Google AI Max: What Advertisers Give Up When Keywords Become Optional

Google now says the full AI Max for Search suite adds an average of 7% more conversions over search term matching alone. At launch it quoted 14%. Google AI Max for Search campaigns trade keyword control for reach, and the official documentation is unusually specific about which controls go. Since 1 September 2026 that trade has been made for many advertisers by default.

The short answer

AI Max for Search adds three features to an existing Search campaign: matching beyond your keywords, generated ad text, and landing pages chosen by Google. You keep negatives, brand lists, URL exclusions and locations. You lose pinned assets once URL expansion is on, and cannot edit generated text.

What Google AI Max for Search actually changes

Google describes AI Max as "an optimization layer within existing Search campaigns". Your campaign keeps its name, budget, ad groups and history. What changes is how much of the matching and copy Google decides.

The Google Ads help documentation lists three features. Search term matching extends your keywords using what Google calls "broad match and keywordless technology". Text customization, formerly automatically created assets, writes headlines and descriptions from your landing page, ads and keywords. Final URL expansion sends a click to whichever page on your domain Google judges most relevant to the query.

Keywordless matching is the real departure. Broad match starts from a keyword you wrote. Keywordless matching starts from your landing pages and assets, so a query can trigger your ad without resembling your keyword list.

The help centre states plainly that "New Search campaigns have the AI Max feature turned on by default". Turning AI Max on also switches on text customization and final URL expansion. Each must be switched off separately.

Google emailed advertisers on 5 August 2026 to say campaigns using the campaign-level broad match setting or automatically created assets would be upgraded to AI Max from 1 September. Dynamic Search Ads were originally on the same timetable. Google moved that part to February 2027 after advertiser feedback.

The AI Max controls Google keeps, and the ones it removes

The headline worry is that AI Max ends keyword control. The documentation does not support that reading. Most targeting controls survive, and two are new. The losses sit in creative and landing page control, where brand and legal teams care most.

Brand and location controls are actually stronger

Brand inclusions work at campaign and ad group level, with ad group lists taking precedence. Brand exclusions work at campaign level. Locations of interest can be set per ad group, and Google says they apply "even in keywordless matches". Standard Search campaigns never had that control, so it is a genuine gain.

Negative keywords also survive. The help centre is direct: "Negative keywords will be respected even with AI Max turned on."

URL controls exist, but only with expansion on

URL exclusions sit at campaign level and stop Google sending clicks to pages you name, such as careers or support. URL inclusions sit at ad group level. With inclusions set and final URL expansion off, Google uses only your listed URLs as dynamic landing pages for that ad group.

Google's explanation of how AI Max works warns that tracking templates without a compatible {lpurl} tag can send expanded landing pages to 404 errors. If your agency hard-coded URLs into tracking templates, check that first.

Text control is where the real concession sits

Text guidelines let you set term exclusions and messaging restrictions on tone. Asked whether advertisers can edit generated assets, the help centre FAQ answers: "No, these assets will fully be automated." You can remove an asset after it runs. You cannot approve it before.

Pinning goes too. If both text customization and final URL expansion are on, pinned responsive search ad assets are not respected. Pinning is also ignored when URL inclusions are set. For a regulated advertiser who pins a disclaimer to headline position one, that rule may decide the question.

Bidding and budget requirements narrow who should try it

Google says search term matching, text customization and final URL expansion "require conversion-based Smart Bidding strategies in order to function effectively". Search term matching will not work at all with manual CPC. The documentation also says AI Max "won't be effective if campaigns are limited by budget".

If your conversion tracking is not trusted inside your own company, do not switch AI Max on. Every feature optimises toward that signal, so a bad signal gets amplified.

Compiled from Google's four-part AI Max help series, read in October 2026. The pages carry no last-updated date, and one still says Editor and API support is "expected later in 2025", so treat specifics as current at reading, not permanent.

What AI Max reporting shows you, and what it cannot

Reporting is the part of the documentation that improved most since launch. It is also the easiest to over-read: a report showing what matched does not show what was gained.

The search terms report now separates the sources

The search terms report adds "AI Max" as a match type. A source column shows whether a match came from "broad match expansion or keywordless matching". A further view shows the search term, the headline served and the URL clicked in one row.

The keywords report gains AI Max summary rows. The landing pages report gains a "Selected by" column showing when Google chose the page. Asset reports now show spend and conversions, not only impressions. Those four views are enough to audit what AI Max did.

Incrementality is the number no report gives you

None of these reports tells you whether a conversion would have happened anyway. A query matched by AI Max might have reached you through an existing broad keyword, a Performance Max campaign, or an organic listing. The report credits AI Max with the match, not with the margin.

Overlap with Performance Max makes this harder. Google's rule is that "if a query is eligible to match to candidates in both AI Max for Search campaigns and Performance Max, then the one with the highest Ad Rank will serve". Spend can therefore move between your campaigns without any change in total demand. The wider measurement gap is set out in the piece on measuring demand that attribution cannot see.

What Google says about AI Max incremental conversions

Google has published two headline performance numbers for AI Max. They are often quoted as if the second replaced the first, but they measure different comparisons.

The May 2025 launch post, by Brian Burdick of Google Ads, said advertisers activating AI Max "will typically see 14% more conversions or conversion value at a similar CPA/ROAS". It added that campaigns "still mostly using exact and phrase keywords" saw a typical uplift of 27%. The footnote defines those as campaigns with more than 70% of conversions from exact or phrase match, and cites Google internal data from non-Retail advertisers.

The April 2026 announcement that took AI Max out of beta gave a new figure. It claimed an average of 7% more conversions or conversion value at similar CPA or ROAS "when using the full feature suite". The comparison is the full suite against search term matching on its own. It also excludes Retail.

So the 7% is not a downgrade of the 14%. It is a claim about how much text customization and URL expansion add on top of matching. Trade coverage that reported a cut from 14% to 7% compared two different experiments. The fairer reading: Google's own data says most of the lift comes from matching, and the creative features add a smaller increment.

Two exclusions matter more than the headline numbers. Both claims omit Retail advertisers, and neither publishes a sample size or test design. For reading vendor figures this way, see the forensic method for checking AI ROI case studies.

What independent AI Max data shows so far

The largest independent dataset we found comes from Smarter Ecommerce. Its head of ecommerce insights, Mike Ryan, analysed more than 250 Search campaigns using AI Max. Search Engine Journal reported the findings on 5 March 2026.

The median campaign saw conversion value up 13% and CPA up 16%. Median ROAS impact looked roughly neutral, but the spread was wide: from 42% above the baseline ROAS to 35% below it. Only 22% of campaigns landed close to their original ROAS targets.

That distribution is the finding. Large wins and large losses averaging near zero make AI Max a high-variance change, not a reliable increment.

About 1 in 6 advertisers ran it with Dynamic Search Ads, about 1 in 4 with Performance Max, and nearly half ran all three at once. Overlap splits conversion data across campaigns, which weakens the Smart Bidding signal AI Max depends on.

The limits are real. The coverage we read does not give the time window or say how the baseline was built. We cannot tell a before-and-after comparison from a controlled experiment. The sample leans towards ecommerce, which is the segment Google's own figures exclude. Read the numbers as directional evidence from one agency's client base.

Microsoft Ads AI Max: the same switches, a firmer keyword rule

Microsoft Advertising announced its own AI Max for Search on 21 April 2026 and opened a pilot in May. It began rolling it out to all accounts globally on 19 August 2026. The global rollout was announced by Microsoft Ads Liaison Navah Hopkins in a LinkedIn post, as PPC Land reported, rather than through the company blog.

The feature set mirrors Google's: search term matching, text customization and final URL expansion under one toggle. Hopkins tied search term matching to conversational queries "in AI experiences on Bing and Copilot". Brand inclusions and exclusions allow up to 20 brand lists per campaign, with term exclusions for generated text and URL exclusions for expansion.

Three differences matter for a buyer. First, Microsoft's exact match keywords keep absolute auction priority, so expansion cannot override a qualifying exact keyword. Second, optimization experiments can test all three features together or one at a time. Third, Microsoft has announced no forced conversion date and has published no performance figure for AI Max at all.

A Google AI Max campaign that was upgraded from Dynamic Search Ads gets converted back into a DSA campaign on import to Microsoft. Microsoft has also posted a help notice saying its older Predictive Matching feature "is moving to Search term matching within AI Max", without an effective date.

Our view is that Microsoft's slower, opt-in defaults are the better design for advertisers. Nothing Microsoft has published shows its reporting can separate useful expansion from waste, so the same tests apply on both platforms.

Where this argument is weakest

This post leans sceptical. The case for AI Max is stronger than sceptics usually allow.

Google can see the counterfactual and you cannot

Google observes every auction across every account. If it says matching beyond your keywords finds converting queries you were missing, it is reporting something it can measure and you cannot. The one independent median we found, at 13% more conversion value, lands close to Google's 14% claim. That is real corroboration of the direction, even if the cost side differs.

Keyword control was also partly an illusion before AI Max. Broad match has expanded queries for years, and close variants loosened exact match long ago. That trend runs through the analysis of paid search click-through rates under AI Overviews.

Our test plan is not free either

A 50/50 experiment halves the traffic each arm sees. A small account may never reach a conversion count that separates a real effect from noise. Smart Bidding also needs time to learn in the treatment arm. For a campaign with a handful of conversions a week, you may not be able to test AI Max to a standard that justifies any verdict.

That leaves a judgement call, not a measurement. In that position we would switch on search term matching alone, keep text and URL control, and watch the search terms report weekly.

A test plan before you switch AI Max on

The help centre FAQ describes a one-click experiment that runs "a 50/50 split test comparing your base Search campaign against an AI Max treatment arm". Inside the treatment arm you can switch search term matching or asset optimization on or off per ad group. Testing the whole suite at once tells you whether the bundle works, not which part did.

Step 3 is the one most teams skip, and it catches the expensive failure. If AI Max wins inside its own campaign while account-level conversions stay flat, it has moved credit, not found demand. Setting a stop rule before the test starts is argued in the case for writing kill criteria before an AI project begins.

We would also keep an exact match copy of your best keywords in the campaign. Google's documentation notes that "broad keywords are not treated as exact match keywords in prioritization", so exact keywords are your only way to keep your highest-value queries on copy you wrote. If you were upgraded on 1 September without testing, start at step 2 with the data you already have.

Paid clicks are one input to a channel shifting toward answers with no click at all, covered in the post on metrics for zero-click search. A test that improves CPA on a shrinking click pool is worth running, but it is not a strategy.

Frequently asked questions

Where to start

This week, check every Search campaign's settings for AI Max. If it is on and you never chose it, filter the search terms report by the AI Max match type and read the source column. That view shows how much spend goes to queries outside your keyword list.

Then decide one thing in writing before any test runs: the CPA or ROAS figure at which you would switch AI Max back off. Without that number, every result will read as a reason to keep going.

Related reading