Is Google AI Max Worth It?

Let's start with what Google won't put in the press release.

An independent analysis of over 250 campaigns found that AI Max delivered a median CPA increase of 16%. A separate 4-month head-to-head test found it cost 128% more per conversion than phrase match. Monks Agency looked at 30,000 search terms and found 99% of AI Max impressions generated zero conversions. A LinkedIn poll of working PPC professionals - people actually running these accounts - found only 16% reported good results.

None of that is in Google's blog post.

What is in Google's blog post: a 263% conversion surge for Royal Canin, 161% growth for Klook, and a headline claim of 14% more conversions at a similar CPA. Those numbers are also real. That's what makes AI Max genuinely hard to call - it's not snake oil, and it's not magic. It lands somewhere messier and more interesting than either.

This is our attempt to give you the full picture.

This is our attempt to give you the full picture.

14%
More conversions promised by Google - for non-retail advertisers with the right setup.

16%
Median CPA increase found in the largest independent study (250+ campaigns).

84%
Of advertisers report neutral or negative results.

22%
Of campaigns in the SMEC study hit anywhere near their original ROAS targets.

Before anything else: AI Max is not a new campaign type. You don't set one up from scratch. It's a toggle - a suite of AI-powered features you layer onto an existing Search campaign. Think of it as Search, but with Google's machine learning turned all the way up.

Google has quietly positioned it as the answer to PMax's biggest complaint. Performance Max is powerful, but it's a black box - you put assets and budget in, conversions come out (hopefully), and you're largely left guessing why. AI Max gives you something PMax doesn't: query-level reporting, URL controls, brand guardrails, and the ability to actually see what your ads matched against.

AI Max is Google's answer to the PMax black-box complaint. Automation with visibility. The question is whether the automation is actually working in your favour.

The three things AI Max actually does:

Search Term Matching: Your keyword list becomes a starting point, not a ceiling. AI Max layers on keywordless targeting - similar to Dynamic Search Ads - finding queries you never explicitly targeted. This is the reach engine, and the source of most of the controversy.

Text Customisation: AI-generated headlines and descriptions pulled from your landing pages and existing copy. As of February 2026, Text Guidelines let you set up to 40 messaging restrictions, which makes this considerably less terrifying than it used to be.

Final URL Expansion: Google dynamically picks the landing page it thinks is most relevant to each query. This one needs careful configuration - if your tracking templates aren't set up compatibly, you'll get 404s.

The Numbers Game: Google's Story vs. Everyone Else's

Google's case studies are real. Royal Canin, 263% more conversions. Klook, 161% conversion value growth in a month. For the right account in the right moment, AI Max clearly works.

The problem is that case studies are selected. No company publishes a case study that says "we tried this and it cost us 128% more per conversion." But those results exist too, and they're more common than the headline numbers suggest.

Mike Ryan at SMEC ran the largest independent analysis published so far - 250+ campaigns, not cherry-picked. The median outcome was a 13% lift in conversion value. Sounds decent, until you see the other side: median CPA went up 16%, and only 22% of campaigns landed near their original ROAS targets. The other 78% either dramatically overperformed or dramatically underperformed. That's not a story of AI precision. That's high variance.

The SMEC finding that sticks: only 22% of campaigns hit anywhere near their ROAS target. The other 78% were either a windfall or a disaster. For agencies managing client budgets, 'high variance' is not a selling point.

Why CPA Goes Up Even When Conversions Go Up

This is the part that trips people up. More conversions sounds good. But AI Max gets those extra conversions by reaching further - into queries that are less intent-rich, less proven, and less efficient than the ones your keyword list already covered.

Google's automation uses a blended CPA. Your first batch of conversions might come in at £30 each. The next batch costs £80. The batch after that, £150. As long as the blended average stays within your target, the system keeps spending. But you're not paying the average for the last conversion - you're paying £150 for it. That's the math that doesn't appear in the dashboard.

There's also a specific culprit worth naming: the Search Partner Network. One campaign in the SMEC data had half a million monthly impressions on Search Partner Network sites - at a 0.07% conversion rate. Compare that to 3.04% on standard Google Search. The AI was technically staying within CPA targets by blending those together. The spend, however, was largely wasted.

So Who Does It Actually Work For?

Not everyone. Probably not even most people. But there's a real segment where AI Max makes sense, and being honest about that is more useful than either blanket enthusiasm or blanket dismissal.

Where AI Max Has a Real Shot

The strongest use case is B2B, professional services, and high-consideration purchases - verticals where the AI's intent modelling has room to find genuinely valuable queries your keyword list missed. If you're a managed IT provider or a specialist recruiter and your campaigns have plateaued on phrase and exact match, AI Max's reach expansion is precisely the problem it was built to solve.

The other strong signal: accounts that are already conversion-rich. 100+ monthly conversions, Enhanced Conversions enabled, clean negative keyword lists. The AI needs data to learn, and it learns faster and more accurately when the signal is strong. Thin accounts with patchy tracking don't give it enough to work with.

Where It's a Bad Idea

Retail and ecommerce: Notice that Google's headline 14% figure specifically excludes retail. That's not an accident. PMax with a product feed is still the dominant stack for ecommerce, and AI Max doesn't change that equation.

Tight budgets: The learning phase costs real money. If a client has a strict CPA target and under £1,500/month in spend, there's no room for the variance that comes with AI Max finding its feet.

Messy accounts: Nearly half of advertisers testing AI Max are simultaneously running Dynamic Search Ads, Performance Max, and Broad Match. If your account already has that much overlap, you won't be able to read what AI Max is actually doing. Fix the architecture before adding another layer of automation.

Bad tracking: The AI optimises toward whatever signal you give it. If your conversion tracking is incomplete, misconfigured, or counting the wrong events, AI Max will optimise toward those wrong events - efficiently and at scale.

How to Actually Test It (Without Getting Burned)

If you've decided AI Max is worth exploring for a client, the cardinal rule is: never enable it directly on a live campaign. Always run it as an experiment first. This isn't optional caution - it's the only way to get a read on whether it's working or just cannibalising conversions that would have happened anyway.

The Experiment Setup

1. Create a campaign experiment in Google Ads (Experiments → Drafts & Experiments). Split traffic 50/50.
2. Configure controls before anything else: URL inclusions and exclusions (block homepages and blog posts), Text Guidelines with your brand guardrails, brand inclusions/exclusions for query matching.
3. Run for at least 4–6 weeks. Shorter than that and you're reading noise, not signal.
4. Enable the ai_max_search_term_ad_combination_view report from day one. You need to see what queries, headlines, and URLs are being matched - not just the aggregate numbers.
5. Segment Search Partner Network separately. This is where efficiency goes to die if you're not watching it.

One thing worth flagging to your teams: AI Max is not supported in Google Ads Editor. Attempting to manage it through Editor causes errors. If your workflow relies on Editor, build that into your process before rollout.

What to Measure at the 4-Week Mark

Compare experiment vs. control on conversions, CPA, and ROAS - but don't stop there. Look at conversion quality. Are the new conversions from AI Max actually closing, or are they softer signals that inflate the count without improving revenue? And look at search term overlap: is AI Max finding genuinely new territory, or just bidding into queries your existing campaign already covered?

Set a concrete benchmark before you start. If your non-AI Max campaign runs at a £50 CPA, that's your standard. Don't let the experiment slide to £65 or £70 because the conversion volume looks nice. Volume without efficiency isn't growth - it's spending more to stay in the same place.

The Decision Framework

Rather than another list of bullet points, here's the honest version of how we think about this:

The Verdict

AI Max is a real step forward in Search automation. It's not vaporware, and the conversion uplifts Google cites aren't fabricated. But the independent data is consistent enough that we can't pretend it's a straightforward win either: higher CPA, high variance, and a majority of advertisers not seeing the gains they expected.

The agencies that will get the most out of AI Max in 2026 are not the ones enabling it everywhere because Google said so. They're the ones treating it like any other channel test: clear hypothesis, controlled experiment, concrete benchmark, honest read of the results.

For the right client - B2B, solid conversion volume, a CPA target with some flex - it's genuinely worth testing. For the wrong client, you're handing Google more of your budget and hoping the algorithm figures it out.

The test that matters isn't whether AI Max can work. It clearly can. The test is whether it works for your specific account, with your specific goals, at your specific budget. That's a question no case study can answer. Only your experiment data can.

The Bottom Line

AI Max isn't a scam, and it isn't magic. It's a tool with a real use case — B2B and services accounts with healthy conversion volume and some CPA flexibility — and a long list of situations where it quietly drains budget while the dashboard tells a happier story than your P&L does. The gap between Google's case studies and the independent data isn't dishonesty on Google's part; it's just what happens when a platform's incentives and your incentives aren't quite the same thing.

The only way to know which side of that gap your account falls on is to test it properly: controlled experiment, real benchmark, four to six weeks, no shortcuts. Most accounts we've seen skip straight to "just turn it on" and then spend months trying to figure out why performance drifted.

If you'd rather not run that experiment blind, Search Ally Agency can set up and monitor the AI Max test properly — controls, guardrails, and a clear read on whether it's actually earning its place in your account or just eating budget. Get in touch with our team to find out if AI Max makes sense for your Search strategy.