Google Ads Keyword Match Types: Broad, Phrase, Exact and 2026 AI-Powered Strategy

by Francis Rozange | Apr 4, 2026 | Google Ads

types-correspondance-mots-cles-google-ads-1

Category: Google Ads | Reading time: 24 minutes | Last updated: April 2026

The keyword match type debate looks settled from the outside. Google’s product marketing tells a clean story: broad match plus Smart Bidding wins, exact match is for brand defense, phrase match is the safe middle. The reality on the ground is messier. Match types in 2026 do not work the way they did in 2020, the way they did in 2022, or the way the Google Ads Help articles describe in the first paragraph. The exact-match keyword you typed five years ago triggers on a different set of queries today. The phrase-match keyword you set up last quarter has absorbed everything that used to be modified broad match. The broad-match keyword you avoided in 2018 is now what the auction rewards, but only if you have the conversion infrastructure to back it up.

This guide unpacks what each match type actually does in 2026, how Smart Bidding rewires the question, where manual CPC still has a role, and how to read the search terms report to keep the system honest. Sources are linked inline: Google Ads Help for the canonical behavior, WordStream for industry benchmarks, Search Engine Land and Search Engine Journal for the platform changes that did not get press releases, Tinuiti and Optmyzr for the agency-side performance data, Ahrefs for the SEO-adjacent keyword research view.

Why match types still matter in 2026 despite Smart Bidding

The argument that “match types do not matter anymore” comes up in every Google Ads forum. The argument is wrong, but it contains a partial truth. What is true: Smart Bidding overrides match-type literalism in ways that did not exist five years ago. The algorithm decides whether a query converts based on signals you do not control, not on whether the search string matches your keyword character by character. What is false: that this dynamic makes match types interchangeable. Match types still control the universe of queries that even reach the auction for your keyword. They control the query density that Smart Bidding has to work with. They control the noise floor.

The cleanest way to think about it: match types in 2026 are no longer a matching contract, they are a steering signal. Exact match tells the system “narrow the candidate query pool aggressively, I want signal density over reach.” Broad match tells the system “expand the candidate pool, I trust your conversion model to filter.” Phrase match sits in between with its own absorption of what used to be modified broad. The bid algorithm reads those signals and behaves accordingly. Get the signal wrong and you starve a high-intent campaign of volume, or you flood a low-conversion campaign with irrelevant queries the algorithm has not yet learned to filter.

According to WordStream’s 2024 industry benchmark data, the gap between best-in-class and median accounts in the same vertical is roughly 3x on cost-per-acquisition. The single largest contributor to that gap is not creative, not landing pages, not bid strategy, it is the alignment between match types and the conversion data feeding the bidder. A retail account with $30 CPA in the top quartile and a retail account with $90 CPA in the bottom quartile usually differ on this exact axis: the first one runs broad match against a clean conversion stream with 200+ monthly conversions, the second runs broad match against a noisy conversion stream with 20 monthly conversions and no negative-keyword discipline.

The other reason match types still matter: account structure. Different match types in the same campaign create different learning surfaces for the bidder. Different match types across campaigns let you run parallel strategies with different risk tolerances. The post-2024 reality is not that match types do not matter, it is that match types matter differently. They are a structural lever, not a literal-matching rule.

Broad match: the rebuild from 2021 to 2024

Broad match in 2018 was a wasteland. The keyword “running shoes” would trigger on “best Netflix shows,” “running for office,” and “shoes for kids,” because Google’s matching logic was a bag-of-related-terms expansion with no intent filter. Agencies told clients to never use broad match. The advice was correct given the technology of the time.

The rebuild began in 2021 and finished, in practice, in late 2024. Three changes converged. First, Google deprecated modified broad match in February 2021 and folded its behavior into phrase match, which freed the broad-match slot to become something else. Second, the BERT and MUM language models were integrated into the match-type expansion logic, which let the system reason about query intent rather than just lexical similarity. Third, broad match was rewired to take the campaign’s conversion history as a primary input rather than a secondary one, which means a broad-match keyword in a campaign with 500 monthly conversions now behaves very differently from the same keyword in a campaign with 5 monthly conversions.

The audience-signal integration that landed in 2023 was the piece that made broad match competitive with phrase match for most accounts. Broad match now reads your customer match lists, your remarketing audiences, your in-market segments, and your custom intent audiences as inputs to the matching decision. A query that would have triggered your ad in 2020 might not trigger it in 2026 because the user does not match your audience profile. A query that would not have triggered in 2020 might trigger now because the user does match.

The performance pattern that emerged across agency case studies is consistent. Tinuiti documented across multiple 2024 client engagements that broad match plus Smart Bidding plus a clean conversion stream produces 25 to 35% more conversions than phrase match at the same or lower CPA, but only past a volume threshold of roughly 50 monthly conversions per campaign. Below that threshold, broad match underperforms phrase match by 15 to 25% on CPA because the bidder does not have enough signal to filter the expanded query pool effectively.

What broad match is actually matching on

The 2026 broad-match decision tree, reconstructed from Google’s Help documentation and the academic papers Google has published on the underlying models, looks roughly like this. The system starts with the query semantic embedding, computes a similarity score against the keyword embedding, applies a campaign-conversion-prior to bias the score toward query patterns that have converted historically, applies an audience-signal weight if the user matches one of your audience inputs, and applies a landing-page-content weight that pulls toward queries semantically aligned with what the destination page actually offers. The result is a single match-eligibility score, and the threshold for that score is dynamic based on the auction context.

The practical consequence: the broad-match keyword “project management software” in a campaign with 200 monthly conversions and a clean B2B audience signal will trigger on “team workflow tool,” “remote collaboration platform,” and “agile sprint planning software,” because the system has learned those queries convert for businesses that look like yours. The same broad-match keyword in a campaign with 10 monthly conversions and no audience signal will trigger on the same queries plus “trello vs asana,” “free kanban board,” “personal to-do list app,” and assorted noise the system has no signal to filter against.

The reach difference is real. Optmyzr’s 2024 study across 2,637 accounts showed that broad match in mature campaigns reaches roughly 3x the unique queries per dollar of phrase match in the same campaign, and roughly 6x the unique queries per dollar of exact match. The conversion-rate gap closes when conversion infrastructure is solid: in the top quartile of accounts, broad match converts within 10% of exact match on CPA. In the bottom quartile, broad match converts at 60 to 80% higher CPA than exact match.

The three prerequisites for broad match

The prerequisites for broad match in 2026 are non-negotiable in a way they were not in 2020. Skip any of them and broad match becomes the budget sink it was a decade ago.

Conversion tracking that captures value, not just count. Smart Bidding optimizes for the conversion signal it receives. If the signal is “a conversion happened, value unknown,” the bidder treats every conversion as equivalent and chases volume regardless of revenue. If the signal is “a conversion happened, value $87.50,” the bidder learns to bid more aggressively for queries that produce higher-value conversions. Per Google’s published guidance on Maximize Conversion Value bidding, accounts that pass true conversion values to the system outperform accounts that pass flat-value or count-only conversions by 10 to 25% on revenue per dollar of spend. For broad match specifically, the value signal is what filters the noise: without it, the bidder cannot tell that “free project management” and “enterprise project management” are different conversions even when both produce a “lead” event.

Minimum conversion-tracking standards before enabling broad match: every revenue-producing event tracked, conversion latency under 24 hours, conversion values passed for transactions, lead-quality scores passed for B2B (offline conversion imports if the lead-to-customer cycle exceeds 30 days), cross-device tracking enabled, no double-counting between GA4 import and the Google Ads native tag.

Conversion volume above the learning threshold. Google’s published threshold for Smart Bidding to exit learning mode is 30 conversions in 30 days for the campaign. The threshold for broad match to outperform phrase match in agency-published data sits higher, around 50 to 100 conversions per month per campaign. Below that, the bidder is making decisions on too-thin a signal and broad match becomes a gamble.

The volume requirement scales with conversion-action diversity. A campaign with one conversion type (purchase) and 50 monthly conversions has more signal density than a campaign with three conversion types (purchase, lead, signup) and 50 monthly conversions split across the three. The latter behaves as if it has 17 conversions per type, which is below the learning threshold for each.

A negative keyword list that has been built from real search-terms data, not from imagination. The most common broad-match failure pattern is launching with 20 imagined negatives (“free,” “cheap,” “DIY”) and never adding to the list. Real search-terms data reveals categories of waste that nobody imagined: a B2B SaaS account discovers it is showing on dating-app related queries because of a stem-matching collision, a footwear retailer discovers it is showing on costume-shoe queries because Halloween is approaching and the algorithm expanded into adjacent intent. Per Search Engine Journal’s 2024 analysis of broad-match waste, accounts that do weekly search-terms reviews and add 10 to 30 negatives per week reduce broad-match waste from a typical 40 to 50% of spend down to 8 to 12% within 90 days.

When broad match works and when it does not

Broad match works well in mature accounts with rich conversion signal, in verticals where query language varies widely (the same intent can be expressed in 20 different phrasings), in campaigns where the audience signal is strong and well-defined, and in product categories with broad enough demand that signal density is achievable.

Broad match does not work in new accounts with under 30 days of conversion history, in regulated verticals (legal, medical, financial) where query expansion can trigger compliance violations, in campaigns targeting hyper-specific niches where the candidate query pool is too small for the expansion to add value, or in accounts where conversion tracking is broken in ways the team has not yet diagnosed.

Phrase match: what it became after the modified broad merger

Phrase match in 2020 was a narrow tool: “running shoes” matched on queries containing the phrase “running shoes” in that order, with allowed words before and after. Modified broad match was the workhorse for most accounts, with its +keyword +keyword syntax letting advertisers require specific terms without enforcing word order.

The February 2021 deprecation of modified broad match folded its behavior into phrase match, which became the primary tool for query control with order-flexibility. The behavior change is sometimes mis-stated. Phrase match in 2026 is not “the old phrase match plus the old modified broad.” It is closer to “the old modified broad with intent-aware close variants and order tolerance for queries where the meaning is preserved by reordering.”

The keyword “women’s hiking boots” in phrase match triggers on queries that contain the keyword’s core meaning with the same intent. Allowed: “women’s hiking boots on sale,” “best women’s hiking boots,” “waterproof women’s hiking boots,” “hiking boots for women” (reordered, meaning preserved), “women hiking boots” (function word removed). Not allowed: “women’s boots” (core hiking concept missing), “men’s hiking boots” (audience flipped), “hiking shoes for women” (product category drifted). The order-tolerance was the modified broad behavior; the meaning-preservation check was the close-variant logic.

Phrase match’s role in the 2026 stack is the volume driver in mature campaigns and the safe-default in new campaigns. The volume profile sits roughly between exact and broad: 40 to 60% more clicks than equivalent exact match, 50 to 70% fewer clicks than equivalent broad match, with conversion rates that typically land within 10% of exact match if the negative-keyword list is maintained.

Why phrase match is the safe default for accounts under the broad-match threshold

The accounts that should run phrase match as the primary match type are the ones that do not yet meet the broad-match prerequisites. New accounts in their first 90 days. Accounts with under 50 monthly conversions per campaign. Accounts in regulated verticals where query expansion creates compliance risk. Accounts where the team does not yet have the bandwidth for weekly negative-keyword reviews.

Phrase match in these contexts gives 80% of the volume of broad match with 30% of the noise. The conversion signal feeding Smart Bidding is denser per click. The negative-keyword list grows more slowly because there are fewer expansion-driven misses to catch. The team can run the campaign with monthly reviews instead of weekly reviews and the performance does not degrade.

The transition pattern that shows up across well-managed accounts: launch on phrase match plus exact match for the first 90 days, accumulate the conversion data and audience profile, then add broad match in a controlled 10 to 20% budget allocation once the prerequisites are met. The accounts that skip the phrase-match phase and launch directly on broad match in their first 30 days consistently report the same problem: 60 to 80% higher CPA in the first quarter, with a recovery curve that takes 4 to 6 months and a lot of negative-keyword work to flatten.

Phrase match’s blind spots

Phrase match has two structural blind spots that the documentation does not surface. First, the order-tolerance is not symmetric. The keyword “women’s hiking boots” matches “hiking boots for women,” but the keyword “hiking boots for women” does not always match “women’s hiking boots” with the same eligibility score. The reason is that Google’s system treats the first keyword as the canonical form and the matching is biased toward queries that map to that canonical form. The practical workaround for accounts that care about this asymmetry is to add both phrasings as separate phrase-match keywords. The cost is a slight increase in keyword-list management; the benefit is more predictable matching across reordered queries.

Second, the close-variant expansion in phrase match has been getting more aggressive each year, and 2026 is no exception. The keyword “project management software” in phrase match now matches on “project management tools,” “project management platform,” “project management app,” “workflow management software,” and a growing list of synonyms. The expansion is mostly accurate but produces occasional drift. The fix is the same as for broad match: review the search-terms report weekly, add negatives for the categories of expansion that do not align with your offering.

Exact match: what “exact” means in 2026

Exact match is the most-misunderstood match type in 2026 because the word “exact” has not described the behavior since 2017. The first close-variant expansion landed in 2014, the function-word and reordering tolerance landed in 2017, the same-meaning expansion landed in 2018-2019, and the AI-driven intent matching layer landed in 2021-2022. The match type still produces the highest signal density of the three, but it is not exact in the literal sense.

The keyword [women’s hiking boots] in exact match triggers on the literal query, on misspellings, on singular and plural variations, on function-word removals, on word reorderings that preserve meaning, on stems and synonyms that the system has classified as same-intent, and on rephrasings that mean the same thing in a different vocabulary. It does not trigger on queries that change the audience (men’s, kids’), the product category (running shoes, dress boots), or the intent (rental, repair).

The 2024 close-variant expansion that drew the most agency complaints was the same-meaning expansion into category-adjacent terms. The keyword [women’s hiking boots] started triggering on queries like “women’s trekking boots” and “women’s outdoor boots” because the system classified the categories as same-intent. The opt-out for same-meaning close variants was removed in 2019 and has not returned. The only control left is the negative-keyword list.

Exact match’s role in the 2026 stack

Exact match remains the highest-signal-density match type. The conversion rate on exact match keywords sits 30 to 50% above phrase match and 50 to 100% above broad match in most verticals, per Optmyzr’s published benchmark study. The CTR follows a similar pattern: exact match averages 6 to 8% CTR in retail, phrase match averages 4 to 5%, broad match averages 2 to 3%. Quality Score on exact match averages 8 to 10 in well-managed accounts, versus 7 to 9 for phrase and 6 to 8 for broad.

The role exact match plays in 2026 account structure is the precision foundation. It captures the queries with crystal-clear intent at the lowest possible CPA, it produces the conversion data that feeds Smart Bidding’s understanding of what a high-value query looks like, and it serves as the brand-defense layer for searches that should never go to a competitor. Accounts that build the exact-match foundation first and add broad match later consistently outperform accounts that launch broad-first.

The accounts that benefit most from exact-match-heavy strategies are B2B with high customer acquisition costs and long sales cycles, regulated verticals where query expansion creates risk, local services where the geographic intent is part of the query, and brand-defense layers across all verticals. The accounts that benefit least are e-commerce with broad consumer demand and high conversion volume, where exact match alone leaves significant volume on the table.

The exact match performance baseline

The performance baseline that emerges across published benchmark data, per Optmyzr’s 2024 study and WordStream’s 2024 industry benchmarks: exact match keywords in retail average 8 to 10% conversion rate, $18 to $25 CPA, $1.50 to $3.00 CPC, and 6 to 8% CTR. Phrase match averages 5 to 6% conversion rate, $26 to $35 CPA, $1.20 to $2.50 CPC, and 4 to 5% CTR. Broad match averages 3 to 4% conversion rate, $30 to $45 CPA in mature accounts, $0.80 to $2.00 CPC, and 2 to 3% CTR.

The volume-per-dollar pattern reverses. Broad match produces 3 to 6x the unique queries per dollar of spend versus exact match. Phrase match sits in the middle. The right metric for comparing match types is not CPA in isolation, it is CPA per unit of incremental volume the match type adds to the account.

Negative keyword match types and how they interact

Negative keywords are the defensive armor of the match-type system. They work differently from positive match types, in ways that confuse most advertisers. The differences matter because the wrong negative match type either over-blocks (kills queries you wanted) or under-blocks (lets waste through).

Negative exact match. Syntax: -[search term]. Behavior: blocks the literal search and its tight close variants (misspellings, singular/plural, function-word removal). Does not block queries that contain the term plus other words. Example: -[free project management software] blocks the query “free project management software” but does not block “free project management software for nonprofits.” Use case: blocking specific high-volume zero-conversion queries that the search-terms report has surfaced.

Negative phrase match. Syntax: -“phrase phrase”. Behavior: blocks searches that contain the phrase in the specified order, with additional words allowed before or after. Example: -“free project management” blocks “free project management software,” “free project management tool,” and “best free project management app,” but does not block “project management free trial” (different word order). Use case: blocking categories of intent (free, cheap, DIY, tutorial) that the search-terms report shows are not converting.

Negative broad match. Syntax: -word -word -word. Behavior: blocks searches that contain all the listed words in any order. The crucial difference from positive broad match: negative broad match does not expand to synonyms or close variants, it requires the literal word. Example: -free -tutorial blocks “free project management tutorial,” “tutorial for free project management,” “is project management free tutorial,” but does not block “complimentary project management guide” (synonym) or “free project mgmt tutorial” (close variant of the negative).

The asymmetry is the trap. Positive broad match expands aggressively through synonyms; negative broad match does not. This means a negative-broad list of “free” does not block “complimentary,” “no cost,” or “$0.” Building negative-keyword lists for broad-match campaigns requires explicitly listing the synonyms, not relying on the negative match type to expand them.

Recommended negative-keyword structure

The structure that works across well-managed accounts uses three tiers, each with a different role.

Account-level negative lists. 200 to 500 universal negatives that should never trigger ads anywhere in the account. Categories: completely off-target intents (jobs, salary, login, support, tutorial), offensive content, wrong audience segments (kids if you sell to adults, B2C terms if you sell B2B). Apply to every Search campaign through the Shared Library.

Campaign-level negatives. 100 to 300 negatives that block intent categories that should not trigger this campaign but might be valid elsewhere. Categories: competitor names if not running competitor campaigns, low-intent modifiers (free, cheap, DIY), wrong-fit modifiers (used, secondhand, rental for a business that does not offer those).

Ad-group level negatives. 50 to 150 negatives that block queries that should go to a different ad group. Categories: cross-contamination between sibling ad groups (women’s vs men’s, exact-vs-phrase variants), product-line crossover (running shoes negatives in the hiking-boots ad group).

Performance Max changed the negative-keyword math in March 2025 when Google raised the negative-keyword limit from 100 to 10,000 per campaign. The change made Performance Max a controllable surface for the first time. Accounts running Performance Max in 2026 should treat the negative-keyword list as a primary control lever, not a footnote.

Match types plus Smart Bidding: the interaction model

Smart Bidding and match types are not independent levers. They interact in ways that determine whether a campaign produces a clean signal for the bidder or a noisy one. The interaction model has three layers.

Layer one: query density. Smart Bidding needs a sufficient number of conversions per match-type-segment of the campaign to learn the conversion patterns. Exact match in a low-volume campaign produces high signal density on a small query pool. Broad match in a high-volume campaign produces moderate signal density on a large query pool. Both can work for Smart Bidding. The trap is broad match in a low-volume campaign, which produces low signal density on a large query pool, and Smart Bidding cannot learn fast enough to filter the expansion before the budget is spent.

Layer two: conversion-value signal. Smart Bidding’s value-based strategies (Maximize Conversion Value, Target ROAS) require conversion values to differentiate query patterns. Match types affect the value distribution. Exact match queries tend to cluster near a single value pattern (high-intent, full-price). Broad match queries spread across a wider value distribution (mixed intent, mixed value). Without conversion values being passed through, Smart Bidding cannot distinguish a $200 broad-match conversion from a $20 broad-match conversion, and the bidder optimizes toward the volume-weighted average.

Layer three: audience and context signal. Smart Bidding incorporates user signals (audience match, device, time, location, search history) into the bid decision. Match types affect the audience profile. Exact-match queries tend to come from users with high topical specificity. Broad-match queries come from a broader audience profile. The bidder uses both as inputs, but the match-type-by-audience interaction is where most of the optimization gain happens. A broad-match keyword in a campaign with strong audience signals (customer match, in-market, custom intent) outperforms the same keyword in a campaign with no audience signals by 20 to 35% on CPA.

Bid strategies and match-type pairing

The pairings that work, based on agency-published case studies and the patterns that show up across mature accounts:

Maximize Conversions plus phrase or broad match. Default starting strategy for accounts with 30 to 50 conversions per month and value-agnostic conversions. Works well in lead generation where every lead is treated equivalently before sales qualification.

Target CPA plus phrase or exact match. Default for accounts with stable CPA targets and predictable conversion-to-revenue ratios. Works well in subscription B2B and services with consistent customer-acquisition economics.

Maximize Conversion Value plus broad match. Default for mature e-commerce with rich product-feed-driven value passing. Requires Enhanced Conversions and accurate transaction values. Underperforms broadly when the value signal is broken or thin.

Target ROAS plus broad match. Default for mature e-commerce with stable margin structures. Set the ROAS target slightly below historical average to give the bidder room to find new conversions; set it slightly above to preserve margin at the cost of volume.

Manual CPC plus exact match. Underrated in 2026. Works well for new accounts before Smart Bidding has data, for regulated verticals where deterministic bid control is required, for brand-defense campaigns where the bid does not need optimization, and for low-volume B2B campaigns where the conversion data will never reach the Smart Bidding threshold.

Match types plus manual CPC: when the manual lever still wins

Smart Bidding has eaten manual CPC for most use cases. The use cases where manual CPC still wins, based on the accounts where it shows up in well-managed portfolios:

Sub-threshold conversion volume. Smart Bidding requires roughly 15 conversions per month per campaign as a hard minimum and 30 to 50 to behave stably. Below that, the bidder is in permanent learning mode and CPCs swing wildly. Manual CPC plus exact match keeps the campaign in the auction at predictable bids while you accumulate the conversion history.

Brand-defense campaigns. The CPC on branded queries is low (often under $0.50 per click in non-competitive verticals), the conversion rate is high (often 15 to 25%), and the optimization opportunity is limited (the queries are already converting). Manual CPC at a fixed bid produces stable performance with no algorithmic surprises. Smart Bidding on brand campaigns occasionally raises bids unnecessarily because it confuses brand-search with high-value generic-search and tries to “scale” the brand campaign past its natural ceiling.

Regulated verticals. Legal, medical, financial, gambling, alcohol. Smart Bidding’s query-expansion logic occasionally triggers ads on queries that violate vertical-specific advertising policies. Manual CPC plus exact match plus tight keyword lists keeps the auction surface deterministic.

Geographic precision plays. Local services with hyper-local geographic targets where Smart Bidding’s audience expansion can dilute the local signal. Manual CPC plus exact match plus tight geo-targeting produces predictable performance without the algorithm trying to expand into adjacent markets.

Test campaigns. A new ad group, a new landing page, a new audience. Manual CPC during the test period isolates the variable being tested. Smart Bidding during the test introduces algorithmic confounds that make the test results harder to interpret.

The search terms report: the only reliable source of truth

The search terms report is where match-type theory meets match-type reality. The report shows the actual queries that triggered each keyword, the match-type classification Google assigned to each query, the impressions and clicks, and the conversions if conversion tracking is configured correctly. Reading the report is not optional in 2026, it is the primary feedback loop for the match-type strategy.

The 2020 report was richer than the 2026 report. Google removed the queries that did not meet a “significant volume” threshold in September 2020, which means roughly 25 to 30% of the search-terms data is now hidden in the “other search terms” aggregate row. The reduction is permanent and there is no opt-out. The remaining 70 to 75% of the data is still the cleanest source of truth for what is happening in the auction.

The four-step search-terms-report optimization process

Step one: weekly identification of high-converters. Filter the report for queries with conversion rate above 2x the campaign average. These are the patterns the system is finding that you did not anticipate. Add the high-converters as exact-match keywords in the most relevant ad group, which moves them out of the broad-match expansion zone and into a controlled exact-match position.

Step two: identification of high-cost zero-converters. Filter for queries with above-median click counts and zero conversions over a 30-day window. These are the budget thieves. Add as negatives at the appropriate level (ad group, campaign, or account) based on whether the query should ever convert anywhere in the account.

Step three: identification of competitor and comparison queries. Queries containing competitor names or comparison structures (“X vs Y,” “alternative to X”) are high-intent but require deliberate strategy. Either route them to a comparison-focused ad group with appropriate landing pages and ad copy, or add the competitor names as account-level negatives if you are not running competitor campaigns.

Step four: pruning the obvious waste. If more than 15% of the search terms in a 30-day window are clearly off-target, the keyword match types are too broad for the current state of the negative-keyword list. Either tighten the match types (move broad-match keywords to phrase match) or expand the negatives until the off-target rate drops below 10%.

The threshold for healthy match-type alignment, per Search Engine Journal’s 2024 analysis: 85 to 90% of search terms in the report should be recognizable as legitimate intent matches. Below 85%, the match types are misaligned with the keyword strategy. Above 95%, the match types may be too tight and you are leaving volume on the table.

Common mistakes

Mistake one: enabling broad match before the prerequisites are met. The single most expensive match-type mistake in 2026. The pattern: a new account or a low-volume account reads the Google product marketing about broad-match-plus-Smart-Bidding, enables broad match across the campaign, and watches CPA double or triple over the first 30 days. The recovery is slow because the negative-keyword list has to be built from scratch on noisy search-terms data, and Smart Bidding is in extended learning while the conversion signal is rebuilt.

Mistake two: confusing keyword match type with search-term match type. The match-type column in the search-terms report is not the match type of your keyword, it is Google’s classification of how the query matched the keyword. An exact-match keyword [women’s hiking boots] can show in the report with match-type “phrase” or “broad” if Google classified the query as a phrase-type or broad-type match of your keyword. The classification reveals how aggressively Google is expanding your keyword.

Mistake three: launching broad match without a negative-keyword foundation. Per Search Engine Journal’s 2024 analysis, accounts that launch broad match with under 50 negatives in place waste 40 to 50% of broad-match spend in the first month. Accounts that launch with 200+ negatives based on industry-typical waste patterns plus a weekly negatives-review cadence waste 10 to 15% in the first month and converge to 8 to 12% by month three.

Mistake four: ignoring Quality Score by match type. Quality Score declines on broad match are an early warning that the match-type strategy is misaligned. Exact match should hold 8 to 10 Quality Score, phrase match 7 to 9, broad match 6 to 8. Persistent Quality Score below 5 on any match type signals a deeper problem (landing-page mismatch, ad-copy drift, keyword-intent collision) that the match-type lever cannot fix.

Mistake five: testing all match types simultaneously on equal budget. The most common diagnostic mistake. The advertiser launches a campaign with exact, phrase, and broad match for the same keywords, gives each equal budget, and tries to compare performance. The comparison is not valid because the match types compete with each other in the auction (the exact-match keyword wins the auction for queries that match it exactly, leaving the phrase-match keyword to compete for everything else, and the broad-match keyword for the residual). The cleaner approach: run the match types in separate campaigns with deliberate budget allocation, or run a phased rollout that adds match types over time.

Mistake six: treating match-type strategy as a one-time decision. Match-type strategy needs to evolve with the account. A new account in month one runs exact-match-heavy. The same account in month six should be running a balanced exact/phrase/broad mix. The same account in month eighteen, with mature conversion data and audience signals, should be running broad-match-heavy with exact-match brand defense. Accounts that lock the match-type structure at launch and never revisit it leave 20 to 30% of available volume on the table by month twelve.

Decision tree by goal, budget, and account maturity

The decision tree for match-type strategy in 2026 has three branches: account maturity, budget level, and primary goal. The intersection determines the right starting structure.

New account, any budget, any goal

Months one through three. 70 to 80% exact match, 20 to 30% phrase match, 0% broad match. Manual CPC or Maximize Clicks for bidding. The goal of this phase is to accumulate conversion data, build the negative-keyword foundation, and establish Quality Score baselines. Budget allocation matters less than data quality. A $1,500 monthly budget on tight exact match produces better data than a $5,000 monthly budget on loose broad match.

Exit criteria for moving to the next phase: 30 to 50 conversions accumulated, search-terms report showing 85%+ recognizable terms, Quality Score averaging 7 or higher across the keyword list, conversion tracking validated end-to-end with values passing correctly.

Maturing account, low to mid budget, conversion goal

Months three through nine, $1,500 to $10,000 monthly budget. 50 to 60% exact match, 30 to 40% phrase match, 5 to 15% broad match in a controlled test allocation. Smart Bidding (Maximize Conversions or Target CPA) on the campaigns with sufficient volume, manual CPC on the rest.

The broad-match test allocation is the critical lever in this phase. Allocate 10 to 15% of campaign budget to broad-match keywords with tight negative-keyword lists. Review weekly. If broad-match CPA stays within 25% of exact-match CPA over a 30-day window, expand the allocation. If broad-match CPA runs more than 50% above exact-match CPA, tighten the negative-keyword list before expanding.

Mature account, mid to high budget, conversion goal

Month nine onward, $10,000+ monthly budget, 100+ monthly conversions. 30 to 40% exact match, 30 to 40% phrase match, 20 to 30% broad match. Smart Bidding (Maximize Conversion Value or Target ROAS) across most campaigns, manual CPC reserved for brand defense and edge cases.

The structural decision in this phase is whether to consolidate or segment. Smart Bidding works better with consolidated campaigns (more conversion data per learning surface), but consolidated campaigns are harder to read in reporting. The compromise that works across well-managed accounts: consolidate by intent and audience (one campaign per intent/audience combination), segment by match type within campaigns through ad groups.

Mature account, mid to high budget, ROAS goal

Same conversion thresholds as above, but with conversion values flowing accurately. 20 to 30% exact match for brand defense and high-value query capture, 30 to 40% phrase match for volume, 30 to 50% broad match for discovery. Target ROAS bidding across the volume drivers, Target CPA on the precision plays.

The match-type allocation tilts toward broad match in ROAS-goal accounts because the value signal lets Smart Bidding filter the expanded query pool more effectively. Without the value signal, the same allocation would tilt toward exact match.

Regulated vertical, any budget, any goal

Legal, medical, financial, gambling, alcohol. 60 to 80% exact match, 20 to 40% phrase match, 0 to 5% broad match (only with heavy oversight). Manual CPC or Target CPA bidding, no Maximize Conversions (which can chase low-quality conversions in regulated verticals).

The match-type discipline in regulated verticals is not a performance optimization, it is a compliance requirement. Broad-match expansion can trigger ads on queries that violate platform policies or vertical-specific advertising rules. The cost of a policy violation (account suspension, lost trust, regulatory exposure) is higher than the cost of leaving volume on the table.

Local services, geographic intent

50 to 70% exact match with explicit geo-modifiers (“plumber Denver,” “real estate attorney Austin”), 20 to 40% phrase match for natural-language variations (“emergency plumber near me,” “best real estate attorney in Austin”), 0 to 10% broad match. Manual CPC or Target CPA.

The exact-match dominance in local services reflects the structural property of the queries: local intent is usually expressed in a small number of canonical phrasings, and the long-tail expansion that broad match captures elsewhere does not exist as cleanly in local search.

Account structure implications

The match-type strategy has cascading effects on account structure. The structure that works for an exact-match-heavy account does not work for a broad-match-heavy account, and vice versa.

Exact-match-heavy structures. Smaller ad groups (5 to 10 keywords each), more ad groups per campaign (10 to 20), tighter ad-copy alignment to the keyword set, more specific landing pages, more ad-group-level negatives.

Broad-match-heavy structures. Larger ad groups (20 to 40 keywords each), fewer ad groups per campaign (3 to 7), broader ad-copy that can serve a wider query range, broader landing pages, more campaign-level and account-level negatives.

Mixed structures. Match types separated into ad groups within the same campaign, ad-group-level bid modifiers if running manual CPC, separate campaigns for match-type-by-intent combinations if running Smart Bidding (the bidder works better with consolidated learning surfaces).

The 2026 trend, per Search Engine Land’s coverage of agency case studies, is consolidation. The hyper-segmented account structures of 2018 (Single Keyword Ad Groups, narrow ad groups by exact-match keyword) have been replaced by semantically themed ad groups with mixed match types and Smart Bidding handling the fine-grained optimization. The driver is the bidder’s appetite for conversion data: more conversions per learning surface produces better optimization, and consolidation feeds the bidder while segmentation starves it.

The 2026 paradigm and where it is heading

Google’s direction is unmistakable. The platform is moving from “you give us keywords, we match queries to those keywords” toward “you tell us your business intent, we find queries that align with it.” The endpoint of that direction is keywordless campaigns, which Google launched in beta in early 2026 and which represent the logical conclusion of the match-type evolution.

Match types still matter in 2026 because keywordless campaigns require conversion-data maturity that most accounts do not yet have, because the platform is mid-transition rather than at the endpoint, and because keywords still produce more granular control than the keywordless alternative. The accounts that will navigate the transition successfully are the ones that build the conversion-data foundation now, learn to read the search-terms report as the primary feedback loop, and treat match-type strategy as an evolving structural lever rather than a one-time decision.

The recommendation that holds up across the 2026 platform: start with exact match plus phrase match for control and data accumulation, expand to broad match once the prerequisites are met, treat broad match as a discovery channel layered on top of a controlled exact-and-phrase foundation, maintain thick negative-keyword lists as the steering mechanism, and revisit the match-type allocation every quarter as the account matures.

Conclusion

Match types in 2026 are a steering signal, not a matching contract. Exact match steers narrow and high-density. Phrase match steers wide enough for volume and tight enough for quality. Broad match steers wide and lets Smart Bidding filter, if the conversion infrastructure is in place to do the filtering. The accounts that win in this environment are the ones that match the steering signal to the account’s actual maturity, budget, and goal, and that revisit the alignment as the account grows.

The platform is unforgiving of broad-match adoption before the prerequisites are met, and forgiving of disciplined exact-match foundations that expand into broader match types over time. Choose the discipline first, the volume second. The volume compounds when the foundation holds.