Google Ads Keyword Research & Match Types Explained

If you learned Google Ads keyword match types more than a couple of years ago, some of what you know is outdated. Match types used to work on literal text matching — phrase match required your exact phrase to appear intact, exact match required (with narrow exceptions) the precise keyword. As of 2026, all three match types work on inferred meaning and intent instead, which changes what "tight control" actually means and, more importantly, what it doesn't mean anymore.

This covers how each match type actually behaves now, why the shift to intent-based matching matters practically (not just conceptually), a real methodology for building a keyword list from scratch, and how to choose match types by what a specific campaign is actually trying to do.

Everything below reflects how match types behave as of late 2026 — worth treating as a snapshot of current behavior rather than a permanent specification, since Google has continued adjusting this specific area of the platform and is likely to keep doing so.

The three match types, how they work now

Broad match — the keyword with no punctuation — has always been the widest net, and it remains so. What's changed is how Google decides what counts as related: rather than matching primarily on shared words or simple synonyms, broad match now draws on the meaning of your keyword combined with contextual signals (your landing page content, your other keywords, and — when you're running Smart Bidding — real-time auction and conversion signals) to decide which searches are close enough in intent to trigger your ad. This makes broad match considerably more capable than it used to be when paired with Smart Bidding and a solid negative keyword list, and considerably riskier without either.

Phrase match — the keyword in quotes — sits in the middle. It's meant to match searches that carry the same meaning as your phrase, with word order mattering less strictly than it used to; a search reordering your phrase's words while preserving its meaning can now match, which wasn't reliably true under the older, more literal phrase-matching logic. Phrase match absorbed the functionality of Broad Match Modifier (BMM) when Google retired BMM as a separate match type in 2021 — if you see BMM referenced in older content, it no longer exists as a distinct option; what it used to do now lives inside how Phrase Match itself behaves.

Exact match — the keyword in brackets — remains the tightest available control, but "tightest" is a relative statement, not an absolute one anymore. Google's own guidance confirms exact match can still trigger on close variants: different word order, singular versus plural forms, common misspellings, and now, meaningfully, same-intent rewrites that don't share the literal wording at all. A search that means the same thing as your bracketed keyword, even phrased quite differently, can match under exact match today in a way it typically wouldn't have five years ago.

It's worth checking the specific matching behavior on your own core keywords directly, using the Ad Preview and Diagnosis tool with a handful of real, varied search phrasings — reading a general description (including this one) gives you the underlying logic, but seeing your own exact-match keyword actually trigger, or not trigger, for a specific reordered or synonym-based search tells you something concrete about how it's behaving in your specific account right now.

Why this shift matters practically, not just conceptually

The practical implication that most guides state without following through on: no match type, including exact, is fully insulated from irrelevant or unexpected queries anymore. Under the old literal-matching system, choosing exact match was itself a meaningful form of protection against wasted spend — the query genuinely had to match your bracketed term closely. Under the current intent-based system, exact match still narrows your reach considerably compared to broad or phrase, but it no longer guarantees the same degree of query-level precision it once did.

This means the search terms report and active negative keyword maintenance — covered in full depth in our dedicated negative keywords guide — matter for every match type, not primarily for broad match the way older conventional wisdom framed it. An account running entirely on exact match keywords with nobody reviewing the search terms report is still exposed to a real, if smaller, stream of unexpected queries slipping through under the current matching logic, and treating exact match as a reason to skip that review entirely is no longer a safe assumption.

The upside of the shift is genuine too, worth stating plainly rather than only covering the risk side: broad match paired with strong Smart Bidding and a solid negative list can now surface genuinely relevant demand that older, more literal broad matching would have missed entirely — new phrasings, emerging search patterns, and related intent that a rigid keyword list wouldn't have anticipated. The tradeoff is real in both directions, not a one-sided downgrade.

This also changes how much weight to put on match type alone when diagnosing a wasted-spend problem. Finding that a chunk of wasted spend came through broad match doesn't automatically mean broad match itself is the problem — under the current intent-based system, tightening to phrase or exact match reduces but doesn't eliminate the same category of risk, and the underlying fix (better negatives, tighter ad group themes) matters more than which match type symbol happens to be attached to the keyword.

Budget a specific, recurring block of time for this regardless of which match types you're running — treating the search terms report as an occasional check rather than a scheduled habit is how even a carefully built, intent-categorized keyword list quietly drifts toward waste over the following months.

A closer look at what changed, concretely

It's worth a few concrete before-and-after examples, since "matching now works on intent" can sound abstract until you see it applied. Under the old, literal exact-match system, the bracketed keyword [running shoes] would reliably match only "running shoes" and perhaps very close variants like plural or minor misspellings — a search for "shoes for running" or "best sneakers for jogging" would not have triggered it. Under the current system, both of those reordered or synonym-based searches can plausibly match, because Google is evaluating whether the search shares the same underlying meaning and intent as your bracketed keyword, not just checking the literal string.

The same shift applies to phrase match. The older system for "running shoes" (in quotes) required that exact two-word phrase to appear somewhere in the query, in that order — "buy running shoes online" would match, "buy shoes for running" would not, since the words were reordered and a word was inserted between them. Under current matching, that second query's shared meaning is enough to plausibly trigger a match despite the reordering, which is a genuinely different behavior than what the same match type produced a few years ago.

This is worth internalizing specifically because plenty of advertisers, and plenty of published guides, still describe match types using the older, more literal logic — testing your own account's actual behavior against a handful of real searches, using the Ad Preview tool or reviewing your own search terms report closely, is more reliable than assuming any written description (including this one) perfectly captures Google's current matching behavior in every case, since this is an area Google continues to actively adjust.

None of this means match types no longer matter or that the distinction between them has collapsed entirely — broad, phrase, and exact still produce meaningfully different reach and control in practice, and the FAQ above on checking your own account's behavior directly remains the most reliable way to understand exactly where those boundaries currently sit for your specific keywords.

A keyword research methodology

Start with seed keywords drawn from your own product or service pages, not from a tool — the words and phrases you'd naturally use to describe what you offer are usually a more accurate starting point than a generic industry list, since they reflect your actual positioning rather than a category-wide average.

Expand each seed using Google's Keyword Planner (inside Google Ads) and, separately, Google's own autocomplete and "People also ask" results for the same seed terms searched directly — the two sources surface genuinely different related terms, since Keyword Planner draws on aggregated search volume data while autocomplete reflects real, current query patterns more directly.

Categorize the resulting list by intent, since not every keyword deserves the same treatment. Commercial and transactional terms — where the searcher is clearly close to a purchase or action decision ("buy," "pricing," "near me," "book") — deserve your tightest match types and highest bids. Informational terms ("how to," "what is," "guide") generally convert at a meaningfully lower rate and are better suited to broader, lower-bid treatment or excluded entirely depending on your funnel strategy, covered in more detail in our beginner's guide's discussion of funnel stage. Navigational terms — searches for a specific brand, including your own or a competitor's — deserve their own dedicated, isolated campaigns, covered in our campaign structure guide.

Assign match types last, once the list is categorized by intent — this ordering matters, since choosing match types before you understand which keywords represent which intent level means making that decision on incomplete information.

Long-tail keywords, and why they still matter under intent-based matching

Long-tail keywords — longer, more specific phrases with lower individual search volume — remain genuinely valuable even though broad match's improved intent-matching can now surface some of that same traffic without you needing to add every long-tail variation explicitly. The reason: bidding on a specific long-tail term directly gives you control over its bid and its dedicated ad copy that letting broad match catch it incidentally doesn't.

A search for "emergency 24 hour plumber for burst pipe" reflects meaningfully higher urgency and intent than the shorter "plumber" it's related to, and a dedicated ad group with copy speaking directly to that urgency ("Emergency service, arriving within the hour") will generally outperform a broader ad trying to cover both the short and long-tail version of the same underlying need with one generic message. Long-tail keywords, in other words, aren't obsolete just because broader match types have gotten better at finding related traffic — they're still worth deliberately targeting whenever the specific phrasing reflects a meaningfully different intent or urgency level worth addressing with dedicated, tailored ad copy.

A practical way to find genuine long-tail opportunities worth this dedicated treatment: scan your search terms report for longer, more specific queries that are already converting well despite not being keywords you deliberately targeted — these are effectively pre-validated long-tail candidates, since real searchers have already demonstrated the phrasing and the intent behind it, rather than requiring you to guess at long-tail variations from scratch.

Keyword research tools beyond Keyword Planner

Google's Keyword Planner is free, directly tied to real Google search volume data, and the right starting point for most accounts — but it has real limitations worth knowing about. Search volume figures are shown as ranges rather than exact numbers for most accounts (exact figures are generally reserved for accounts with meaningful active spend), and it can undersell genuinely long-tail, highly specific query variations that don't aggregate into its broader keyword groupings.

Your own site search data, if your website has an internal search function, is an underused and genuinely valuable source — the exact phrases visitors type into your own search bar reflect real intent from people already on your site, often surfacing phrasing your Keyword Planner research wouldn't have generated on its own.

Your existing search terms report, once a campaign has been running for a while, becomes one of the best keyword research sources available, ironically after the fact — queries that are already converting well but that you never deliberately targeted as keywords are strong candidates to add explicitly and bid on directly, rather than continuing to rely on broad or phrase match to catch them incidentally.

Third-party paid tools (Ahrefs, SEMrush, and similar) add competitive keyword intelligence — what terms competitors are ranking or bidding on — that Google's own free tools don't expose, useful supplementary research though worth the same directional-rather-than-precise treatment covered in our automation tools guide regarding third-party estimates generally.

Whichever combination of sources you use, cross-check any candidate keyword against actual, current search behavior before committing meaningful budget to it — a keyword that looked promising in a planning tool can behave quite differently once it meets real searchers and Google's current intent-based matching, which is exactly why the search terms report closes the loop that upfront research alone can't.

A worked example: applying the methodology

Say you run a local moving company. Seed keywords pulled directly from your own service pages: "local movers," "long distance moving," "packing services," "moving company." Expanding "local movers" through Keyword Planner and autocomplete surfaces related terms: "movers near me," "affordable local movers," "same day movers," "moving company quotes," "how to hire movers," "moving company jobs."

Categorizing by intent: "movers near me," "affordable local movers," and "moving company quotes" are commercial/transactional — someone close to hiring. "How to hire movers" is informational — someone still researching, potentially weeks from a decision. "Moving company jobs" is navigational-adjacent but entirely wrong intent — someone looking for employment, not a moving service, and it belongs on a negative keyword list rather than anywhere in the positive keyword set at all.

Assigning match types: the clearly commercial terms go in as phrase match initially, in a tightly themed "Local Moving - Commercial" ad group. The informational term either gets excluded entirely (if your funnel strategy is bottom-of-funnel only) or goes into a separate, lower-bid ad group if you're deliberately running upper-funnel content too. "Moving company jobs" goes straight onto the negative keyword list, never into the positive keyword set at all — a common mistake is treating a clearly mismatched term as simply "low priority" rather than actively excluding it, which leaves the door open for it to trigger spend anyway.

This same three-step process — seed from your own pages, expand and categorize by genuine intent, assign match types last — scales to a much larger keyword list than this small moving-company example shows; the specific number of keywords changes, but the sequence and the reasoning behind each step stay the same regardless of account size or industry.

Choosing match types by campaign goal

For a new campaign or a keyword you haven't proven yet, start with phrase or exact match rather than broad, even in the current intent-based system — you don't yet have the conversion history Smart Bidding needs to use broad match's wider net well, and starting tighter lets you build that history on more controlled, interpretable data first.

Once a specific keyword theme has proven itself — meaningful conversion volume at an acceptable CPA over a real evaluation window — testing broad match on that same proven theme, with Smart Bidding and a solid negative list already in place, is a reasonable way to discover additional relevant demand the tighter match types were never going to surface on their own.

This progression — start controlled, expand deliberately once proven — is a different structural decision than the single-keyword-ad-group approach some older content still recommends, loading one keyword into all three match types simultaneously inside dedicated per-match-type campaigns. As covered in detail in our campaign structure guide, that approach fragments conversion data in a way that works against Smart Bidding's need for pooled volume; a single-theme ad group with a deliberately chosen match type, expanded to a second match type only once proven, keeps data pooled while still giving you real match-type control.

Keep your own brand terms on exact match specifically, regardless of your broader match-type philosophy elsewhere in the account — brand terms are the one case where the tightest possible control is almost always worth it, since you already own the relevance and cost advantage on your own name and there's little upside to broadening that specific traffic's reach.

None of this is a rigid formula to apply identically to every keyword in every account — the underlying principle (start controlled, expand deliberately once proven, keep brand terms tight regardless) is the transferable part; the specific match type you start any individual keyword on should still reflect your own read of how proven and how commercially clear that specific term's intent already is.

Should I still use all three match types for the same keyword?

Generally not simultaneously in the same ad group — running the same keyword across multiple match types at once tends to create internal competition between your own keywords in the same auction and complicates understanding which match type is actually responsible for a given result. A more effective approach, covered above, is choosing one match type deliberately based on how proven the keyword is, and expanding to a broader match type only once you have a specific reason to test additional reach.

How do I know if broad match is working or just burning budget?

Check the search terms report specifically for broad-match-triggered queries, filtered by cost and conversion — the same review process covered in our negative keywords guide applies here directly. Broad match "working" means it's surfacing genuinely relevant, converting queries you wouldn't have thought to add manually; broad match "burning budget" means cost accumulating on searches only tangentially related to your actual offer. The two patterns look different in the search terms data even though both technically fall under the same match type.

Does Smart Bidding actually need broad match, or can it work with exact match alone?

Smart Bidding functions with any match type — it doesn't require broad match specifically to operate. What it needs is sufficient conversion volume, covered in our KPIs guide, regardless of which match types are generating that volume. Broad match tends to accelerate reaching that volume threshold faster, since it captures more total traffic, but a tightly-matched account with genuinely high search volume on its exact and phrase-match terms can feed Smart Bidding adequately without broad match playing any role at all.

This is worth testing directly on your own account rather than assuming either way — some accounts genuinely reach strong Smart Bidding performance on exact and phrase match alone, while others see a meaningful volume and performance lift once broad match is added with proper guardrails, and the only reliable way to know which describes your specific account is checking your own data after a deliberate test.

How often should I revisit my keyword list and match type choices?

A full keyword and match-type review is worth doing on the same quarterly cadence covered in our optimization tips guide — search behavior and your own product offering both shift over time, and a keyword list built a year ago may no longer reflect either your current business or how people currently search for it. The search terms report review itself, by contrast, deserves a much tighter weekly or monthly cadence, since that's where near-term drift and waste actually show up first.

What is Dynamic Search Ads, and does it replace keyword research entirely?

Dynamic Search Ads (DSA) generates ads and matches them to searches based on your website's content directly, rather than a keyword list you've built manually — genuinely useful for discovering keyword gaps your manual research missed, or for covering a large product catalog where building individual keywords for every item isn't practical. It doesn't replace deliberate keyword research and match-type strategy for your core, highest-value terms, though; most accounts running DSA use it as a supplementary discovery layer alongside a deliberately built Search campaign, not as a full substitute for one.

Should I include my competitors' product names as keywords, not just their brand name?

This falls under the same territory covered in our dedicated guide on competitor brand keywords — the same legal and economic considerations (Quality Score disadvantage, lower conversion rate) apply to a competitor's specific product name as much as to their overall brand name, and it's worth evaluating with the same break-even scrutiny covered there rather than treating product-name targeting as automatically safer or more effective simply because it's more specific than the brand name alone.

The short version

Match types work on intent now, not literal text — which means broad match is more capable than it used to be, and exact match offers less absolute protection than it used to. That means the search terms report and negative keyword discipline matter for every match type, not just broad. Build your keyword list from your own product pages first, categorize by genuine intent before assigning match types, start controlled and expand deliberately once a theme is proven, and keep your brand terms on exact match regardless of your broader philosophy elsewhere in the account.

Treat every match type choice as a starting point to validate against real data, not a permanent, set-once decision — the underlying platform keeps evolving, and the accounts that adapt their keyword strategy alongside it consistently outperform the ones still operating on assumptions from a few years ago.

Can I exclude a specific close variant that exact match now catches, if I don't want it?

Yes — if exact match's expanded, intent-based matching pulls in a specific close variant or same-intent rewrite you genuinely don't want, adding that specific query as a negative keyword (at whatever level is appropriate, covered in our negative keywords guide) excludes it going forward, the same way you'd exclude an unwanted broad or phrase match query. This is exactly why the search terms report review matters for exact match now too, not just the broader match types — it's your mechanism for catching and correcting exactly this kind of unwanted expansion.