Google AI Max: how to open this new black box and adapt your Google Ads structures

AI Max is not a break: it is the culmination of an automation Google has been orchestrating since 2016. The real question is no longer whether the direction is right, but how to keep control within this new frame.

The convergence is no surprise: it was programmed

For several years Google has been running two transformations in parallel, and they are coming together.

On the user side, search is no longer what it was. AI Overviews answer before you even click. Google Lens identifies a product from a photo taken in the street. Voice search, generative results, multimodal: the interface between the user and information is being deeply reinvented.

On the advertiser side, Google Ads has followed exactly the same path, in carefully calculated stages:

• Smart Bidding: the algorithm takes over the bids.
• DSA (Dynamic Search Ads): the algorithm generates ad headlines from your site.
• Broad Match pushed further: the algorithm extends query targeting well beyond your keywords.
• Performance Max: the algorithm handles creative, placements and targeting across Google’s whole inventory.
• AI Max: the algorithm takes over the entire Search chain — targeting, creative, destination.

AI Max is not a break. It is the logical culmination of a gradual automation Google has been orchestrating since 2016. The question is no longer whether that direction is right — it is settled. The real question is: how do you keep control within this new frame? In 2019 we were still writing “Do not let Google optimise your bids”: that advice has had its day.

“Users can now phrase their searches as they would in a conversation: ask a complex question, bring context, set out their needs and follow up on the answers they get.”

That shift on the user side is the same one pushing Google to automate the whole chain on the advertiser side — AI Max is its expression on Search campaigns. Google showed us the future of Search

Under AI Max’s bonnet: what really changes

AI Max for Search campaigns is a set of features you can switch on in your existing campaigns. Three main blocks:

1. Search term matching This is the most structuring feature. Google runs your ads on queries it judges relevant to your assets and your site — even if those queries hold none of your keywords. We move from classic Broad match to something we might call “intent matching”: the algorithm interprets the overall sense of your assets to decide where to appear.

2. Text customization (asset personalisation) The AI generates and adapts your ads’ headlines and descriptions in real time, against the context of the query, the user and the landing page. That feature absorbs and replaces the old automatically created assets (ACA).

3. Final URL expansion The algorithm can send users to the page of your site it judges most relevant — not necessarily the one you set. A logic inherited from DSA, which AI Max builds in natively.

The point that changes everything compared with PMax: modularity.

Unlike Performance Max — where you give Google all or nothing — AI Max lets you switch each feature on independently. You can, for instance, switch on Search term matching alone while leaving the automatic generation of text assets off. That granularity is a real opportunity for advertisers who want to test gradually, without losing control of their branding or their message.

“We have to trust the black box of Google’s algorithm, and we therefore lose visibility.”

That observation, made about Performance Max, applies almost as it stands to AI Max — except that modularity changes the game: here you can choose how far to delegate, rather than give everything or keep everything. How to open the black box of Performance Max campaigns

What the field teaches us

Here are the three realities we see in the accounts that have switched AI Max on.

The quality of the tracking settles everything

AI Max optimises towards the objective you give it with formidable precision. That is where everything is decided.

• Optimising on a TOFU objective — filling in a first-contact form, signing up to a newsletter: volume rises, costs per action look attractive, but quality degrades noticeably further down the funnel. The algorithm does what it is asked — it generates micro-conversions, not customers.
• Optimising on a BOFU or value-based signal — a qualified quote request, a purchase, a target ROAS: the search terms AI Max generates often outperform the account average. The algorithm finds queries you would not have thought of, but that genuinely convert.

The rule is simple: if your tracking is imprecise or poorly aligned with your business reality, AI Max will amplify the problem. If your tracking is solid and aligned with what really creates value, AI Max can become a real lever.

The “broader than broad” warning

With Search term matching, you will run on queries sharing no keyword at all with your account. That is no longer Broad Match — it is semantic, intent-based matching driven by AI.

The direct consequence: negative keyword lists become your main lever of control. Without rigorous discipline on exclusions, you end up appearing on queries that have nothing to do with your offering, at a real cost and with no immediate visibility in the standard interface.

Review your negative lists before switching AI Max on. And audit them regularly afterwards.

Creative friction: how far do you let Google take the wheel?

Automatic asset generation is powerful on paper. In practice it means giving Google editorial freedom over your ads — headlines, hooks, and potentially the pages you send your visitors to.

For brands with firm constraints on their tone, their positioning or their visual identity, that freedom is hard to accept. And rightly so. Switching asset personalisation on without guardrails means risking off-brand messages running at scale.

The right approach: start with Search term matching alone, measure the results, then decide whether the potential gain from asset generation justifies the loss of control over the message.

The restructuring it calls for: the end of thematic siloing

AI Max changes the logic of account structure — and that is probably the deepest impact for teams running complex accounts.

Why ultra-segmented siloing becomes counterproductive

The traditional structure — ten non-brand campaigns, segmented by theme, by intent, by product — rests on the idea that precise targeting comes from precise segmentation. With AI Max that logic is reversed: the more you fragment the signals, the less data the algorithm has to learn from. Each extra campaign dilutes the volume of conversions available and slows the optimisation.

The ideal structure for AI Max

The aim is to consolidate, not to fragment:

• One Brand campaign — to protect your brand terms with bids and messages you control. That campaign is not meant to carry AI Max: it is your safety net, and it stays on exact or phrase match.
• One Generic campaign per language — with AI Max switched on, optimised on BOFU or value signals, and fed by robust negative lists.

That simplification may look aggressive. It is. But it matches the reality of what the algorithm can absorb and process effectively.

The accounts that will suffer

Those that keep a hyper-segmented structure with AI Max switched on will create internal competition between their own campaigns (cannibalisation), scatter the conversion signals, and try to steer by hand a mechanism designed to work consolidated. The predictable result: mediocre performance, wrongly blamed on the tool.

The new sandbox Google imposes

We have lived through this scenario before, with Performance Max. Google first presented PMax as an opportunity. Then the automatic migrations arrived. Then standard Shopping campaigns were gradually deprioritised. The pattern repeats with AI Max: automatic migration notices are already planned for eligible accounts. Google has since announced that AI Max will extend to Shopping and Travel campaigns: our reading of Google Marketing Live 2026.

The conclusion is nuanced: AI Max is a genuinely effective tool — provided you do not hand it the keys without guardrails. The advertisers who will succeed with AI Max are those who have:

• Simplified their account structure, so as to concentrate the signals.
• Strengthened their negative lists, so as to frame Search term matching.
• Aligned their conversion objectives with bottom-of-funnel signals, representative of their business reality.
• Chosen consciously which features to switch on, and which to set aside at first.

Automation is not the performance marketer’s enemy. But automation badly configured, driven by signals too high in the funnel in an account too fragmented, will produce exactly what you fear: volume without value, and a black box you do not hold the keys to.

The tool is there. The structure and the tracking are your responsibility.