AI Max isn't a campaign type. It's a transfer of control. Here's what you still own.
Google Ads AI Max is sold as a feature but behaves like a transfer of control. Here's what you still own — and where the real levers are hidden.
Google has spent eighteen months convincing advertisers that AI Max is a feature you toggle on. That framing is doing a lot of work. AI Max isn't a feature in the sense that Sitelinks or audience signals are features. It's a campaign architecture decision. Once enabled, it changes what you can target, what you can exclude, what data you see back, and which variables your team is allowed to influence. The right question isn't "should we turn on AI Max." It's "given that we've turned it on, what do we still own?"
What AI Max actually does
AI Max takes a Search campaign and lets Google's models expand keyword matching, generate ad variants, and route traffic to the URL the system thinks will convert best. The advertiser supplies the budget, the seed keywords, and the conversion goal. The system supplies almost everything else.
There is real performance lift in many accounts. We've seen it. The issue isn't whether AI Max can outperform a manually managed campaign — sometimes it does. The issue is that the levers you used to pull no longer exist, and the levers that replace them are not the ones being talked about in the platform's onboarding flows
What you've lost — and need to stop arguing about
Granular keyword control. Match types are heavily blurred. Your "exact match" keyword in an AI Max campaign is treated more like a strong signal than a constraint.
Negative keyword precision. Negatives still apply but the system's interpretation of "related search" is wider than it used to be. Detailed search term reporting. You can see categories of queries; you cannot reliably see every
query that triggered an ad.
URL routing control. The system selects the landing page from your inventory based on signals you
don't have full visibility into. Bidding signals at the campaign level. Most manual bid modifiers are gone or muted. Spending team meetings re-litigating these is wasted energy. They're not coming back. The faster you
accept the new control surface, the faster you can work on it.
What you still own
1. The conversion goal definition
This is the single biggest lever. If your conversion action is "all purchases," the system optimises for unit volume; if it's "purchases above £80 with positive contribution margin," the system optimises for something that actually matters. The distinction between these two settings is the difference between "AI Max didn't work" and "AI Max transformed our return." Almost every account that complains about AI Max has the first setup, not the second.
2. The asset inventory you give it
Headlines, descriptions, sitelinks, callouts, image assets, business names. AI Max can only mix what you supply. If you give it twenty mediocre headlines, you get mediocre ads. If you give it twelve sharp ones in three distinct angles, you get sharper ads. Asset quality has become the new bidding strategy.
3. The seed keyword set
Yes, match types are blurred. But the seed set is still where the model starts. Pruning it down to high-intent terms and removing brand-defence keywords that should live in a separate campaign is real work that pays.
4. Account-level negatives and brand controls
Account-wide negative lists, brand exclusion lists (now mandatory for any advertiser who cares about brand discipline), and the new content suitability controls. These are blunt instruments but they're yours.
5. The decision to use AI Max at all in a given campaign
Brand campaigns, competitor campaigns, branded shopping, and tightly targeted high-CPL B2B campaigns rarely benefit. AI Max is a tool, not a default. The campaigns where it earns its keep are mid-funnel volume drivers where the cost of letting the model explore is acceptable.
The reporting you should actually run
If you only look at the headline metrics AI Max gives you, you'll be flying blind. The reports we run on AI Max accounts:
Asset performance by combination, not in isolation. Find which headline-plus-description-plus-sitelink
groupings drive conversions and reinforce those.
Search category split. Even with reduced search term transparency, AI Max provides category-level
breakdowns. Use them to detect category drift — when the system starts spending on a category that
doesn't convert.
Conversion-value-per-pound at the conversion goal level, not ROAS as reported. ROAS as reported
includes value that may not be commercial.
Negative keyword incrementality. Test removing a negative for two weeks on a small budget split. If
nothing changes, the negative wasn't doing work.
The trap nobody is naming
The biggest mistake we see is teams treating AI Max as set-and-forget because "the AI handles it." It
doesn't. AI Max accounts need more attention than manual ones, not less — but the attention shifts.
You're no longer adjusting bids and keyword match types; you're auditing conversion goal hygiene, refreshing asset inventory, and reading category-level drift reports. Teams that don't rebuild their weekly cadence around the new surface find themselves three months in with a black box and no idea why performance shifted.
FAQ
Should we enable AI Max on every Search campaign?
No. Use it on volume campaigns where exploration is acceptable. Avoid it on brand, competitor, and tightly targeted high-CPL campaigns.
Will AI Max replace Performance Max?
Not yet. PMax is a separate campaign type with different placements; AI Max is a layer on Search campaigns. They will likely converge over time.
How long should we leave AI Max running before judging it?
Six to eight weeks minimum, assuming a clean conversion goal and adequate asset inventory. Less than that and you're reading noise.
Do this week
1. Audit every conversion action used by an AI Max campaign. Redefine any that optimise for unit
volume rather than commercial value.
2. Pull asset performance reports. Identify the bottom-decile assets and replace them, not edit them.
3. Pull search category reports. Flag any category whose conversion rate is below half of the
campaign average — these are candidates for negatives or campaign segmentation.
4. Review which campaigns have AI Max enabled. Reconsider it for brand, competitor, and high-CPL
campaigns.
5. Build a weekly AI Max review cadence — distinct from your manual campaign cadence. Conversion
goal hygiene, asset rotation, category drift. That's the new job.
23.png)
