety you want to invite, and how tightly you want to fence. Phrase is a tactical tool for anchoring a specific phrase in its own ad group, not the resting default for a whole account.
The real decision: reach versus control, gated by data
The choice that matters is how much search-term variety your account can absorb without leaking money — and that is gated by your conversion data, not by preference. Broad match plus Smart Bidding is a powerful combination, but it only works when the conversion signal is strong enough to steer the algorithm toward the good queries inside the wide net. With a trained signal, broad can find converting searches you would never have thought to add as keywords. Without one, broad is just an expensive way to discover which queries you should have negated first.
So the gate is data maturity. A high-volume account with clean conversion tracking and months of history can run broad match and let Smart Bidding do the sorting, catching the misses with a weekly negative pass. A thin or new account cannot — it should stay tight, keep the early search terms readable, and widen only once the signal can guide the wider net. The mistake in both directions is the same: choosing a match type that does not match the amount of search-term review the account can actually sustain. Match type is a throttle on how much query variety you invite; set it to what you can read.
A decision framework you can run this week
Work it as a short sequence rather than a preference. First, check the conversion signal: does the campaign have enough recent conversions for Smart Bidding to steer? If not, default to exact and phrase and stop there until it does. Second, check your review cadence: can you commit to a weekly search-terms pass and a maintained negative list? If not, a wide match type will outrun your ability to police it, so keep the net narrow. Third, check the goal of the campaign: prospecting and discovery justify broader matching; protecting a proven, high-intent term argues for exact.
Then structure so the decision is legible. Put exact keywords in their own ad groups and broad keywords in others, so that when you read the report you can see which match type produced which query and fence each independently. Segmenting by match type is what lets you add a negative to the broad ad group without starving the exact keyword that shares the theme — the discipline the negative fencing post covers. Run the framework per campaign, not once for the account: a brand campaign and a cold prospecting campaign have different data and different goals, so they earn different match types.
Match type is only half the system
Choosing a match type sets how wide the net is; it does not decide what the net keeps. That is the job of negatives, and the two levers only work together. A broad keyword with a strong negative list behaves like a well-fenced discovery engine; the same keyword with no negatives behaves like a budget leak. Conversely, an exact keyword still needs negatives because exact match degradation lets it trigger on close variants and paraphrases you never chose — as the exact match degradation post details, even the tightest type leaks in 2026.
Read the two together every time. When you widen a match type, expect the search-term variety to rise and budget a heavier negative pass to match. When you tighten it, you can ease off the negatives but never drop them to zero. The same care applies to the negatives themselves: over- negating a wide match type strangles the reach you widened for, the failure mode the over-negating post lays out. The account that wins in 2026 is not the one that picked the "right" match type — it is the one that matched its match type to its negatives and its review cadence, and kept all three in balance.