Google begins auto-upgrading Search campaigns to AI Max in September 2026, starting with campaigns that use automatically created assets or the campaign-level broad match setting. This is not an opt-in beta any more: AI Max left beta in April 2026, and the upgrade path now runs on Google’s timetable rather than yours. The single most important thing to understand is that AI Max is three separate features bundled under one switch, and the risk profile of each is completely different. Accounts that turn the bundle on without setting brand exclusions and URL restrictions first are the ones that get hurt.
Quick reference:
- Auto-upgrade begins September 2026 for campaigns using automatically created assets or campaign-level broad match
- Dynamic Search Ads sunset and auto-upgrade to AI Max begins February 2027
- Google reports an average 7% lift in conversions or conversion value at similar CPA/ROAS when the full feature suite is used
- Independent account-level testing is far less positive — a large majority of advertisers report neutral or negative results
- Three features, one switch: search term matching, text customisation, final URL expansion
- Brand exclusions and URL restrictions are the two controls to set before upgrading, not after
What is AI Max for Search, exactly?
AI Max is a bundle of three automation features applied to a standard Search campaign. It is not a new campaign type, which matters: your campaign keeps its structure, budget and bidding strategy, and the changes happen to matching, creative and destination.
Search term matching expands which queries your campaign is eligible for, going beyond the behaviour of traditional broad match. Google describes it as using the full context of your landing pages and existing keywords rather than keyword-to-query similarity alone.
Text customisation generates ad copy variations automatically, pulling from your website content and existing assets, and adapts headlines to the specific query.
Final URL expansion lets Google send traffic to whichever page on your site it judges most relevant, rather than the final URL you set on the ad.
Those are three genuinely different decisions. You can enable them independently, and in most accounts you should. Google’s AI Max documentation covers the mechanics; what it does not tell you is which of the three is likely to cost you money in your specific account.
What does the September 2026 auto-upgrade actually do?
The upgrade applies first to campaigns already using automatically created assets or the campaign-level broad match setting. If you run either, treat September as a deadline rather than a suggestion.
Google’s framing is that these campaigns already delegate the equivalent decisions, so the upgrade is continuity rather than change. That is defensible for automatically created assets, which are close to text customisation in spirit. It is much weaker for campaign-level broad match, because broad match still resolves against your keyword list, while AI Max search term matching also reads your landing pages. The set of queries you become eligible for genuinely widens.
The practical consequence is that accounts drift into AI Max without an explicit decision, and the first sign is a change in the search terms report four to six weeks later, after the budget has already moved. Diarise a search terms review for the first week of October. If you have never built negative keyword lists, build them before September, not after.
Does AI Max actually improve performance?
This is where the honest answer diverges from the marketing.
Google reports an average 7% increase in conversions or conversion value at a similar CPA or ROAS when advertisers use the full feature suite rather than individual features. That figure is a comparison between full adoption and partial adoption, not between AI Max and no AI Max, which is a meaningfully different claim from the one most people hear.
Independent testing paints a more mixed picture. Across a body of controlled tests, accounts using all three features together did outperform accounts using single features, and text customisation in particular lifted ad relevance and Quality Score in a majority of tests. But measured at account level rather than campaign level, the large majority of advertisers saw neutral or negative outcomes.
Both things can be true. Campaign-level lift with account-level neutrality is the signature of query cannibalisation: the AI Max campaign wins traffic that another campaign in the same account would have won anyway, reports it as incremental, and the account total barely moves. This is the same pattern we see with Performance Max in lead generation accounts, and it is why the only test worth running is measured on account totals over at least six weeks.
Which of the three features should you actually turn on?
Across the accounts we audit, the three features have very different risk profiles, and treating them as one decision is the most common mistake.
Text customisation — lowest risk, turn it on first. The downside is bounded: at worst you get ad copy that is slightly off-brand, and you can see it and fix it. The upside is real, because relevance improvements feed Quality Score. Set your brand voice through pinned headlines if there are claims you must control, and check the asset report monthly.
Search term matching — medium risk, budget-dependent. On an account with tight negatives and a healthy budget, the wider net finds genuinely new demand. On a small budget it dilutes: you fund experimentation with money that was previously buying your best-converting terms. Below roughly $3,000 AUD per month in ad spend, we generally leave it off.
Final URL expansion — highest risk, off by default. This hands page selection to Google. If your site has thin category pages, an outdated blog, or landing pages that do not match ad intent, Google will find them and send paying traffic there. Turn it on only after a landing page review, and always with URL restrictions set.
Which controls must be set before you upgrade?
Two, and both exist specifically because the early version of AI Max caused problems without them.
Brand exclusions. Without them, search term matching will serve your ads against competitor brand queries and, more expensively, against your own brand terms inside campaigns you did not intend to be brand campaigns. That corrupts your reporting: brand traffic converts at a high rate and a low cost, so it flatters whichever campaign absorbs it, and you make budget decisions on a blended number that hides how your non-brand prospecting is really doing.
URL restrictions. These constrain final URL expansion to a defined set of pages. If you enable expansion at all, restrict it to your actual commercial pages. The failure case we see most often is paid traffic landing on blog articles: high engagement, near-zero conversion rate, and a cost per acquisition that quietly climbs while everything else looks fine.
Set both before the upgrade lands. Retrofitting them after a month of spend means you have already paid for the lesson.
How should you test it without losing a quarter?
Treat it as a structured test with a defined end date, not a setting you flip and forget.
Baseline first. Record four weeks of pre-upgrade data on the metrics that matter: total account conversions, cost per acquisition, and revenue if you are e-commerce. Not campaign metrics — account metrics. Our ROAS calculator will tell you the return you actually need to hold.
Change one thing. Enable text customisation alone for four weeks. Then add search term matching if the budget supports it. Adding all three at once makes it impossible to attribute the result.
Give it six weeks minimum. Your bidding strategy re-enters a learning period when eligibility changes. Judging week two is judging the learning phase, which is how good tests get killed and bad ones get kept. The same discipline applies to any Smart Bidding change.
Read the search terms report weekly. This is the early warning. If irrelevant queries appear, add negatives immediately rather than waiting for the test window to close.
Judge on account totals. If account conversions did not move but the AI Max campaign looks brilliant, you have measured cannibalisation, not improvement.
Who should stay on standard Search campaigns?
Some accounts genuinely should not upgrade beyond the mandatory minimum, and it is worth being clear about which.
Accounts under roughly $2,000 AUD per month have too little data for broad automation to learn from. Accounts in regulated sectors — legal, financial services, health — where generated ad copy creates compliance exposure should keep text customisation tightly pinned or off. Accounts with a small set of high-value commercial pages and no content depth get nothing from URL expansion. And any account whose conversion tracking is not verified accurate should fix that first: automation optimises toward the signal it receives, and a broken signal at scale is worse than a broken signal at small scale.
For everyone else, the realistic position is that AI Max is not optional in the medium term. Dynamic Search Ads are being upgraded to AI Max from February 2027, and the direction of travel across every campaign type has been consistent for three years. The choice is not whether, but whether you set the controls before or after Google sets them for you.
Frequently asked questions
When exactly does the AI Max auto-upgrade start? September 2026 for campaigns using automatically created assets or the campaign-level broad match setting. Dynamic Search Ads campaigns follow, with their sunset and auto-upgrade beginning February 2027. If you run neither, you are not on the current auto-upgrade list, but you can still enable AI Max manually.
Can I opt out of AI Max? You can disable the individual features on most campaigns, and you can avoid the September trigger by not using automatically created assets or campaign-level broad match. What you cannot do is opt out permanently — Dynamic Search Ads are being retired into AI Max, and the pattern of previous Google transitions suggests the manual controls narrow over time.
Will AI Max increase my CPCs? Not directly, since it does not change your bidding strategy. Indirectly it can, because wider query matching makes you eligible for auctions with different competitive dynamics, including more expensive head terms. Watch average CPC alongside conversion volume for the first six weeks rather than either on its own.
Is AI Max the same as broad match? No. Broad match expands matching against your keyword list. AI Max search term matching also uses your landing page content as a signal, so it can make you eligible for queries with no close relationship to any keyword you have. If you want the underlying mechanics, our guide to broad match in 2026 covers how the matching layer works.
Does AI Max work with Performance Max? They are separate products and can run in the same account. Doing so without brand exclusions on both is how accounts end up with campaigns bidding against each other on the same queries. If you run both, set exclusions on both and check the search terms overlap monthly.
How much budget do I need for AI Max to work? There is no official minimum, but the practical floor for search term matching is around $3,000 AUD per month in ad spend. Below that, the widened query set spreads a small budget across more auctions and slows the learning that Smart Bidding depends on. Text customisation has no meaningful budget floor. You can model what your target requires with our budget calculator.
If your account uses automatically created assets or campaign-level broad match, the September upgrade will reach you whether or not you plan for it. We audit accounts before that kind of change lands, set the exclusions and restrictions properly, and measure the result on account totals rather than campaign vanity metrics. Get in touch and we will tell you what your account is actually exposed to.