How broad, phrase, and exact really work in 2026, how each one feeds Smart Bidding, and the decision rule that covers 80% of accounts we audit.
Google Ads has three keyword match types in 2026: broad, phrase, and exact. Broad shows your ad on any related query and is the modern default with Smart Bidding. Phrase shows your ad on queries containing the meaning of your keyword. Exact shows it on same-meaning queries plus close variants.
Most "broad vs exact" guides on the internet still read like 2019. They miss the two things that decide whether match types work for you today: how each one feeds Smart Bidding, and whether you actually run a 14-day negative-keyword review. This article is the explanation of the system. If you came here because your CPC is high, that is a downstream symptom; we wrote the diagnostic version separately at Google Ads cost per click too high.
Triage rule of thumb. Under 10 conversions per month: start exact. 10 to 30 per month: phrase with Maximize Conversions. 30 or more per month: broad with Target CPA or Target ROAS, with disciplined negatives. The detail is in the rest of the article, but that rule alone covers 80% of accounts we audit.
Three changes broke the way most guides talk about match types.
September 2019: Google expanded close variants on phrase and exact match to include same-meaning queries. Before that, "exact" meant the keyword or a misspelling, plural, or reordering. After 2019, "exact" includes paraphrases ("lawn mowing service" matching "grass cutting service") the system reads as identical intent. The official notes live in the close variants documentation.
February 2021: broad match modifier (BMM, the +keyword syntax) was deprecated. Phrase match absorbed BMM behaviour. Most account audits we see in 2026 still reference BMM as if it exists. It does not.
2022: Google publicly repositioned broad match as the recommended pairing with Smart Bidding. The pitch is that the algorithm needs query diversity to find converting intent humans miss. Partly true. Without Smart Bidding and tight negatives, broad match is what it always was: a budget incinerator.
In 2026, match type is a signal-tightness choice, not a literal string-match instruction. You tell Google how wide a net to cast. Google decides what falls into the net.
Broad match shows your ad on any search Google considers related to your keyword. "Related" is doing heavy lifting. The system looks at semantic similarity, user intent, the searcher's recent activity, and other campaign signals before deciding to show. The official definition lives in the keyword matching options documentation.
Sample search-term width. The keyword "running shoes" on broad match can trigger "marathon footwear," "trail sneakers for women," "best brands for plantar fasciitis," and (with Smart Bidding looser) "Nike sale 2026." Some convert beautifully. Some burn $80 of spend in a day.
Why it works in 2026. Smart Bidding uses the broader query stream as training data. With enough conversion volume, the algorithm finds buyers in queries you would never have added as exact keywords. We have seen broad outperform exact on lift in well-instrumented accounts with 30 plus monthly conversions.
Why it fails. Without that conversion volume AND a robust negative list, broad match spends budget on intent that does not match the offer. Google's algorithm optimizes for your bidding target. If the target is loose or the conversion tracking is broken, broad match will faithfully spend your money chasing the wrong signal.
One concrete. On a Goodevas It client in Sprint 5, we moved 4 keywords from phrase to broad with Target ROAS. Result over 30 days: 18% more conversions at the same CPA. It worked because we added 47 negatives before flipping the switch.
Phrase match shows your ad on searches that contain the meaning of your keyword in the order you specified. Since February 2021 it has absorbed what BMM used to do.
Sample search-term width. The keyword "running shoes for women" on phrase triggers "best running shoes for women," "women's marathon shoes," "running shoes women 2026," but not "trail sneakers" or "marathon footwear." Order plus meaning together cut out the wildest broad-match drift.
Why it works. Tighter intent signal than broad, broader signal than exact. Pairs well with Maximize Conversions for accounts in the 10 to 30 conversions per month range. Wide enough for the algorithm to learn, narrow enough to keep waste under control.
Why it fails. It still requires negatives. Less predictable than exact, less expansive than broad. People reach for phrase as a comfort default when they cannot decide. Comfort is not a strategy.
One concrete. We ran phrase match on "plumber near me" for an HVAC client. Over 90 days, 312 unique queries appeared; 87% had clear local intent. The other 13% (mostly "what does a plumber do") became negatives in week 3.
Exact match shows your ad on searches with the same meaning as your keyword, including close variants like plurals, misspellings, reorderings, and (since 2019) paraphrases. The brackets in your account still say [exact], but the matching is no longer literal.
Sample search-term width. The keyword [blue widget] triggers "blue widget," "blue widgets," "bluewidget," and depending on Google's same-meaning model, "azure widget" or "navy widget" if read as identical intent. Queries you did not bid on appear in your search terms report under your exact-match keyword. This is by design.
Why it works. Highest control, lowest waste. The right choice for proven converters with 30 plus days of data and a healthy conversion rate.
Why it fails. Volume is 5 to 10 times lower than broad on the same keyword. The query stream is too narrow for Smart Bidding to learn much, and bids get competitive because everyone else wants exact match on the same proven query.
One concrete. Across our audit set, exact-match conversion rate runs 1.5x to 2x phrase and 2x to 3x broad. Volume runs the inverse. You buy precision with reach.
Match type is signal width. Smart Bidding is what does the deciding. The pair matters more than either part alone.
Broad plus Manual CPC is a budget bonfire. You gave Google maximum query width and offered nothing for the algorithm to optimize against. Your bid is a fixed number, not a value target. Spend goes everywhere, value goes nowhere.
Broad plus Target CPA or Target ROAS is the 2022-and-after default. The wider query stream gives Smart Bidding more conversion data to learn from. The target tells it which queries are worth winning. Google recommends this pair for accounts with 30 or more monthly conversions; production data agrees.
Exact plus Target CPA works, but the narrow query stream gives Smart Bidding thin training data per keyword. Use it when control matters more than scale.
Phrase plus Maximize Conversions is a sensible early-stage default for accounts in the 10 to 30 conversion per month band. We start most new client accounts here, then expand to broad as conversion volume crosses 30.
| Match type | Manual CPC | Maximize Conversions | Target CPA / Target ROAS |
|---|---|---|---|
| Broad | Avoid | Tolerate | Use |
| Phrase | Tolerate | Use | Use |
| Exact | Use | Use | Use |
Decision-matrix view: broad with conversion-targeting strategies, phrase with volume-targeting strategies, exact with control-targeting strategies. The cross-cells exist and we have seen them work in narrow cases. They are exceptions, not defaults.
Every match-type decision is undone if you do not run a negatives review every 14 days. This is the rule. No exceptions.
The process is five steps.
On a typical $5K per month SMB account with broad match enabled, 4 hours of negative-list work in week 1 cuts 18% to 28% of wasted spend by week 3. We see this repeatedly. The work is not glamorous. It is not optional either.
| Negative match type | Blocks | Risk | When to use |
|---|---|---|---|
| Exact negative | One specific query | Low | Surgical blocks on specific bad queries |
| Phrase negative | Queries containing the phrase | Medium | Category words like "free" or "tutorial" |
| Broad negative | Queries containing any of the words | High | Rare; only for clearly off-brand themes |
The right match type depends on three things: monthly conversion volume, your bidding strategy, and whether you will commit to the 14-day negatives cycle.
What if I have multiple campaigns at different stages? Match type is set per ad group, not per account. Calibrate per campaign conversion volume. A new prospecting campaign can be exact while your evergreen converters run broad. Mixing match types across campaigns is fine; mixing them inside the same ad group is not (see the next section).
We see mistake #4 in roughly 1 in 3 SMB accounts we audit. Splitting into separate ad groups by match type lifts Quality Score 0.5 to 1.0 points within 21 days because ad relevance improves when ad groups are themed by intent width, not just keyword theme.
Sage scans the search terms report nightly. Any query with cost above your target CPA and zero conversions for 14 plus days surfaces as a negative-keyword candidate, with predicted savings from the last 30 days of waste.
Mira regenerates ad copy when match type changes. Move from exact to broad and Mira drafts three responsive search ad variants tuned for the wider intent set. Move from broad to exact and Mira tightens copy to the specific keyword theme.
Buzz models the expected CPC distribution after a match-type change weekly, so you can decide whether the predicted variance is acceptable before flipping the switch.
Aegis blocks rash match-type switches. Any change on a keyword with 5 or more conversions in the last 30 days requires operator confirmation with a 24-hour cooling delay. The window catches more bad changes than the rule book.
One concrete from production. Sage flagged 47 negative-keyword candidates from one week of Goodevas It search terms. Operator approved 31. CPA dropped 12% in 21 days. No bid was changed. No match type was changed. Just disciplined negatives, run on the schedule.
Should I still use broad match modifier in 2026? No. BMM was deprecated in February 2021. Phrase match absorbed BMM's behaviour. If your account still has BMM keywords (with the + symbol), Google treats them as phrase match.
What is the difference between phrase match and exact match in Google Ads? Phrase match shows your ad on queries that contain the meaning of your keyword in the order specified. Exact match shows it on same-meaning queries including close variants like plurals, misspellings, and paraphrases. Exact is tighter; phrase has more reach.
Does Smart Bidding work better with broad or exact match? Broad, in most cases, once you have 30 or more monthly conversions feeding the algorithm. The wider query stream gives Smart Bidding more learning data. Below 30 conversions per month, the math reverses and exact or phrase tends to perform better.
How often should I review my search terms report for negative keywords? Every 14 days, minimum. Weekly if you are running broad match on a new account. The work is not glamorous; the savings are 18% to 28% of wasted spend on a typical $5K per month SMB account.
Will Google penalize me for using too many exact-match keywords? No. There is no penalty for match-type composition. The trade-off is volume: more exact match means less impression share, which can starve Smart Bidding of learning data. Use exact when control matters, not as a default.
Can I use all three match types in the same campaign? Yes, in separate ad groups. Same ad group with mixed match types creates an internal auction and is the highest-frequency structural mistake we see. One match type per ad group.
Three things to take away. Match type is signal width, not a literal instruction. Bid strategy is what actually does the deciding. Negatives are the seatbelt that lets you ride either one safely.
If you want Sage to read your search terms report and propose the negatives this week, the free B6 audit runs in 10 minutes and surfaces the top 20 candidates with predicted savings. No card, no rewrite of your account, just the list of negatives most likely to cut waste in the next 30 days. See pricing tiers if you want the nightly cadence after that.