Broad Match Isn't a Strategy. It's a Tax on Weak Account Structure

Two settings in your account are about to change themselves, and nobody is going to ask you first. Google emailed advertisers the date on August 5. Starting September 1, 2026, campaigns running automatically created assets or the campaign level broad match setting get upgraded to AI Max for Search on their own. The two cohorts land differently. Automatically created assets pick up search term matching and text customization; campaign level broad match campaigns get search term matching only. There is no opt in step. You turn the legacy feature off first, or you turn AI Max on yourself and keep hold of the setup.

Broad match is a setting. It lets Google serve your ad against searches it judges related to your keyword rather than only against the keyword. I have watched accounts run it well and I have watched accounts get quietly emptied by it, and the difference between those two outcomes was never the setting. September 1 does not change what broad matching costs an account. It changes how many accounts are about to find out what theirs costs.

Broad match is not the mistake people think it is

The case for broad match is stronger than its reputation among people who write about it. Optmyzr looked at February 2026 data across 30,000 Google Ads Search accounts and its conclusion was not that anyone should avoid the thing: "The accounts getting the most from match types tend to use all three deliberately: exact for high intent, high value keywords; phrase as the day to day workhorse; broad for discovery and scale." In the ecommerce segment of that study, broad and phrase finish nearly level on ROAS. That is not a rounding error in our favour.

Google's own position deserves to be stated properly too, because it is a specific argument rather than marketing. Ginny Marvin, asked by Smarter Ecommerce in February 2026 whether splitting campaigns by match type is an outdated approach, agreed and explained the mechanism: "Segmenting your traffic this way can reduce the overall efficiency of your Smart Bidding strategy, resulting in a higher marginal cost for some conversions." The tight, sharded, match type by match type structure a lot of us are proud of fragments the conversion data the bidder learns from. That is a real cost and it is worth weighing.

Exact match has been losing budget share to broad for four years now, and Optmyzr is blunt about why that happened: "advertisers have grown more comfortable with automation, and Google has made broad match easier to justify." We have written up the efficiency side of that shift already, so I will leave it there.

Broad match vs phrase match in 2026: what actually sets the gap

The gap between broad and phrase changes size depending on the account, and the variable is the account rather than the match type. Optmyzr's segment split is the cleanest evidence anyone has published on this. Phrase match carries the largest share of non brand spend across the study and "consistently punches above its weight on conversions." In ecommerce it posts the highest conversion rate of any match type in the entire dataset. In lead gen it dominates both spend and conversions, with the gap over the other two wider than in any other segment, and broad match "loses its footing more noticeably in lead gen than in e commerce" for a reason Optmyzr names outright: there is no conversion value for Smart Bidding to learn from.

Read that again with an account in mind rather than a dataset. Broad match performs worst where the feedback signal is thinnest. The accounts least equipped to absorb the cost are the ones being charged the most for it.

Optmyzr's guidance comes with conditions attached, and the conditions are where the money is: "Without strong Smart Bidding signals and a well maintained negative keyword list, broad match can spend a lot chasing traffic that was never going to convert." Every write up I have seen treats that sentence as a checklist to tick before switching broad on. It works better as a rate card. Whatever you have not built, you pay for, at a rate the platform sets and never itemises.

Chris Chambers, who runs paid search at Understory, put the same thing in practitioner terms: "In ecom nowadays you don't need negatives as much and smart bidding for mid large brands works great because it has thousands of conversions per month to learn from. That's not true for lead gen usually." A B2B account with 40 form fills a month and a single conversion value on all of them is buying broad match traffic at the worst rate available, and nothing in the interface says so.

The tax lands on your tightest keywords first

Here is the finding that should have changed more minds than it did. Smarter Ecommerce traced where AI Max expansion actually originates across roughly a million impressions, sorted by the match type of the keyword that let it in. Exact match accounted for 80.11 percent of it. Phrase match 19.52 percent. Broad match 0.38 percent.

Sit with that distribution for a second. The loosening is being applied to the most disciplined part of the account. An exact match keyword is the strongest intent signal you own, which makes it the most attractive thing in the account to expand from, and the keywords you built specifically to keep loose traffic out are the ones carrying it in.

Google's documentation explains how that is possible, in a footnote. The prioritization page defines an exact match keyword as taking first priority when it is "identical" to the search term, then defines identical: it "includes spell corrected search terms, such as skydiving license, but doesn't include plurals or synonyms like license for skydiving." Your exact match keyword loses its top tier priority to a plural. That is not a loophole somebody found, it is the published behaviour.

Smarter Ecommerce also found AI Max landing on queries its clients' existing broad match keywords were already covering, 49 percent overlap in one account and 63 percent in another, and traced the cause to legacy Broad Match Modified keywords that were converted years ago and still behave like phrase match. Google retired Broad Match Modified as a match type in 2021. Those keywords are still sitting in accounts, still spending, reading as broad in the interface and matching like something else in the auction. So a rename from 2021 is still deciding what some accounts pay this month.

Is AI Max just broad match with a new name?

No, and the honest answer is more useful here than the satisfying one. Xavier Mantica published a cost per conversion ladder from four months of his own testing in November 2025 that made the rebrand argument hard to ignore: AI Max at $100.37 against $43.97 for phrase match, with exact match and the close variant tiers in between. PPC Land called it a 90 percent higher cost than phrase. His read was blunt: "They couldn't convince advertisers to use broad match because we know it burns budget on irrelevant searches. So they rebranded it as an 'AI-powered feature' with a shiny new name." That is one practitioner's account set, with no sample size, spend or vertical published, and nobody has replicated it in the nine months since. It also contains no broad match figure at all, which matters if you want to use it as a broad match comparison.

The rebrand reading is too easy, though, and it lets AI Max off lightly. Broad match is one setting. AI Max is a bundle that arrives with search term matching, final URL expansion, generated and customised assets, and AI based routing between ad groups, and the September 1 upgrade delivers different pieces of it depending on which legacy feature triggered your campaign. Calling it broad match with a new logo understates what changes. The part that should worry you is not that it matches loosely, it is that one of the things it removes is the tier where your keyword gets consulted at all.

Worth keeping this proportionate: at the time smec measured it, AI Max was 0.57 percent of account level spend across more than 600 accounts, active in 12 percent of them, and 5 percent of spend inside the campaigns where it was enabled. This is not majority behaviour yet. September 1 is what moves that number without asking anyone.

Why the tax does not show up on any report

No report in Google Ads prices a query against what it turned into, which is why the cost of loose matching never appears as a line item you could go argue with. The prioritization documentation lays out four tiers: exact match keywords identical to the search term, then phrase and broad match keywords including AI Max that are identical to the term, then AI based ad group prioritization determined by "the meaning of the search term, all the keywords in an ad group, and landing pages within the ad group," then Ad Rank to break ties.

Then comes the exception, and it is the part worth reading twice. For what Google calls advanced search experiences, meaning Lens, AI Mode, AI Overviews and autocomplete, the documentation says that "because these searches are not considered technically identical to a keyword, keywords may not automatically be prioritized. Instead, AI based ad group prioritization ensures that the most relevant ad groups or asset groups are selected to match the user's overall intent." Tiers one and two get skipped. Neither your keyword nor its match type is consulted on that impression.

Reporting makes the accounting harder still. After Brad Geddes documented AI Max claiming credit for conversions that exact and phrase keywords had already delivered, PPC Land described the mechanism: AI Max traffic gets assigned to exact and phrase keywords in reporting when no corresponding broad match keyword exists, which makes evaluating any single match type on its own numbers impossible.

Michael Hulsmann of SearchSavior put the account side of it better than I am going to: "Google's matching keeps getting looser while negatives still use strict rules from a decade ago. Block a junk query and the same intent reappears under a different reported term next week. You're blocking symptoms while the cause keeps producing new ones."

How to work out what you are actually paying

Segment your own non brand spend by match type over a window long enough to mean something, and hold it against conversion values that reflect a downstream outcome rather than a form submission.

Go find the legacy Broad Match Modified keywords still sitting in your ad groups, converted in 2021 and never revisited since. That is a named root cause with published overlap figures behind it and you can check it in an afternoon.

Reconcile queries across campaigns instead of inside them, because two campaigns buying the same intent is invisible from either campaign's own report.

Then go group by group and ask whether the price you are paying reflects a decision somebody made or a default somebody inherited. That question is the whole job, and it is the work we productised as query level paid search optimization.

A price, not a plan

September 1, 2026 is the date on the two settings that upgrade themselves. The Dynamic Search Ads migration moved out to February 2027 and got most of the coverage, which has left a lot of advertisers believing their deadline moved as well. It did not. That extension applied to DSA and nothing else.

Broad match was never the strategy. It was the invoice, and the amount was decided by work nobody can see from the dashboard.

Frequently asked questions

Is broad match bad in 2026? No, and the framing is the problem. Broad match is a price you pay for reach, and the rate depends on whether your conversion values mean anything downstream, whether your negatives steer rather than tidy up, whether your ad group themes are tight enough to route a query on their own, and whether you kept a query history you can actually price. Optmyzr's read of 30,000 accounts is that the advertisers doing best with match types use all three on purpose.

Why is broad match showing up on my exact match keywords? Because expansion is applied to your keywords rather than routed around them, and it lands hardest on the tightest ones. Smarter Ecommerce found 80.11 percent of AI Max expansion originating from exact match keywords against 0.38 percent from broad. Google's documentation also defines an identical search term in a way that excludes plurals and synonyms, so an exact match keyword surrenders its first tier priority on a plural.

Do I have to switch to AI Max on September 1? Campaigns running automatically created assets or the campaign level broad match setting are upgraded on September 1, 2026 with search term matching switched on, and there is no opt in step to skip. Turning the legacy feature off beforehand, or adopting AI Max deliberately, are the two ways to stay in control of the configuration. If you read that the deadline moved to February 2027, that was the Dynamic Search Ads migration only.

What is the difference between broad and phrase match if Google expands both anyway? The difference is the starting point expansion gets applied to, and it still shows up in the data. Phrase match holds the largest share of non brand spend in Optmyzr's study while posting the highest conversion rate of any match type in the ecommerce segment. A tighter starting point produces a tighter set of expanded searches, which is the reason the starting point is worth choosing at all.

Do negative keywords still control broad match? Less than they used to, and the tooling has grown without the mechanics changing. Google raised the Performance Max negative keyword limit from 100 to 10,000 in March 2025 and added campaign level negative keyword lists in Google Ads Editor in November 2025, both of which are capacity rather than control. Matching kept getting looser over the same period, so a blocked query tends to return as a different reported term carrying the same intent.