Stop Managing Keywords. Audit the Matches.

The report lives inside the familiar search terms interface, one dropdown switch away. A row now carries five things at once. The search term: what the user actually typed. The headlines and descriptions: which ad copy appeared alongside that query.

The landing page: where the click went. The match source: whether the traffic came through AI Max expanded matches or through landing page and URL inclusions. The asset source: whether the headline and descriptions came from your assets or were generated by Google.

The match source column is the part that will change how people read their accounts. Google documents the distinctions in its developer resources, and WebProNews' read of those docs spells out the match types. AI_MAX_KEYWORDLESS means the match came from your site content with no keyword involved.

AI_MAX_BROAD_MATCH means the system expanded from your existing keywords. Vertical data feeds cover matches driven by structured product or service data. Kaurra's line in the LinkedIn post is the one to keep: "This report gives us another way to identify mismatches between search intent, ad messaging, and landing pages.

Two properties of the report deserve attention before anyone builds a process on it. First, it has appeared in at least one account. Google has not confirmed when, or whether, it rolls out broadly. Second, it is still a stitched view of what the system decided. It shows the matches, not the reasoning.

You get the trio, not the transcript of the model's thought process. Both limits matter for the playbook below. They do not change the direction. The direction is that the match, not the keyword, is the unit of management.

What AI Max changed about paid search is the background read. That post, from August, covered the shift of decisions from marketer to machine. This report is the machine's first attempt to show its work.

Three features make up AI Max. Search term matching goes beyond your keyword list, serving ads on queries that are similar in meaning to your keywords, assets, and landing page content. Text customization rewrites your headlines and descriptions for the specific query, drawing on material from your assets and site pages.

Automatic landing page selection picks the page to send the visitor to. The list you built over years still shapes campaign structure and signals direction to the system. It no longer defines where your ads can appear.

The migration already happened in most accounts, and many marketers did not choose it. From September 2026, Google moved Search campaigns using automatically created assets to AI Max with search term matching and text customization on, and campaigns using the campaign-level broad match setting moved with search term matching on. Some advertisers accepted those settings through a recommendation or a rep's call.

The result is the same either way. Search terms reports now carry a new match type, AI Max, and the keyword list is one input among several.

Google's pitch for the shift is a number: an average of 14% more conversions at similar cost, rising to 27% for campaigns built mainly on exact and phrase match. The independent reading is more complicated. An analysis of more than 250 campaigns found a median revenue lift of 13% alongside a median 16% rise in cost per conversion. And in the accounts that analysis audited, 22% to 33.5% of spend already goes on queries that never convert.

AI Max expands matching further. Those three numbers are the whole argument for the audit habit. More reach, higher unit cost, and a fifth to a third of spend on queries that never convert.

The keyword list is now a suggestion. The match is the campaign.

Read that sentence as a job description change. The marketer who spent a decade building keyword lists, mining search terms, and layering negatives owned a boundary. The system now crosses that boundary by design.

What the marketer owns instead is the review of what the system did with the freedom. Google just shipped the first interface for that review. Use it like the new job depends on it, because it does.

A match audit needs a vocabulary before it needs a spreadsheet. Every bad trio falls into one of four patterns. Name the pattern first, because the fix follows the pattern.

Intent mismatch. The query means something different from the offer. The classic example from the trade coverage: a basic car wash query triggering a premium ceramic coating ad. Nothing is wrong with the headline or the page.

The pairing is wrong. This is the most expensive mismatch because it burns money on people who were never buyers. It is also the one AI Max creates most often, because meaning-similar matching is the feature, not a bug.

Copy mismatch. The query is right, but the headline the system wrote does not say what you sell. Text customization rewrites your copy for the query, and it can paraphrase its way into claims you would not make.

In regulated categories like healthcare, finance, and legal, an uncontrolled paraphrase lands on the compliance desk, not the style desk. The asset source column exists to catch exactly this. When the copy came from Google and the claim came from nowhere, fix the asset or the permission that let the system write it.

Landing mismatch. The query is right and the headline is right, but the system sent the click to the wrong page. A query about pricing lands on the features page. A query about a specific product lands on the category page.

Automatic landing page selection optimizes for relevance at scale, and at scale it gets some picks wrong. This mismatch is quiet. The campaign metrics look fine while the conversion rate bleeds, because the user never found what they came for.

Brand cannibalization. The query is your own brand name, matched through AI Max expansion, served with a rewritten headline, at a cost per click you would not have paid for a query you already own.

Or the match is a competitor's brand term your URL inclusions pulled in. This is the mismatch that shows up in board-level conversations, because someone asks why the brand campaign's cost rose and the answer is that the system decided your keywords were a suggestion too.

Learn the four patterns until you can spot them in a scroll. The audit below is just the discipline of looking for them on a schedule.

Run this every week on every AI Max account that spends enough to matter. It takes under an hour once the habit is set, and it is the most valuable hour in the account.

First, open the report and sort by spend, not by clicks. Mismatches hide in cheap clicks. They cost money in expensive ones. Look at the rows that spent the most first. A trio that spent 10% of the week's budget on a query that never converts is the whole audit.

Second, read each top-spend trio as a sentence. Query, headline, landing page. Say it out loud if you have to. "Someone searched basic car wash, saw premium ceramic coating, landed on the detailing packages page." If the sentence sounds like a misunderstanding, tag it with one of the four patterns.

Third, check the match source and asset source on every tagged row. The source columns tell you which lever to pull. A keywordless match on a query you never want means a URL inclusion or page content problem, not a keyword problem.

A Google-generated asset with a claim you would not make means tightening the asset controls, not adding a negative keyword. Pulling the wrong lever is how audits become busywork.

Fourth, fix by pattern, not by row. An intent mismatch at scale means the campaign's direction signal is wrong: the keywords, the assets, or the landing pages are telling the system to find the wrong neighbors. A copy mismatch means the text customization permission is too loose for the category.

A landing mismatch means the URL inclusions are too broad. Fix the pattern and you fix every row like it. Fix rows and you will be back next week.

Fifth, write down the three worst trios and what you changed. This is the part everyone skips and the part that compounds. Six weeks of trio logs show whether the system is learning or drifting. A system that keeps producing the same mismatch after you fixed the pattern is a system you should be negotiating with, by turning settings down, not one you should be hoping about.

One caution from the trade coverage applies to the whole ritual. Brandstart's autumn roundup of the Google Ads changes flags it: filtering the search terms report by the AI Max match type leaves out "Other search terms," so the filter understates what AI Max actually spent. Run the audit on the full view, not the filtered one.

The filtered view is a highlight reel. The full view is the account.

The table is the cheat sheet. Print it mentally or literally. The report below it is the workflow.

Honesty first. The report has been spotted in at least one account. It is a test, not a rollout, and Google has not said when every account gets it.

Build the habit on the standard search terms report today, where the AI Max match type already appears alongside exact, phrase, and broad. When the combinations view arrives in your account, it makes the habit faster. It does not start it.

Second, the report shows decisions, not reasoning. You see which query got which headline on which page. You do not see why. That is enough for the audit and not enough for the argument. Do not try to reverse-engineer the model's intent matching from a few dozen rows. Manage the outputs.

Third, the September migration is still settling. Brandstart's autumn roundup notes that confirmation that every eligible campaign has migrated has not been seen. Some accounts are running a mix of old and new settings. Audit what is actually running in each account, not what the rollout schedule says should be running.

Fourth, the numbers cut both ways. A 13% median revenue lift with a 16% median rise in cost per conversion does not settle the AI Max debate. It describes a trade.

More volume at higher unit cost, with a fifth to a third of spend on queries that never convert. Whether that trade is good depends on the account's margins and on whether someone is running the audit. Without the audit, you are buying the lift and donating the waste.

Two things. First, the rollout. If the combinations view reaches every account, the match audit becomes standard practice within a quarter, and the agencies that already run it will have a head start they did not have to earn. Watch your own account for the dropdown before you read anyone's rollout timeline.

Second, the settings. AI Max arrived through defaults and recommendations, and settings that arrive quietly tend to get tightened loudly later. The marketers who document which permissions they granted, and why, will be the ones who can defend their setup when the questions come. The audit log is that documentation. Start it this week.

Your keyword list was the map. The matches are the territory. Walk the territory.