Google Ads Keyword Match Types in 2026: What Broad, Phrase, and Exact Actually Mean Now

Ask five people who run Google Ads accounts what “exact match” means and you’ll get five slightly different answers, and at least two of them will be wrong. Not because they’re bad at their jobs, but because the definition has moved three or four times since 2017 and nobody sent out a memo. The same is true of phrase match and broad match. If the mental model you’re using dates back more than a couple of years, you’re probably making targeting decisions based on rules that no longer apply.

This matters more than it sounds like it should. Match types are the mechanism that decides which searches your ads are even allowed to compete for, and in 2026 that mechanism runs through Google’s AI systems rather than a fixed rulebook. Get the mental model wrong and you’ll misread your search terms report, misjudge why a campaign is underperforming, or leave real money on the table because you’re avoiding broad match for reasons that stopped being true two product updates ago. This post walks through what each match type actually does today, why exact match isn’t exact, why Google is actively pushing advertisers toward broad match, and how to decide which approach fits your account without just copying whatever a blog post from 2021 told you to do.

What Each Match Type Actually Does in 2026

There are still only three match types you set directly on a keyword: broad, phrase, and exact. What changed is how much latitude Google gives itself within each one.

Broad match (no symbols, just the bare keyword: running shoes) is the widest setting. Your ad can show for searches Google considers related to the meaning of your keyword, even if the query shares no words with it at all. A broad match keyword for running shoes can trigger on “best trainers for marathon training” with zero literal overlap, because Google’s matching is working off intent and topical relevance rather than shared vocabulary. This is the match type Google’s own systems lean on hardest, and it’s also the one that can burn budget fastest if nothing is managing it.

Phrase match (in quotes: “running shoes”) requires that the meaning of your phrase be preserved in the query, with other words potentially added before or after. According to Google’s own definition, phrase match triggers when a search includes the meaning of your keyword, and word order matters enough that a query with unrelated words split through the middle of the phrase generally won’t match. In practice, phrase match in 2026 behaves considerably looser than it did five years ago, because “meaning” is now interpreted with the same intent-matching logic that powers broad match, just with tighter boundaries around it.

Exact match (in brackets: [running shoes]) is the narrowest, but “narrow” is relative. Google’s own documentation is explicit that exact match now covers close variants: the same search intent expressed with different words, not just the literal keyword string. That’s the piece most advertisers haven’t fully absorbed, and it’s worth its own section.

The practical throughline across all three: every match type in 2026 is intent-based to some degree. Broad match just gives Google the most room to work with, phrase match narrows that room, and exact match narrows it further, but none of them are purely literal string matching anymore. If you’re auditing a search term report and seeing queries that don’t obviously relate to your exact-match keyword, this is usually why, not a bug and not Google ignoring your settings.

A Short History of How We Got From Six Match Types to Three

It helps to know this wasn’t one decision, it was a decade of small ones that added up. Google originally offered a fourth option alongside broad, phrase, and exact: broad match modifier, or BMM, which used a plus sign in front of individual words to lock them in place (+running +shoes) while still allowing flexible matching around them. For years, BMM was the workhorse setting for advertisers who wanted more reach than exact match without the wide-open risk of plain broad match.

Close variant matching itself started earlier and separately, first extended to exact match back in 2014, letting exact-match keywords catch misspellings and simple plural variations without an advertiser having to manually list every spelling variant as its own keyword. That was a genuinely uncontroversial change at the time; almost nobody argued that a keyword should fail to match its own misspelling.

The bigger shift came in 2021. Google began folding BMM’s behavior into phrase match starting in February of that year, and by July, phrase match and BMM keywords behaved identically across every language, both matching based on the meaning of the keyword rather than a fixed word-order rule. Advertisers could no longer create new BMM keywords, and everything with a legacy + prefix simply became phrase match going forward. That single change is the real reason phrase match feels broader today than it used to: it absorbed an entire match type’s worth of flexibility into itself, in addition to picking up its own close variant expansions along the way.

From there, the close variant definition kept widening in both directions. Function word handling, synonym matching, paraphrase recognition, each of these arrived as its own incremental update rather than one dramatic overhaul, which is exactly why the cumulative change is so easy to underestimate if you haven’t been tracking it update by update. Six match types in 2015 became three by 2021, and the three that remain now each do more of what the old six used to split between them.

Why Exact Match Stopped Being Exact

The phrase “close variant” is the single most important piece of vocabulary in this entire topic, and it’s worth understanding precisely because it’s easy to misread as a minor technicality. Close variants for exact match include misspellings, singular and plural forms, abbreviations, and reordered words that don’t change the meaning of the query. But the definition goes further than that: it also includes the addition or removal of function words that don’t affect intent, implied words, and synonyms and paraphrases that express the same search intent as your keyword.

That last part is the one that surprises people. An exact match keyword like [emergency plumber near me] can now legitimately match a search for “urgent plumbing service” or “plumber available now,” because Google has judged that those queries carry the same underlying intent even though barely any words overlap. Ten years ago, exact match meant the literal string and nothing else. Today it means “the same thing you meant, expressed however the searcher happened to phrase it.”

This isn’t a secret Google is trying to hide. It’s documented plainly in their own help center, and it’s been expanding gradually since close variants were first introduced for exact match back in 2014, then extended further over the following decade to cover more categories of variation. The reason it still trips people up is that the expansion happened in increments small enough that no single update felt like a fundamental redefinition, but the cumulative effect over ten years absolutely is one. If you set up an exact-match campaign in 2019 and haven’t revisited your assumptions since, you’re running a campaign whose actual behavior has quietly drifted away from what you configured it to do.

Phrase match went through a parallel evolution and largely absorbed what used to be called “modified broad match,” the old system where a plus sign in front of a word (+running +shoes) locked that specific word in place while leaving flexibility around it. Modified broad match was retired as a standalone option back in 2021, and keywords still carrying the + prefix are now treated as phrase match. If you’re managing an older account and you spot plus-sign keywords still sitting in a campaign, they haven’t stopped working, but they’re not doing what they were originally set up to do either, and it’s worth cleaning that legacy syntax up so the account’s intent is legible to whoever looks at it next.

Google’s Case for Broad Match, and Why It’s Not Just Marketing

Google’s official position, laid out in its Smart Bidding help documentation, is that broad match paired with an automated bid strategy is the recommended default for most advertisers in 2026. That’s a real shift from the conventional PPC wisdom of a decade ago, where the standard advice was to start narrow with exact match and expand outward cautiously. Google now argues the opposite: start broad, let Smart Bidding find the valuable queries within that wider net, and use the bid strategy itself as the control mechanism instead of the match type.

The numbers Google cites to support that position are worth knowing, along with their limits. According to Google’s own data, advertisers who switch keywords from phrase match to broad match see, on average, about 25% more conversions in Target CPA campaigns and about 12% more conversion value in Target ROAS campaigns, while still meeting their targets. That data point traces back to Google’s internal analysis rather than an independent audit, and it’s aggregate advice across a huge range of account types, so it’s not a guarantee for any specific account. But it’s directionally consistent with what a lot of PPC practitioners have separately observed: broad match under Smart Bidding often outperforms phrase or exact match on volume without blowing up cost efficiency, provided the bid strategy has enough data to work with.

That “provided” is doing a lot of work. Google’s own guidance is explicit that broad match performs best when paired with Target CPA or Target ROAS bidding, not manual CPC, because the whole mechanism depends on Smart Bidding evaluating each individual auction’s context, the device, the time of day, the searcher’s history, and setting a bid tailored to that specific moment rather than a flat bid per keyword. Run broad match on manual bidding and you’ve removed the part of the system that was supposed to be doing the targeting work, which is a large part of why broad match earned its reputation as a budget-eating mistake in the pre-Smart-Bidding era. If you haven’t settled on which automated strategy fits your account, it’s worth working through the tradeoffs in choosing between Target CPA, Target ROAS, and Maximize Conversions before testing broad match on top of it.

Where Broad Match Still Goes Wrong

None of the above means broad match is automatically the right call for every account, and it’s worth being honest about the failure modes rather than just repeating Google’s talking points.

• Thin conversion history. Smart Bidding needs a real sample of conversion data to make good auction-by-auction decisions. Practitioner guidance generally puts the useful threshold somewhere around 30 to 50 conversions a month per campaign before broad match under automated bidding starts behaving reliably; below that, the algorithm is working with too little signal and results tend to be noisier and less predictable.
• Regulated categories. Healthcare, legal, financial services, and similar industries often have compliance constraints on what claims or services they can advertise for. Broad match’s willingness to expand into related-intent territory can pull in queries that sit outside what a regulated advertiser is actually permitted to bid on, which is a real operational risk, not just a wasted-spend one.
• No negative keyword discipline. Broad match without an actively maintained negative list is the classic way accounts bleed money. The wider net catches more irrelevant queries along with the relevant ones, and if nobody is reviewing the search terms report and adding negatives on a regular cadence, that irrelevant traffic just keeps getting paid for. This is the direct link between match type strategy and building a negative keyword list that actually cuts wasted spend; the looser your match types, the more that negative list is carrying the weight of keeping your account on-target.
• Brand and competitor bleed. A broad match keyword tied to a generic product category can occasionally surface next to competitor brand terms in ways an advertiser didn’t intend, depending on how Google interprets the relevance. Worth checking for specifically rather than assuming it won’t happen.

None of these are reasons to avoid broad match outright. They’re reasons to treat it as a strategy that requires active management, not a set-and-forget toggle. The accounts that get burned by broad match are usually the ones where someone flipped the setting and then didn’t look at the account again for a month.

The Negative Keyword Asymmetry Nobody Warns You About

Here’s the detail that catches even experienced advertisers off guard: everything above about close variants applies to positive keywords, the ones telling Google what to show your ad for. It does not apply the same way to negative keywords, the ones telling Google what to block. Negative keywords are matched far more literally, and they generally do not expand to catch close variants the way positive keywords do.

Concretely, that means if you add running shoe as an exact-match negative, it won’t automatically block the plural “running shoes,” a misspelling, or a synonym the way an exact-match positive keyword would automatically expand to include those. You have to add each meaningful variant as its own separate negative if you want it excluded. This is the exact opposite of how most advertisers assume the system works, since it’s natural to expect symmetry: if positive exact match is loose, negative exact match should be too. It isn’t, and the asymmetry is by design, not a bug, since a negative keyword that over-matched due to close variant logic could silently block legitimate traffic Google decided was a false-positive exclusion.

The practical upshot is that a negative keyword list built with the same mental model as a positive keyword list will systematically under-block. If you’re relying on broad match to find new query territory and negative keywords to fence off the parts that don’t convert, remember that fence has to be built with more granular, explicit entries than the keyword list it’s protecting against. This is one more reason the discipline behind building an effective negative keyword list matters more, not less, as your positive match types get broader: the two sides of the equation don’t expand at the same rate, and only one of them is doing you a favor by expanding automatically.

The AI Max Wrinkle: When Your Match Type Setting Gets Overridden Anyway

There’s a complication worth flagging if you’ve adopted AI Max for Search campaigns: it changes the match type conversation in a way that makes manually chosen match types partially moot. When AI Max is switched on at the campaign level, it treats your keywords as an input signal rather than a strict boundary, matching much closer to how broad match behaves regardless of what match type you originally assigned. We covered the mechanics of this in more detail in what AI Max for Search actually changes and what you still control, but the short version for this topic is: if you’re running AI Max, spending a lot of time debating broad versus phrase versus exact at the keyword level is largely academic, because the campaign-level setting is already doing the widening for you.

That’s not automatically a bad thing, Google’s internal data suggests campaigns that were previously running mostly on exact and phrase match see some of the largest gains from turning AI Max on, since it’s opening up query coverage those campaigns were previously missing entirely. But it does mean match type strategy and campaign-level settings need to be considered together now, not as two separate decisions. Setting careful exact-match keywords and then turning on a feature that treats them as broad match anyway is a mismatch worth catching before it costs you a month of budget you didn’t mean to spend that way.

A Practical Framework for Choosing Match Types

Rather than a single universal recommendation, here’s how to think about it based on where your account actually is.

New campaign, no conversion history yet. Start with phrase match as your primary match type, with a handful of tightly targeted exact-match keywords for your highest-intent terms. You don’t have the conversion volume yet for Smart Bidding to make good use of broad match’s flexibility, and you want to build a search terms report you can actually learn from before opening the net wider. Treat this as a data-gathering phase, not a permanent structure.

Established campaign, healthy conversion volume, running Target CPA or Target ROAS. This is the scenario Google’s guidance is actually written for, and it’s worth a genuine test. Duplicate the campaign or a subset of ad groups, switch the copy to broad match, keep the bid strategy and budget comparable, and let it run long enough to clear the learning period, generally a couple of weeks at minimum, before comparing results against your existing phrase or exact match structure. Watch conversion volume and cost efficiency together, not just one or the other.

Regulated industry or tightly constrained offer. Lean toward phrase and exact match with an actively maintained negative list, and be cautious about broad match even under Smart Bidding, since the compliance risk of showing on the wrong query can outweigh the efficiency gains. This is one of the few scenarios where the older, more conservative match type philosophy still holds up well against Google’s current default recommendation.

Running Performance Max or AI Max alongside Search. Recognize that keyword-level match type decisions matter less than they used to for the campaigns where those systems are active, and put more of your attention into negative keyword lists, audience signals, and asset quality, the levers that still function as real constraints inside an AI-driven matching system.

Any account where nobody has reviewed match type strategy in over a year. Assume your assumptions are stale. Pull a fresh search terms report, look at what’s actually triggering your ads today, and compare it against what you’d expect based on the match types you think you’re running. The gap between those two things tells you how much close variant expansion has quietly reshaped your account without anyone deciding it should.

How to Test This Without Wrecking Your Account

The temptation with any match type change is to flip the switch on your whole account at once and see what happens. Don’t. A few habits make this much safer to test properly.

Change one variable at a time. If you’re testing broad match, don’t also change your bid strategy, your budget, and your ad copy in the same week. You won’t be able to tell which change caused which result, and if performance dips, you’ll spend more time untangling the cause than you would have spent just testing sequentially in the first place.

Give it a real sample size before judging it. Smart Bidding strategies go through a learning period after any significant change, typically one to two weeks, during which performance can look worse than it will settle into once the algorithm has enough fresh data. Pulling the plug on a broad match test after three days because the numbers looked rough is one of the most common ways advertisers talk themselves out of a strategy that would have worked if they’d let it run.

Check the search terms report more often during the test than you normally would. This is where you’ll catch irrelevant matches early, before they’ve accumulated into real wasted spend, and where you’ll build the negative keyword list that makes broader match types sustainable rather than just expensive.

Keep a control group. Whether that’s a duplicated campaign, a held-out set of ad groups, or simply comparing to the prior month under the old match type structure, you want something to measure against. A single before-and-after comparison across an account that’s also dealing with seasonality, competitor changes, or other simultaneous shifts is a weak signal on its own.

Write down what you changed and when, even if it’s just a note in a shared doc. This sounds like overkill until three months from now, when someone, possibly you, is trying to explain why a campaign’s conversion volume shifted in September and can’t remember whether that was the broad match test, a seasonal pattern, or something else entirely. A one-line log entry per change (“switched Ad Group X to broad match, MM/DD, Target CPA unchanged”) takes fifteen seconds to write and saves an afternoon of guessing later. If you manage accounts for clients, this same log doubles as the explanation you’ll eventually need to give them, and it’s a much stronger answer than reconstructing the timeline from memory during a call.

This kind of testing is exactly where an account benefits from being checked more frequently than a weekly glance allows. Match type changes interact with bid strategy learning periods, seasonal query shifts, and negative keyword coverage all at once, and a problem that starts small on day one can compound by day ten if nobody’s watching the interim results. That’s part of what continuous daily account monitoring is built for: catching a broad match test that’s drifting into irrelevant queries, or a search term spike that needs a negative keyword added, before it turns into a wasted week of budget rather than a two-day blip that gets corrected quickly.

What This Means for Your Search Terms Report

Once you internalize that every match type now operates on intent rather than literal text, your search terms report reads differently. A query that seems unrelated to your exact-match keyword on the surface isn’t necessarily a sign that Google is ignoring your settings; it might be a legitimate close variant that shares intent without sharing vocabulary. The useful question shifts from “did this query literally match my keyword” to “does this query represent the same buyer intent I was targeting, regardless of the specific words used.”

That reframing changes how you build negative keywords too. Instead of only excluding queries that are obviously off-topic, you also want to watch for queries that share surface-level similarity with your target intent but represent a genuinely different buyer need, the kind of near-miss that close variant matching is more likely to let through than old-style literal matching ever would have. A search terms review done with 2026’s matching behavior in mind catches a different set of problems than the same review would have caught in 2018, even on an account that hasn’t changed its keyword list at all.

It’s also worth checking this reasoning against a second signal when you have one. If a query looks like a legitimate close variant but conversion rate on it is meaningfully worse than your core terms, that’s a useful sign the intent match isn’t as clean as Google’s system judged it to be, even if the words are close. Treat the search terms report as a starting point for judgment, not a verdict, since the algorithm is inferring intent statistically and it won’t always get that inference right for your specific offer.

Summary

Broad, phrase, and exact match are still the three levers you set, but what each one actually does has shifted substantially, and exact match in particular no longer means what the name implies. All three match types now factor in searcher intent through close variants, misspellings, synonyms, reordered words, and paraphrases that mean the same thing, with broad match giving Google the widest interpretive latitude and exact match the narrowest, but narrowest is no longer the same as literal.

Google’s own guidance now points advertisers toward broad match paired with Target CPA or Target ROAS bidding as the default approach for accounts with enough conversion data to support it, and the performance data behind that recommendation is real, even if it’s not universal. Whether that’s the right call for your account depends on your conversion volume, how regulated your industry is, how disciplined your negative keyword process is, and whether AI Max or Performance Max are already doing some of that widening for you at the campaign level regardless of what match type you’ve assigned. Test deliberately, one variable at a time, with a real sample size before you judge the result, and keep an eye on the search terms report throughout, not just at the end. The mechanism behind match types has changed more in the last few years than most advertisers realize, and the accounts that adapt their strategy to match, rather than running on assumptions from an earlier version of the platform, are the ones getting the benefit of where this is actually headed rather than fighting against it.