Broad, phrase and exact match govern which searches a Search keyword can reach. Exact and phrase give a constrained service campaign a focused starting point.
Broad match deserves assessment with Smart Bidding and useful conversion signals, not a blanket ban or recommendation. Because no match type guarantees relevant leads, check what your keywords matched in the search terms report.
What broad, phrase and exact match mean
Keyword versus search term
The first thing to separate is the word you typed from the search a person made.
Match types set how far apart those two things are allowed to be. That is the whole job. They do not choose the location, the schedule, the bid or the page. Nor do they protect a campaign from a bad offer. Our guide to setting up Google Ads for a small business covers the prerequisites around them.
Syntax and matching meaning
Google's keyword matching options page defines the three settings by meaning rather than by words. Broad match keywords can show on searches related to the keyword. That includes some searches that do not carry its direct meaning.
Phrase match keywords can show on searches that include the meaning of the keyword. That meaning can be implied. The search can also be a more specific form of it. Exact match keywords can show on searches with the same meaning or intent as the keyword.
Broad is the default setting for every new keyword. Broader match types capture all the queries of the narrower ones plus more. A match only makes the ad eligible for the auction. Whether it shows still depends on the auction, the budget and the campaign's targeting.
Match type and syntax | What it can match | Relative reach | Practical limit |
|---|---|---|---|
Broad: plumber brunswick | Searches related to the keyword, including some without its direct meaning | Widest, and the default for new keywords | Needs Smart Bidding, useful conversion data and weekly review to stay relevant |
Phrase: "plumber brunswick" | Searches that include the meaning of the keyword, implied or more specific | Between the other two | Word order still matters where it changes the meaning |
Exact: [plumber brunswick] | Searches with the same meaning or intent, including close variants | Narrowest | Low volume terms can go inactive and a long list needs maintenance |
The syntax is the easy part. Plain text is broad, quotation marks are phrase, square brackets are exact. The hard part is the word "exact". It sounds literal, and it is not.
Google's close variants page lists what counts. For exact match that includes misspellings, singular and plural forms, stemmings, abbreviations and accents. It also includes:
reordered words with the same meaning
function words added or removed
implied words
synonyms and paraphrases
searches with the same intent
The next two examples are made up, not matches from an account. The keyword [plumber brunswick] could also be matched by "plumbers in brunswick" through a function word. It could be matched by "brunswick plumbing" through a stem and the same intent.
Phrase match does not require a literal uninterrupted string either. Google's page on the phrase match change gives "moving services NYC to Boston" matching "NYC corporate moving services to Boston". It does not match a search about moving from Boston to New York. That's because the order changed the meaning.
The same page records that broad match modifier was merged into phrase match in 2021. New keywords cannot be added with the plus sign notation. The plus sign is not a fourth match type. So an account still carrying it is running phrase match under an old label.
Choose a starting point for your business
Match types come after the keyword list. Our guide to Google Ads keyword research helps you build that list. The choice then depends on the offer, the measurement and the budget. So here are four short contrasts. The searches in them are examples, not matches from Elev8d accounts.
A plumber in Brunswick wants calls for urgent repairs. The service terms are "emergency plumber brunswick" and its close relatives. Exact and phrase on those terms give a small budget a focused start. The call conversions that arrive over the first weeks become the evidence a broader test would need.
Broad match on day one starts with no conversion history and a $40 a day budget. That asks the system to learn what a lead is from almost nothing.
A conveyancer in Geelong sells a defined service with a defined price. Exact match on the service terms that convert fits a narrow offer. So does phrase match on the service plus place variants. Broad match sits next to adjacent meanings: DIY contract kits, free templates and jobs.
If broad match is tested, the test runs with Smart Bidding. Its conversion means a booked appointment rather than a page view. The firm's name sits in its own ad group so brand demand never pads the result.
An online store running a Search campaign for work boots faces the opposite problem. Product searches arrive as brand, model, size and colour in every order. An exact list that tries to enumerate them never finishes.
Phrase match on the product themes is a reasonable starting point. So is broad match with value based bidding, well structured pages and a clean search terms routine.
A Shopping campaign is a separate decision, not a precondition for broad match in Search.
A physiotherapy clinic has to decide what to do about symptom research. "Physio near me" and "sports physio richmond" are service searches that suit exact and phrase. "Lower back pain when sitting" is research. But research can precede a booking. So it is neither banned nor assumed to convert.
If the clinic tests symptom research, it does so deliberately. The test gets its own ad group, its own page and its own budget. A poor result is contained and a good one is visible. Our guide to high intent vs low intent keywords sorts searches like these into five intent groups. None of the four businesses needs a separate ad group for every service and every suburb.
The tree asks about the business, not about a conversion count. Google's Target CPA page recommends measuring performance over the last 30 days including at least 30 conversions. That is evaluation guidance for the bid strategy, not an entry requirement for broad match.
Our rule of thumb is 15 to 30 conversions a month before trusting a target. Our guide to Smart Bidding strategies explains this. Google itself sets no minimum for Target CPA.
Where all three checks pass, broad match with Smart Bidding is a legitimate first choice. The first is measurement that reflects a business outcome and counts each lead once. The second is an offer and a team that can absorb more varied demand. The third is a spend limit with a review date.
Where they do not pass, the focused start is the safer call. Our Google Ads minimum budget guide covers how narrow that start may have to be.
How Google selects overlapping keywords
Several keywords in one account can match the same search. That does not make them rivals in the auction.
Google's prioritisation page says keywords targeting the same domain do not compete with each other in the auction. Only one of them triggers the ad for a search. The page lays out the order in which that candidate is chosen.
Read it as an order of tie breaks, not five steps every search goes through. An exact match keyword identical to the search term takes priority over any phrase or broad keyword. It also takes priority over a Performance Max campaign. Next come identical phrase and broad keywords, including AI Max, and identical Performance Max search themes.
After that, an AI based relevance assessment takes over. It weighs the search's meaning and context against the ad group's keywords and landing pages. It keeps only the most relevant ad and asset groups. Ad Rank then decides between candidates of equal priority. "Identical" includes spell corrected search terms and does not include plurals or synonyms.
Two compact cases show the difference. First, an account holds [emergency plumber brunswick] as exact, "emergency plumber" as phrase and emergency plumber as broad. Someone searches "emergency plumber brunswick". The exact keyword is identical and, if it is eligible, it serves.
Second, someone searches "urgent plumber brunswick". With no identical keyword, the identical steps find nothing. All three keywords can still match it, the exact one as a close variant. So the choice is made on relevance and then Ad Rank.
The exact keyword does not automatically win because it is exact. That's why the ad group's other keywords and its page matter. Our guide to Google Ads account structure covers the structure that keeps relevance pointing where you want it.
The exceptions on the same page are where to look when a keyword does not fire. A keyword with low search volume status is temporarily inactive and will not trigger ads. An ad group whose creatives or landing pages are all disapproved cannot serve.
Campaign or ad group targeting can rule a candidate out, including audience and location targeting and IP exclusions. A budget restricted campaign means the keyword will not always serve even when it otherwise could.
The preference rules do not apply to Shopping. Google lists its advanced search experiences as Lens, AI Mode, AI Overviews and auto complete searches. In those, keywords may not be prioritised automatically. Relevance to the person's overall intent decides instead.
When a keyword you expected to serve did not, ask the questions in the order of those exceptions.
Routing is the last explanation to reach for, not the first. The search terms report with the Keyword column turned on shows which keyword actually served.
Read match types in the search terms report
Google's search terms report page describes the report's match type column. The column shows how closely the search terms that triggered your ads relate to your keywords. A search term's match type might differ from the one selected for the keyword that triggered the ad.
That one sentence is the key to reading the report.
The search roof repair shows Exact match, so it looks as if an exact keyword served.
The Keyword column shows roof repair with no quotes or brackets. The keyword is still broad, and Exact match describes only this search.
The roof repair row is an example. The report's values are:
Exact match
Exact match (close variant)
Phrase match
Phrase match (close variant)
Broad match
Each describes the relationship between one search and the keyword that matched it.
The Keyword column is hidden by default. Turn it on through the columns icon under Attributes. Read the configured setting, the matched keyword, the search term and the reported relationship together. The report sits under Campaigns, then Insights and reports, then Search terms. Searches matched to Dynamic Search Ads or Shopping targets show Exact with no keyword.
Four example rows from a plumbing campaign show how that turns into decisions. The Keyword column gives each keyword's own setting. Match type is the report's relationship label.
A service in the area served. Judge it on calls and booked jobs
A job seeker, not a customer, and nothing on offer for them
Price research can lead to a booking, but two months brought no calls
Brand demand, kept in its own ad group so it never flatters the service terms
The verdicts depend on the offer and the evidence, not on a banned word list. On the price row, pausing the keyword stops the research clicks. It does so without blocking "cost" from the searches buyers make.
Two more facts about the report change how it should be read. Some search terms without enough query activity are omitted to meet Google's privacy standards. So the visible rows are not the full denominator for any claim about account wide waste.
Search partner strings can also look odd. An odd string is data to investigate, not a new topic to target. A relevant search that has not converted may be immature or simply have too little evidence behind it.
Not every irrelevant looking row needs to become a negative. No fixed share or count of negatives makes an account healthy. Look at the focused local service campaigns we have launched. In their first month, we added 15% to 30% of visible search terms as negatives, by count. Which terms they were mattered more than the share.
Exclusions do not mirror positive matching. Google's negative keywords page says negatives do not match close variants. So synonyms and singular or plural forms need their own entries. But casing and misspellings are handled automatically.
A loose negative can block customers. A broad "free" added for course searches would also block "free quote plumber brunswick". Our guide to Google Ads negative keywords covers the three negative match types, scope and the cleanup routine.
Reach the keyword list does not show
The report can also show reach that none of the three settings explains. Google's AI Max for Search page describes search term matching. It uses broad match, asset based and landing page based technology to extend a campaign's reach.
Google's AI Max setup page says new Search campaigns have AI Max turned on by default. Search term matching turns on with AI Max. You can toggle search term matching off in each ad group's settings.
In the search terms report, AI Max matches get their own match type, labelled AI Max. A source column shows whether each one came from broad match expansion or from keywordless matching. Keywordless matching works from your assets and landing pages rather than a keyword. The keywords report shows AI Max summary rows. The controls are brand inclusions and exclusions, locations of interest and URL inclusions and exclusions.
Separately, the campaign level broad match setting converts the campaign's existing phrase and exact keywords to broad. It leaves the originals with a Removed status and their history visible. Google's page says that from September 2026 campaigns using that setting are automatically upgraded to AI Max.
That upgrade changes priority too. While the setting is on, Google prioritises those keywords as if they were both exact and broad. After the move to AI Max, broad keywords are not treated as exact in prioritisation. That's why Google's AI Max for Search page suggests adding exact match copies where that priority matters.
So a list of phrase and exact keywords does not prove the reach stops at phrase or exact. Check the campaign settings for AI Max and the broad match setting. Check the ad group setting for search term matching and read the source column.
Open Change history to see who or what made each change. Open the Recommendations page to see which recommendation types are set to apply automatically. An Exact label on one row says nothing about whether expansion is active on the rest.
Test broader reach against qualified outcomes
Broad match earns its keep only when the bidding has something to learn from. The keyword matching options page explains why.
To understand the keyword's intent, broad matching considers the person's recent search activity. It also weighs the landing page and asset content and the ad group's other keywords. Google's page says it is critical to use Smart Bidding with broad match.
Which bid strategy to choose is a separate decision. What matters for match types is the quality of the signal the bidding learns from.
Good data means a conversion representing an appropriate business outcome: a booked appointment, a qualified call, a purchase. It is deduplicated so one person is not counted three times. It is not a high total of low value events. Those teach the system to buy more of the wrong thing.
A fair comparison between match types is harder than it looks, because the two arms rarely start equal. Brand searches come from people who already know the business. So an arm that includes them starts ahead. Different service themes carry different margins and different close rates. A wider geography changes travel time and competition, and with them the cost per click.
Raw leads mature at different speeds too. The arm that produced more enquiries last week may be the one producing fewer customers next month. An account wide cost per lead leaderboard by match type carries all of that selection bias. Because of that, it proves nothing about matching.
The fair version is a bounded experiment:
one service theme
the same offer, destination and geography in both arms
brand demand separated out
outcomes compared after they have had time to mature
a hypothesis written down before the test starts
Use a campaign experiment where the volume allows. A low volume business may have to run a carefully logged sequential test instead. Seasonality and a competitor's new budget limit what a sequential test can prove about cause. Our guide to the first 30 days of Google Ads covers how long to wait before reading anything.
Here is a comparison that shows why the denominator matters more than the match type.
The focused arm produced 12 raw leads, six of them qualified. That is $100 per raw lead and $200 per qualified lead. The broader arm produced 20 raw leads, four of them qualified. That is $60 per raw lead and $300 per qualified lead.
Read on raw leads, broader matching won by 40%. Read on qualified leads, it lost by 50%. The example teaches which number to divide by, not that broad match performs worse.
A real test sets the qualified definition, the maturity cutoff and the spend before the first click. The result is then read against them.
Service theme and hypothesis: [theme and what broader reach should change]
Current settings and proposed change: [match types, AI Max status and the edit]
Qualified outcome and maturity cutoff: [definition and days allowed]
Affordable spend and review date: [test allowance and date]
Brand and targeting controls: [brand separated, geography and schedule held]
Observed result and limitations: [qualified CPL by arm and what could confound it]
Continue, revert or investigate: [the decision and the evidence behind it]
Decision owner and next check: [person and date]
Before any match type edit, export the keyword and search term reports as a baseline. Then write the change into the record. Google's keyword editing page says editing a keyword or changing its match type removes the original keyword. A new one is created in its place.
The original's statistics can be viewed when the filters are adjusted to include removed keywords. The edited keyword only appears in reports once it has statistics in the selected period.
A new row with no observations does not erase the account's history. It does not establish a platform wide learning reset.
Define the three outcomes before the test starts. That way, the result decides the next step.
None of those is "expand in week four because the calendar says so".
Check match settings and common mistakes
A ten minute review covers the settings that change what your keywords reach. Work through it with the keyword list, the campaign settings and the change history open.
Leaving new keywords on the broad default
Set the match type as you add each keyword, then check the match type column
Plus sign keywords still in the account
They run as phrase match. Rebuild them as phrase or exact with the current syntax
Inheriting AI Max or the campaign broad match setting
Decide deliberately. Check the setting, the ad group search term matching toggle and the source column
Treating a suburb in the keyword as targeting
Location words make a keyword eligible for searches with that meaning. Location settings decide where ads show
Blaming matching when an exact keyword gets nothing
Check low search volume status, disapprovals, targeting, budget limits and whether another keyword is serving
The same keyword in several ad groups
Relevance decides the ad group for close variants. Keep one home per theme with its own page
Judging match types on raw CPL
Compare qualified, matured outcomes by arm, with brand separated and a recorded baseline
Two of those deserve a sentence more. A keyword is not a targeting guarantee and it does not force an ad to show.
A keyword containing "brunswick" is eligible for searches with that meaning from anywhere the campaign's location settings allow. Eligibility is still not the same as winning the auction. Winning it depends on Ad Rank, budget and the rest of the list. Set where ads can show with Google Ads location targeting. Treat a suburb in brackets as a word, not a fence.
The second is the automatic change. Our audit sample is 180 Google Ads account audits from August 2021 to September 2026. Some routes widen traffic beyond the written keyword list. In that sample, about 4 in 5 accounts had at least one of them on. Those routes range from broad match keywords to Search Partners.
In the same audits, about 1 in 3 accounts had an owner in the dark. Asked at audit, those owners didn't know one of three things was active. They were AI Max search term matching, the campaign level broad match setting and an auto applied recommendation. About 1 in 4 accounts had auto apply switched on without the owner knowing.
If someone else runs the account, monitoring a broad test is part of the job. So is reporting on it. Our Google Ads account management page shows how we handle both.
We'll check your match types, AI Max and auto apply settings. Then we'll show you which searches are widening your reach.
What we recommend at Elev8d
For a local service business on a tight budget, we start with exact and phrase match. They go on the core service terms, with brand searches kept apart. We decide AI Max search term matching and the campaign broad match setting before launch, not by default. In a new account, we switch search term matching off. It stays off until there is a reliable conversion signal and someone reading the search terms weekly.
We widen reach once the tree's three checks pass. The first test is one service theme, judged on qualified leads once they mature.
We'd go broad sooner for an online store with a wide catalogue and value tracking that works. There, an exact list never finishes. We'd hold back longer where the offer is narrow or the diary is already full.
When a campaign moves to AI Max, we treat it as a new test with a fresh baseline. We also add exact copies of the terms whose priority matters. Our guide to Google Ads settings that waste budget covers the other defaults worth deciding before launch.
FAQ: questions before changing your match types
Should I add the same keyword in all three match types?
Not for coverage, because the broader types already capture the narrower queries. The one reason to keep an exact copy of an important term is control.
An identical exact keyword takes priority for that search. It also gets its own row in the keyword report. Add the copy when you want to see and steer that search on its own. A template saying three of everything is no reason to add it.
Is phrase match still needed if broad match works with Smart Bidding?
Phrase match keeps a middle setting. It includes the meaning of the keyword without the full reach of broad. It is still the simplest way to hold a service theme together when conversion data is thin.
A phrase and exact stage can also build the conversion history that broad match bidding later learns from. Phrase match is not obsolete.
Can I stop a close variant I do not want?
Not by switching close variants off, because Google offers no opt out. Add the unwanted search as an exact match negative in that ad group or campaign instead. Negatives do not match close variants. So the negative blocks that search without touching the keyword's other matches.
Next steps
Read the report before touching the settings
Turn on the Keyword column and read a month of search terms. Keep the configured setting beside the reported relationship. Then check AI Max, the broad match setting and automatic changes. That's where reach changes without anyone choosing it.
Write the test before you run it
If broader reach is on the table, write the hypothesis and the qualified outcome into the template. Add the maturity cutoff, the spend limit and the review date. Separate brand demand and export the baseline. Then change one thing.
Decide who owns the weekly read
Someone reads the search terms, the source column and the qualified outcomes every week and records the decision. If that is nobody in the business, that is the first gap to fix. We are happy to talk about doing it for you.
