Pull up the search terms report on almost any Google Ads account right now and filter to exact match. Look at what's in there. On a lot of the accounts we manage, you'll find queries that share an idea with the keyword but almost none of its words. Phrase match is doing the same thing, only more of it.

That's not a glitch. It's the product working as designed in 2026. And it means the question most advertisers ask ("which match type should I use?") needs a different starting point. The real question is what each match type is actually buying you now, what it costs, and how much cleanup it creates on the back end.

We covered the bigger picture of lost control in our earlier piece on match types, negatives, and search terms. This one is the practical follow-up: how each match type behaves today compared to a few years ago, when to use each one, what the cost data says, and how to use negative keyword match types to keep the whole thing from leaking money.

How we got here: a short history of match types getting looser

If you started running Google Ads in the last couple of years, you might not realize how literal match types used to be. For a long time, exact match meant the query had to be your keyword. Character for character. Phrase match meant your words had to appear in that order, with stuff allowed before or after. Broad match meant "anything vaguely related," which is why experienced advertisers avoided it.

Then Google started relaxing the rules, one step at a time.

In 2014, close variants (plurals, misspellings, abbreviations) became mandatory for exact and phrase match. You could no longer opt out. In 2017, exact match started ignoring word order and function words like "in" or "for." In 2018, exact match began matching queries with the same meaning, including paraphrases and implied words. In 2019, phrase match and broad match modifier got the same treatment.

In 2021, Google folded broad match modifier into phrase match and retired BMM entirely. (We wrote about that change when it happened.) Phrase match stopped strictly caring about word order when the meaning held up.

Then came AI Max for Search. It launched in beta in May 2025 and went to general availability in April 2026. Starting September 1, 2026, Google began automatically converting campaigns that used the campaign-level broad match setting or automatically created assets into AI Max. Dynamic Search Ads got a reprieve until February 2027.

Two other 2026 changes matter here. In May, Google updated its help documentation to say that for searches coming through Lens, AI Mode, AI Overviews, and autocomplete, the search term you see in your report "represents the best approximation of the user's intent." In plain English: some of the search terms in your report are Google's interpretation, not what the person typed. And in September, Google confirmed a small test letting exact and phrase match keywords serve ads inside AI Mode for queries with "explicit and direct user intent."

If you've noticed looser matching on phrase and exact this year, you're right. It's not your imagination, and it's measurable. Mike Ryan at Smarter Ecommerce analyzed 383 million impressions in AI Max-enabled ecommerce campaigns and found that by July 2026, only about 71% of the impressions attributed to exact match keywords were genuine exact matches. In early 2025, that number was close to 100%.

What each match type actually matches in 2026

Here's how to think about the three match types today. We'll use a home services example since a lot of our clients are Metro Detroit service businesses.

Exact match: [water heater repair]

Exact match now targets the meaning of your keyword. Google will show your ad for the keyword itself, plus close variants, reordered words, and searches it decides mean the same thing.

Searches that will likely trigger it: "water heater repair," "water heaters repair," "repair water heater," "water heater fix," "hot water heater repair."

Searches that might trigger it, depending on the account: "water heater not working," "no hot water," "water heater leaking repair."

That second group is where people get surprised. "No hot water" has the same intent as water heater repair for most people. It could also be a renter looking for a landlord complaint template. Google is making that call for you.

Exact match is still your tightest option. It's just not a lock anymore. Think of it as a strong preference.

Phrase match: "water heater repair"

Phrase match will show your ad for searches that include the meaning of your keyword, with extra words around it. Word order matters less than it used to as long as the meaning stays intact.

Likely matches: "emergency water heater repair near me," "water heater repair cost," "best water heater repair in Royal Oak," "how much to repair a water heater."

Possible matches: "tankless water heater service," "water heater replacement," "who fixes water heaters."

Phrase is the match type where we've noticed the most drift in client accounts this year. "Water heater replacement" is a different job with a different ticket size. Sometimes that's a welcome match. Sometimes your client doesn't do installs and every one of those clicks is wasted.

Broad match: water heater repair

Broad match uses your keyword as a starting point and then pulls in context: the other keywords in the ad group, your landing page content, the searcher's recent search activity, location, and whatever conversion data your bidding strategy has learned from.

Likely matches: everything above, plus "plumber near me," "water heater brands," "Rheem tankless reviews," "why does my water smell like eggs."

Some of that is good discovery. "Plumber near me" might be one of your best converting queries. "Water heater brands" is someone researching a purchase who may never call a repair company. Broad match doesn't distinguish between those on its own. Your bidding strategy and your negatives do that work.

The cost question: is broad cheaper, or does it just spend faster?

This is where a lot of the conversation online goes sideways. People compare cost per click and conclude broad match is cheaper. Sometimes it is, per click. But clicks aren't the goal.

The best recent data we've seen comes from Optmyzr, which analyzed 30,000 Google Ads accounts using February 2026 data. A few findings stood out.

Exact match has lost nearly 10 percentage points of spend share since 2022, and broad match is now the largest match type by budget. Advertisers have shifted money toward broad, but the performance hasn't followed at the same rate. Across non-brand campaigns, exact match led on efficiency metrics. Broad match drove the most volume, but at a higher cost per conversion.

The phrase match finding is the one we'd point most businesses to. In lead gen accounts (those without revenue tracking), phrase match held the largest share of both spend and conversions, and broad match lost more ground there than in ecommerce. Optmyzr's explanation matches what we see: lead gen campaigns live on intent-adjacent keywords, and phrase reaches beyond the exact term without opening the floodgates.

In ecommerce, exact match led on ROAS and CPA, phrase posted the highest conversion rate in the whole study, and broad and phrase ended up nearly tied on ROAS. On branded campaigns, exact match won on almost every metric.

Here's how that plays out in real accounts. Exact match keywords often carry a higher cost per click, because they tend to be your most competitive, most obvious commercial terms and everyone is bidding on them. But the conversion rate is usually high enough that cost per lead comes in lower. Broad match often shows a lower average CPC because a lot of its traffic is on cheaper, less competitive queries. Some of that traffic converts. A lot of it doesn't. The CPA ends up higher unless the bidding strategy is well fed and the negatives are well maintained.

There's also a hidden cost that never shows up in the CPA column: management time. A broad match campaign that "works" may need two or three times the search term review that a tightly built exact and phrase campaign needs. For a small business paying an agency or doing it in-house, that time is real money.

When to use each match type

There's no single right answer, but there are clear patterns. Here's how we decide.

Use exact match when

The keyword is your brand name. Branded traffic is cheap and high intent, and the Optmyzr data shows exact winning almost every branded metric. Broad or phrase on your brand name tends to pull in lookalike searches and competitor-adjacent terms. (If you want to go deeper on branded spend, we wrote about whether your brand campaign is buying customers you already had.)

The keyword is one of your proven money terms, with a track record of converting at a cost you're happy with. You want those protected and funded, not competing with exploratory traffic for budget.

Your budget is tight. If you're spending $1,500 a month, you don't have room for discovery. Every dollar should go to the searches most likely to turn into a call.

You're running lead gen with low conversion volume. Smart Bidding needs data. Without it, looser match types have nothing to steer by.

Example: A family law firm with a $3,000 monthly budget runs exact match on [divorce attorney macomb county], [child custody lawyer sterling heights], and its own firm name. Those are the searches that turn into consultations.

Use phrase match when

You want reach beyond your exact list but still need guardrails. For most of the service businesses we work with, phrase is the day-to-day workhorse.

Your keywords have a lot of useful modifiers. "Roof repair" pulls in "roof repair near me," "emergency roof repair," and "roof repair after storm," all of which you'd want.

You're building out a new campaign and want search term data to find what converts before committing to exact on the winners.

Example: An HVAC company runs phrase match on "furnace repair," "ac repair," and "heating and cooling service," then reviews search terms weekly. Converting variations like "furnace not igniting" get promoted to their own exact match keywords. Junk like "furnace repair jobs" gets added as negatives.

Use broad match when

You have enough conversion volume for Smart Bidding to learn from. That's the most important condition and we'll come back to it.

You've already captured the obvious searches with exact and phrase, and you're looking for incremental volume you can't find on your own.

Your conversion tracking is accurate and reflects real business value. If your conversion action counts every form fill including spam, broad match will learn to find more spam.

Example: An ecommerce brand with 300+ purchases a month and accurate revenue tracking runs broad match on its core product categories under a target ROAS strategy, with exact and phrase campaigns protecting brand and top sellers.

Broad match best practices (if you're going to run it)

We're not anti-broad. We run it in accounts where it makes sense, and it can find converting searches nobody would have thought to add. But it only works when certain conditions are met. Here's our checklist.

Have the conversion volume. Our rule of thumb is at least 30 conversions in the last 30 days at the campaign level, and more is better. Below that, the bidding algorithm is guessing, and broad match gives it a lot of room to guess wrong. Lead gen accounts with 10 leads a month should not be running broad match as their main strategy.

Pair it with Smart Bidding. Broad match with manual CPC is the worst of both worlds: loose matching and no algorithm deciding which of those loose matches are worth bidding on. Use Maximize Conversions, Target CPA, or Target ROAS.

Fix your conversion tracking first. Broad match learns from whatever you tell it is a conversion. If you're counting page views, newsletter signups, or duplicate form fills as conversions, broad match will chase those. If you're running lead gen, consider importing offline conversions so the algorithm knows which leads became customers. Our 30-minute conversion tracking audit is a good place to start.

Review search terms weekly, especially early. For the first 60 to 90 days of a broad match campaign, we review search terms every week. After it settles, every two weeks is usually enough. Remember that some of those terms may be Google's approximations, so you're looking for themes, not just individual strings.

Build negative keyword lists before you launch. Don't wait for the waste to show up. Every service business has predictable junk: jobs, careers, salary, free, DIY, how to, training, course, used, and so on. Add those as shared lists on day one.

Don't let it starve your best keywords. If broad match lives in the same campaign as your proven exact match terms, it competes with them for the same budget. Google's own documentation warns that AI Max won't be effective in budget-limited campaigns. We usually keep broad in its own campaign, or run it as an experiment against the existing setup so we can measure whether it's actually adding anything.

Check what happened on September 1. If any of your campaigns used the campaign-level broad match setting or automatically created assets, they may have been converted to AI Max. Look at your campaign settings and your search terms. A lot of advertisers haven't noticed yet.

Know when to pull it. If CPA has been 30% or more above target for three or four weeks after the learning period, and the search terms are full of irrelevant queries even after negatives, broad match isn't working for that campaign. Go back to phrase and exact and revisit when you have more data.

Negative keyword match types: how they actually work

Negative keywords are the other half of match type strategy, and most advertisers misunderstand them in one specific way: negative keywords do not match close variants. Google's documentation is direct about this. If you add "flowers" as a negative, your ad won't show for "red flowers," but it can still show for "red flower."

So while positive keywords have gotten looser every year, negatives still follow mostly literal rules. Michael Hulsmann of SearchSavior summed it up this spring: matching keeps getting looser while negatives still use strict rules from a decade ago. That mismatch is why so many accounts leak.

There is one helpful exception. Google says negatives automatically account for casing and misspellings. Plurals and synonyms, though, you have to add yourself.

Here's how each negative match type behaves, using "free estimate" as a negative keyword.

Negative broad match: free estimate

This is the default. Your ad won't show if the search contains all of the negative terms, in any order. It can still show if the search contains only some of them.

Blocked: "free estimate roof repair," "roof repair estimate free"
Not blocked: "roof repair estimate," "free roof inspection"

Negative broad is not as broad as it sounds. It's really "contains all these words." That's useful for blocking a concept regardless of word order.

Negative phrase match: "free estimate"

Your ad won't show if the search contains the exact words in the same order, with anything before or after.

Blocked: "free estimate roof repair," "get a free estimate"
Not blocked: "roof repair estimate free," "free estimates" (plural is a different word)

Phrase negatives are what we use most. They block a specific modifier wherever it appears in a query.

Negative exact match: [free estimate]

Your ad won't show only if the search is exactly those words in that order, nothing more.

Blocked: "free estimate"
Not blocked: "free estimate roof repair," "free estimates"

Negative exact is a scalpel. Use it to block one specific query without accidentally blocking longer, valuable searches that contain it. For example, a roofer might add [roofing] as a negative exact to avoid paying for a one-word research query, while still showing up for "roofing contractor near me."

One technical detail worth knowing: Google only checks the first 16 words of a query against your negatives. If your negative appears as the 17th word or later, the ad can still show. With AI Mode queries running about three times longer than traditional searches, that limit is starting to matter.

Where negatives live and how to use each level

Negatives can sit at several levels, and using the right level saves a lot of maintenance.

Ad group level. Use this to steer traffic between ad groups in the same campaign. If one ad group targets "water heater repair" and another targets "water heater installation," you can add each as a negative in the other to keep queries going to the right ad and landing page.

Campaign level. Use this for terms that don't fit a specific campaign but might be fine elsewhere in the account. A commercial roofing campaign might negate "residential" and "house."

Shared negative keyword lists. These are lists you build once and apply to many campaigns. We keep standard lists for employment terms, DIY and education terms, free and cheap terms, and competitor names (when we don't want to bid on them).

Account level. Account-level negatives apply across all Search and Shopping inventory in Search, Performance Max, Shopping, and several other campaign types. Google caps this at 1,000 negatives per account, so save it for terms you never want to show for anywhere: things like "jobs," "careers," "lawsuit," or a product you don't sell.

Performance Max. PMax campaigns now accept campaign-level negative keywords, with a limit of 10,000 since March 2025. If you're running PMax alongside Search, this is one of the most useful controls you have. We covered the broader PMax picture in our Performance Max for lead generation guide.

How negatives improve campaign performance

Good negative keyword management does three things.

It cuts wasted spend directly. Every irrelevant click you block is money that goes to a relevant one instead. In our experience auditing Metro Detroit service business accounts, 15% to 25% of spend is commonly going to searches that were never going to convert, and negatives are the fastest way to recover most of it.

It gives Smart Bidding cleaner data. When the algorithm stops spending on junk queries, its learning is based on traffic that actually has a chance to convert. That tends to improve conversion rate and CPA over time, not just immediately.

It improves ad relevance. When a query reaches the ad group built for it, the ad copy and landing page match what the person searched. That helps Quality Score and click-through rate, which lowers what you pay per click.

Here's the process we follow on most accounts.

1. Review the search terms report weekly for new campaigns, every two weeks for mature ones.
2. Sort by cost, not by impressions. The expensive junk matters more than the frequent junk.
3. Look for patterns, not just individual queries. If you see "plumber salary," "plumber apprenticeship," and "plumbing school," the theme is employment and education. Add phrase negatives for "salary," "apprentice," "school," and "jobs."
4. Use negative exact for single bad queries you don't want to over-block.
5. Use negative phrase for bad modifiers.
6. Be careful with negative broad on single words. Adding "free" as a negative broad blocks "free estimate," which might be your best converting search.
7. Add plurals and common variations by hand, since negatives don't cover close variants.
8. Check your negatives against your keywords every few months. Conflicting negatives (a negative that blocks one of your own keywords) are one of the most common issues we find in audits.

The bottom line

Match types still matter in 2026. They're just looser guidelines than they used to be. Exact match is the strongest signal you can send about what you want, even if Google interprets it more freely than the brackets suggest. Phrase match is the workhorse for most lead gen and local service accounts. Broad match can find real incremental volume, but only when you have the conversion data, the tracking, and the time to manage it.

Negatives are what hold it all together. As positive matching keeps loosening, the advertisers who maintain disciplined negative keyword lists are the ones who keep their cost per lead under control.

If you're not sure how much of your budget is going to searches that will never convert, that's exactly what our free Wasted Clicks Audit is built to find. Or if you'd rather talk through your account structure, schedule a consultation with our Google Ads management team.