What a match type controls
Every search campaign is a list of keywords, and every keyword carries a match type written in its notation: a bare keyword is broad match, a keyword in quotation marks is phrase match, and a keyword in square brackets is exact match. The notation tells Google how much latitude it has to connect a search to that keyword. Nothing else about the keyword changes: the bid, the ad group and the ads are the same whichever type is used.
What changes is which auctions the keyword enters. A broad match keyword enters many; an exact match keyword enters few. Since the advertiser pays per click (CPCAverage cost of each click.) on whatever searches the keyword matched, the match type is in effect a decision about which searches the budget is allowed to buy. That makes it a targeting decision rather than a technical setting, and it interacts with everything downstream: the ads written for the ad group, the landing page, the bidding strategy and the negative keywordsKeywords that stop an ad from showing for searches containing them. that trim what the match type lets through. The negative keywords guide covers the other half of that pair; this page covers the positive side and the choice between the three types.
One keyword can match a search only if the search is eligible under its type, and when several keywords in an account could match the same search, Google applies documented preferences to pick one. The broad strokes: an exact match keyword identical to the search is preferred, then other signals of relevance, then Ad Rank. In practice this means an account rarely needs to duplicate the same keyword in several match types to "cover" a search; the more useful question is which type the keyword should be. When a wanted search keeps landing in the wrong ad group or campaign, the diagnosis for searches matching the wrong ad group traces it back to the preference rule or missing negative that sent it there.
Broad, phrase, exact: documented behaviour
Google's documentation describes the three types in terms of meaning rather than words, and it is worth quoting the intent of each definition closely because older descriptions still circulate.
| Match type | Notation | Documented behaviour | Example keyword | Searches it may enter | Searches it should not enter |
|---|---|---|---|---|---|
| Broad | keyword | Searches related to the keyword's meaning, which can include searches that do not contain the keyword's terms; Google uses signals such as the landing page and other keywords in the ad group to understand intent | running shoes | trainers for jogging, best footwear for marathons, running shoe shop near me | Unrelated meanings, but the boundary is set by the system's understanding, not by the words |
| Phrase | "keyword" | Searches that include the meaning of the keyword; the meaning can be implied and the search can be a more specific form of it | "running shoes" | cheap running shoes, running shoes for flat feet, shoes for running | Searches where the meaning is absent, such as running socks or shoe repair |
| Exact | [keyword] | Searches that have the same meaning or intent as the keyword | [running shoes] | running shoes, runningshoes, shoes for running, running shoe | Searches that add a different intent, such as running shoes for kids |
Two points in the table deserve emphasis. First, the old broad match modifier, the plus-sign notation, no longer exists; Google folded its behaviour into phrase match, which is why phrase match now accepts reordered words when the meaning is kept. Second, the "should not enter" column describes the documented intent, not a guarantee. Google states what each type is designed to match; whether a particular search falls inside or outside that design is decided by its systems, and the search terms reportThe Google Ads report of the actual searches that triggered an account's ads. is the only record of the decision. Treat the table as the rule and the report as the evidence.
Broad match also differs in kind, not only degree. Google documents that broad match uses more signals than the keyword text: the content of the landing page, the other keywords in the ad group and, with Smart BiddingGoogle Ads' automated bid strategies that set each auction's bid from conversion signals., the signals available at auction time. A broad match keyword is less a keyword than a theme the system is allowed to interpret, which is why the ad group it sits in and the page it leads to shape what it matches.
Close variants and what "same meaning" means
Close variants are the reason none of the three types is a literal match. Google documents that all match types may show ads for close variants of the keyword, and lists the kinds of variation this covers: misspellings, singular and plural forms, stemmings such as "floor" and "flooring", abbreviations, accents, reordered words where the meaning is unchanged, added or removed function words (prepositions, articles and conjunctions that do not change the meaning), implied words, synonyms and paraphrases, and searches with the same intent.
"Same meaning" is therefore a judgement made by Google's systems rather than a rule an advertiser can apply by hand. For exact match, the documentation gives the shape of the boundary: the search must have the same meaning or intent, so "lawn mowing service" and "lawn mowing services" match the same exact keyword, while "lawn mowing equipment" should not. For phrase match, the keyword's meaning must be present in the search, so additional words that narrow the search are fine and words that change its subject are not.
Three practical consequences follow. An account no longer needs to list every spelling and plural as its own keyword; doing so creates duplicates that compete with each other for no gain. The search terms report will show searches that look unlike the keyword that matched them, and the match type column there labels which were close variants. And negative keywords, which do not use close variants, must list the misspellings and plurals explicitly if those are to be blocked. Where a close variant repeatedly matches a search the business does not want, the remedy is a negative, not a narrower positive keyword, because the narrower keyword would still accept the same variant.
Match types with Smart Bidding vs manual bidding
The match type decision cannot be made without knowing how bids are set, because the two were designed together. With manual cost-per-click bidding, the bid is fixed per keyword, so every search a keyword matches is bought at roughly the same price whether it is valuable or not. Under manual bidding, the match type is the main control over waste, and the conservative structure, exact and phrase match with a thorough negative list, is the sensible one.
Smart Bidding changes the calculation. Google documents that Smart Bidding sets a bid for each auction using signals available at that moment, including the actual search query, and recommends broad match with Smart Bidding for that reason: the strategy can bid high on a related search that looks likely to convert and low on one that does not, which is exactly the judgement a fixed bid cannot make. The Smart Bidding strategies guide covers the strategies themselves; the point here is that broad match without a conversion-based strategy hands the system reach without giving it a reason to be selective.
| Bidding | Broad | Phrase | Exact |
|---|---|---|---|
| Manual CPC | Avoid except with strong negatives and a monitoring routine | Workhorse for most ad groups | Core terms and brand |
| Maximise clicks | Avoid; the strategy buys cheap clicks wherever it finds them | Acceptable with negatives | Preferred |
| Target CPA / Maximise conversions | Viable once conversion data exists; pair with negatives and brand exclusions | Safe default | Core terms, and where tight control of the ad is needed |
| Target ROAS / Maximise conversion value | Viable with value-based tracking in place | Safe default | Core terms, brand, high-value products |
Industry practice adds a condition the documentation does not spell out: broad match with Smart Bidding needs enough conversion data for the strategy to have something to learn from. A new account with a handful of conversions a month is not giving the system the signal it needs, and the related searches broad match introduces will be bought on thin evidence. Start narrower, let conversions accumulate, then widen.
Choosing match types per campaign goal
The goal sets the tolerance for reach against control, and the table gives a starting point per common campaign. It is a decision aid, not a rule; the search terms report will adjust it within weeks.
| Campaign goal | Starting match types | Why |
|---|---|---|
| Brand protection | Exact and phrase on brand terms | Brand searches are precise; broad match would spend brand budget on generic searches |
| Lead generation, established account | Phrase for core services, broad on the best converters with Smart Bidding | Conversion data lets the strategy exploit related searches |
| Lead generation, new account | Phrase and exact | No data yet to make broad match selective |
| E-commerce, non-brand | Phrase for category terms; exact for product terms | Product searches are specific; category searches benefit from phrase match's flexibility |
| Competitor terms | Exact only, with careful ads | Broad or phrase on a competitor name wanders into their product lines and support queries |
| Research or discovery | Broad on a few themes with a modest budget | The purpose is to find searches; the report is the output |
Keyword lists for these campaigns come from research that the SEO cluster covers; the intent classification guide is the useful one here, because a keyword's intent, informational, commercial or transactional, is a good predictor of which match type it tolerates. Transactional terms tend to need the control of exact or phrase; informational themes are where broad match discovery pays, if the campaign's purpose is traffic rather than immediate leads. Where a term sits in the marketing funnel tells you the same thing from the other direction. Recording each keyword's match type and intent on one row of a search campaign build sheet before the build lets a reviewer check those choices against the campaign goal.
Reading the search terms report for match-type decisions
The search terms report lists the searches that triggered ads, with the keyword that matched and, in the match type column, whether it was a direct match or a close variant. It is the only evidence of what a match type is actually doing in an account, and reading it on a schedule is how match type decisions are made after launch. Google withholds searches below a privacy threshold and groups them as other search terms, so the report never shows everything; it shows enough.
Four readings are useful. A search that appears repeatedly, converts, and is not a keyword in its own right is a candidate for adding as a phrase or exact keyword, so that it gets its own ad and bid rather than riding on a broad match. A search that is irrelevant is a negative keyword candidate, and the match type that let it in tells you whether the problem is one keyword or the type as a whole. A broad match keyword whose searches are mostly off-topic is a candidate for narrowing to phrase, especially under a strategy that is not conversion-based. And an exact or phrase keyword whose close variants convert as well as its direct matches is evidence that the type is working as designed and does not need "protecting" with duplicates.
The report also reveals cost. Segmenting it by match type shows the cost per click and conversion rate of broad against phrase against exact within the same campaign, and the CPC calculator converts spend and clicks into a comparable rate. In most accounts broad match searches cost less per click and convert less often; whether the trade is worth it is a question the report answers and the match type theory does not.
A software company widens one ad group
A payroll software company runs an ad group on phrase match terms around "payroll software for small business" with a target CPA strategy and six months of conversion history. It adds broad match versions of the three best-converting keywords, keeps the phrase versions, and reviews the search terms report weekly. In the first month broad match spends £1,900 of the ad group's £6,400 and brings 41 trials at £46 each, against £58 for the phrase match terms. The report shows the new conversions came from searches such as "pay staff online" and "run payroll myself", which no phrase keyword would have reached. It also shows £310 spent on "payroll jobs" and "payroll course" before negatives were added. The company keeps broad match on these three keywords and declines to widen the rest until they have similar conversion history.
Migrating a legacy structure
Accounts built before close variants and broad match modifier were retired often carry a structure that no longer makes sense: the same keyword in three match types across three ad groups, hundreds of misspellings as separate exact keywords, modified broad keywords that were converted to phrase match automatically, and single-keyword ad groups that exist to control a match that is now controlled by the system anyway.
Migration is best done in stages, because the structure is also the history that Smart Bidding and quality ratings draw on. Start by consolidating duplicates: where the same keyword exists in several types in one ad group, keep the type that the search terms report shows doing the work and pause the rest. Next, remove the variant keywords that close variants already cover, checking each for conversion history first; a misspelt keyword with a long record of cheap conversions can stay. Then merge single-keyword ad groups whose keywords share meaning and intent into themed ad groups, so that one responsive search ad can serve them; the responsive search ads guide explains why the ad needs a theme rather than a single word. Finally, if the account moves to a conversion-based strategy, introduce broad match on a few proven keywords rather than converting the list wholesale, and watch Impression ShareShare of eligible impressions you received. and click-through (CTRShare of impressions that result in a click.) on the core terms to confirm nothing was lost. Quality ratings (Quality ScoreGoogle's 1–10 keyword diagnostic.) on consolidated keywords may reset or shift; that is expected and settles with history. To judge the move on evidence rather than a before-and-after comparison, the broad match expansion test runs it as a campaign experiment read on conversion value at target.
Google does not document how long the systems take to adjust to structural changes, and in practice a migration in a mature account is judged over weeks, not days. Keep a record of what was paused and when, so the change can be reversed if the report says so. The Google Ads expertise page describes the author's approach to restructuring inherited accounts.
If you are considering extending search campaigns beyond Google, the Google Ads versus Microsoft Ads comparison explains how match types, imports and audiences differ between the two platforms.
Two troubleshooting pages apply the ideas above to a symptom: the irrelevant search terms diagnosis traces why a given query matched, and the underspending campaign diagnosis covers keyword lists too narrow to use the budget.
Before comparing click prices across match types, check each keyword against its affordable ceiling using the CPC calculator for paid search at keyword level.
For the strategic choice of which end of the range a keyword or an account should default to, the comparison of broad match and exact match sets out when each is the right starting point and how to run them side by side.
Common mistakes
- Treating exact match as a string match. It is a meaning match; close variants apply, and the report will show searches that do not contain the keyword's words.
- Running broad match on manual bids. Broad match is designed for Smart Bidding; without a strategy that can bid selectively, it buys reach indiscriminately.
- Duplicating keywords across match types "for coverage". The preference rules mean one keyword usually does the work; the duplicates compete and confuse reporting.
- Listing misspellings and plurals as keywords. Close variants cover them; the list adds noise and duplicates.
- Forgetting that negatives have no close variants. A negative for the singular does not block the plural; both must be listed.
- Widening before there is conversion data. Broad match with a conversion strategy that has nothing to learn from is broad match with no steering.
- Never reading the search terms report. The match type is a rule; the report is what happened. Decisions made without it are guesses.
