Google quietly rolled AI Max out to Shopping and Travel campaigns at the end of April, and Dynamic Search Ads are being phased out in the same breath. The replacement is here, and the question every PPC manager is now being asked is the same one: should we turn it on?
For most accounts, my answer is no, not yet. The reasons aren't ideological. They come from running real tests across a portfolio of accounts and watching the same pattern emerge across the agency community. The headline summary is straightforward. AI Max generates more impressions but not more performance, and the use case it most cleanly fills is one most modern accounts have already covered through other levers.
What AI Max actually does
AI Max sits at the intersection of dynamic targeting and generative ad copy. It crawls your landing pages, extracts the products, services, and language you use, and then matches search queries to that content automatically. It generates and rotates headlines on its own. It chooses which final URL to send a click to. And it expands beyond what your structured campaigns could reach by interpreting intent rather than relying on the keyword lists you've built.
The pitch is that it's the AI-native evolution of DSA. The reality is that it's a much bigger swing. DSA was a tool you bolted onto an existing structure. AI Max is being positioned as the structure.
DSA had a clear job. AI Max is being asked to do everything.
DSA earned its place by catching long-tail searches that exact and phrase match couldn't reach. Most agencies ran it as a complement, with conservative budgets, capped CPCs, and aggressive negative keyword hygiene. It sat alongside the structured campaigns and filled a known gap. Nobody used DSA to replace their account.
AI Max is being marketed differently. The Google narrative is that it's the smarter, more sophisticated successor that should drive a larger share of your spend. That's a meaningful jump in scope, and the testing data so far doesn't support the jump.
The pattern in testing
Across my own testing and across the agency community, the pattern has been consistent. AI Max generates a step change in impressions. It does not generate a step change in conversions, revenue, or qualified leads.
Higher impressions without performance lift is not a neutral outcome. It inflates the top of the funnel, dilutes whatever quality signals your account is sending Google, and makes it harder to read whether your other levers are still working. If you can't tell whether your existing campaigns are pulling their weight because the impression mix has shifted under you, you've made the account harder to optimise rather than easier.
People I trust who run mid to large accounts are reporting the same thing. The community consensus right now is wait, watch, and run small controlled tests if you must. Almost nobody is restructuring their accounts around AI Max, and nobody I respect is recommending that anyone else do so either.
The real question is whether you already have AI coverage
The strongest argument for AI Max is being visible in AI Overviews and conversational SERP results. That's a real coverage need. AI search is not a hypothetical anymore. Users are running queries that look more like conversations than keywords, and Google is increasingly answering those queries with AI-summarised responses that contain ad units.
The catch is that for most accounts, you already have that coverage.
If you're running Performance Max, you already have AI Overview coverage. PMax serves into AI-generated SERP experiences as part of its standard placement set. If you're running broad match keywords in Search, you already have AI Overview coverage. Broad match has been quietly extended to feed Google's conversational query matching for some time. Most modern accounts run one or both.
So the question stops being "is AI Max ready?" and starts being "do I have a coverage gap that AI Max would fill?" For most accounts, the honest answer is no.
Where AI Max actually fills a gap
There are accounts where AI Max is worth a controlled test today, and the diagnostic is straightforward.
If you're running only exact and phrase match in Search with no PMax in the mix, you have a real coverage gap in AI Overviews. AI Max could fill it.
If you don't have PMax running for whatever reason, whether brand sensitivity, channel separation, or historical preference, you have a real coverage gap. AI Max could fill it.
If you've deliberately kept your account narrow and structured, and you want a controlled way to extend reach into AI-driven SERP placements without committing to a full PMax build, AI Max gives you that lever.
If none of those describe your account, you're paying to duplicate coverage you already have.
The trust problem and the setting most people miss
Even when AI Max does fill a coverage gap, the next question is whether you're comfortable with what it's doing.
By default, AI Max generates and rotates headlines for you. You can turn this off via the text optimization toggle, but the default is on. Most accounts that switch AI Max on won't think to check, and Google's generated headlines often don't read like the brand would write them. Clients notice. They start asking why their ad copy says what it says, and the agency answer of "Google generated it" is a conversation neither side enjoys.
The deeper issue is that text optimization is most of what AI Max is doing. Turn it off and you've stripped out a big chunk of the engine. Leave it on and you've handed your ad copy to Google. There isn't a clean middle position, which is why the "is AI Max worth running" question and the "am I comfortable with Google writing my copy" question are really the same question.
For larger accounts with copy review processes and brand teams, the answer is usually that text optimization needs to come off. Once it's off, the value proposition narrows considerably.
Where it makes sense today
There are accounts where AI Max is a reasonable lever to add now.
Large accounts with budget to absorb the impression bloat without disturbing pacing across the rest of the structure.
Accounts that have already maxed their existing levers, where exact, broad, and PMax are all running at the limit of what the budget allows, and the next dollar genuinely has nowhere structured to go.
Brands comfortable letting Google write copy on their behalf, or with the operational discipline to police what gets generated.
If your account fits that profile, AI Max is a tool to test thoughtfully. If it doesn't, the structured stack you already have is doing the job AI Max would be asked to do, and doing it more reliably.
What to do instead
For accounts that don't fit the AI Max profile, the better path is to make the levers you already have work harder.
Maximise exact and broad match in your Search campaigns first. Run PMax with proper asset and audience hygiene. Increase budgets across the structured stack until performance flattens. Only when your existing campaigns have plateaued and you're confident the budget genuinely has nowhere structured to go does a controlled AI Max test start to make sense.
This sequence isn't conservative for the sake of it. It's the same logic that worked for PMax when it launched in 2022. Most accounts that saw the strongest returns from PMax weren't the ones that rebuilt their stack on day one. They were the ones that kept structured Search and Shopping pulling at full strength for a year, and then integrated PMax as a controlled addition rather than a replacement.
DSA is being replaced. AI Max is the replacement. That doesn't mean switch.
DSA isn't being replaced in spirit. The functionality of dynamic targeting is moving inside an AI Max wrapper. The intent of catching long-tail searches that structured keywords miss is going to live on in some form, and over the next year or two it will become the default approach.
That's not the same as the answer being yes today. For most accounts, the question right now isn't "is AI Max good or bad." It's "do I already have the coverage AI Max promises, and if so, what is AI Max actually adding?" For most accounts, the answer to the first half is yes, and the answer to the second half is not enough to justify it. Yet.