Google’s AI Max for Search has officially entered the chat. But is it really the next big leap forward in search campaign performance Google promised?
Or, just another automated system that swallows your ad budget while providing little to no visibility in the process?
In our latest webinar, we brought together three seasoned marketing experts to share their AI Max findings so far:
Scott Carruthers (Senior Paid Search Director at Journey Further),
Kat Sale (Co-founder at House of Performance), and
Diane Talhelm (Senior Search Specialist at ZATO Marketing).
They’ve all been running hands-on AI Max experiments across B2B, B2C, ecommerce, and lead gen. And they didn’t hold back on the results - both the good, and the underwhelming.
Here’s everything you need to know, including exclusive Lunio data on how AI Max impacts invalid traffic (spoiler: it’s not looking good), plus real-world advice for when, why, and how to use it.
Watch the full session here, or keep reading for the written highlights & key takeaways:
-2.png?width=300&name=AI%20Max%20Webinar%20(1)-2.png)
0:00 - Intro & panel introductions
04:41 - Poll: Do you trust Google's automation?
07:11 - Session Agenda
09:14 - Mike Ryan's AI Max summary
10:57 - Jyll Saskin Gales' AI Max discussion
12:17 - AI Max: Initial results & findings (panel)
19:56 - Other AI Max data & results
23:07 - How to keep AI Max spend under control
28:13 - AI Max traffic quality (exclusive data)
31:21 - How AI Max fits into your account
36:51 - When to NOT use AI Max
38:35 - Capitalizing on AI Max going forward
43:38 - Q&A from the live audience
What is AI Max? (Cutting through the hype)
First, a quick note to clear up any confusion: AI Max isn’t a new campaign type.
It’s an optional feature within standard search campaigns. Think of it as a turbocharged version of Dynamic Search Ads (DSAs) and Broad Match, with Performance Max-style ad generation.
It pulls two key automation levers:
- AI-powered matching - Goes beyond your keywords to match search terms based on intent.
- Final URL expansion & ad customisation - Automatically adjusts landing pages and headlines.
As Mike Ryan aptly put it in a recent LinkedIn post:
"Earlier, in November 2023, Google held focus groups asking questions such as "How do you feel about migrating regular keyword Search campaigns to PMax campaigns?" The response, presumably, was NOT good. This was about a year after Google completed a forced migration of Smart Shopping to PMax, and, seemingly having learned a lesson, they were already offering *voluntary* migration of DSA & GDA campaigns.
That's when someone at Google had a brilliant idea: "what if, instead of bringing Search to PMax, we bring PMax to Search?"
So they pulled the Final URL optimization (DSA-like) and Search Term Matching (keywordless) capabilities from PMax into Search Max. As the crowning touch, riding the wave of AIO and AI Mode hype, they renamed it...
AI Max for Search Campaigns.
Ladies and gentlemen, dear esteemed colleagues: A Star is Born.”
So now we’ve cleared up what AI Max is (and isn’t), how well does it actually perform?
AI Max testing: Real-world results from the panel
All three panellists had been testing AI Max across multiple accounts. Their verdict? It’s not the game-changer many hoped for - at least, not yet.
Kat Sale: “Underwhelming… but in a good way”
Kat’s tests focused on a B2C health client spending £20K/month. AI Max was turned on for a high-volume campaign, and the immediate impact was… nothing too dramatic.
“It spent a little more, and yes, the terms got broader. But it learned quickly that those broader terms didn’t convert, and it reined itself in.”
Kat shared a chart showing a drop in impressions for phrase match, with a slight rise in broad match - but after a few days, things stabilised.
“If I looked at the campaign today without knowing we’d switched it on, I probably wouldn’t notice.”
Kat also skipped the ad customisation feature entirely:
“Most clients want their ads to look a certain way. I’m not in a rush to test a feature that might go rogue with copy.”
Scott Carruthers: “It’s a mixed bag. Six accounts did better ROAS. Six did worse.”
Scott has been testing AI Max across 12 accounts. In his first test, AI Max drove higher impressions but lower CTR, with no meaningful performance lift. Over time, the results varied wildly.
- 6 accounts improved ROAS
- 6 accounts saw ROAS decline
- CTRs generally dropped when AI-generated headlines were used
- Impressions jumped significantly in every case
Scott also ran semantic scoring experiments to compare AI Max search terms against existing keyword intent. The results showed AI Max reaching further - often as far as broad match - but less aligned with existing keyword themes.
Interestingly, when Broad Match was enabled, AI Max’s semantic alignment dropped even further. Why? Because it had to find completely new terms not already covered by Broad.
“AI Max plays in the same area as broad match. If you’re already running broad, expect diminishing returns.”
- Scott Carruthers
Diane Talhelm: “It’s not as expansive as expected. But the DSA-style URL expansion delivers.”
At ZATO, Diane saw similar results. The matching itself wasn’t groundbreaking, but the DSA-like URL expansion seemed to drive most of the value.
“CTRs are improving over time. AI matching isn’t great, but the landing page expansion part does well.”
One newer test using the built-in experiment feature delivered much higher spend immediately. Her takeaway is that results depend heavily on:
- Industry
- Budget flexibility
- Campaign structure
- Existing match types
Controlling spend: Guardrails to stop budget blowouts
One of the biggest concerns with AI Max is letting it run wild and waste spend. But as Scott put it:
“If your biggest concern is spend getting out of control - just don’t run it.”
If you do want to test it safely, here’s what the panel recommends.
1. Exclude the obvious upfront
Following advice from Mike Ryan and others, the panel agreed:
- Exclude brand terms
- Exclude competitors (unless you’re deliberately targeting them)
- Exclude irrelevant categories
- Exclude mobile app placements (especially if you’re seeing IVT spikes)
Kat summed it up well:
“It’s only doing what it’s designed to do. Google has given us the guardrails. Use them before switching it on.”
2. Use page feeds instead of letting it crawl your entire site
Diane flagged that most of the spend tends to go through the URL expansion feature. If your site has lots of irrelevant or broad content, AI Max might pull in blog pages or low-intent areas.
“Build a small, targeted page feed with only the URLs you want it to use. Way more control.”
3. Don’t over-constrain too early
Scott’s view was clear:
“If you’re running AI Max, it’s presumably for exploration. Don’t strangle it with exclusions on day one.”
Start with the basics. Let it explore. Then tighten based on what the data tells you.
AI Max vs. other campaign types: where does it fit?
AI Max isn’t here to replace standard search campaigns - at least not yet. All three panellists saw it as an add-on, not a starting point.
“To me, it’s just another lever. If I was launching a brand new account tomorrow, I wouldn’t start with AI Max. I’d layer it on once search is working” - Kat Sale
Scott added that it’s best used for incremental reach - not core performance. Especially if you’re already running Broad Match.
Diane raised an important point about campaign overlap:
“We saw our DSA campaigns drop when AI Max was enabled. There’s definitely some cannibalisation.”
If you’re using AI Max, carve out space for it by adjusting or excluding from other campaigns.
The IVT problem: AI Max increases exposure to fake traffic
Now for the not-so-great news. Lunio’s own data science team analysed the impact of AI Max on invalid traffic rates (IVT) - and the results are pretty clear.
Key findings from a luxury retail client:
- 404 search campaigns with AI Max
- 491 without
- Baseline IVT rate before: 3.7%
- After enabling AI Max: 5.0% average IVT, with peaks at 6%
- That’s a 35% increase in bot traffic exposure
For brands spending big on Google Search (like £1M+/year), that uplift represents serious wasted spend.
If you’re testing AI Max, we recommend auditing your traffic early. Even if you’re happy with performance, knowing how much fake traffic you’re paying for is essential.
When not to use AI Max
There were a few clear scenarios where the panel advised marketers to steer clear of AI Max:
- You’re already budget-constrained
- If the campaign is already limited by budget, AI Max will just dilute results.
- You’re in a highly regulated industry
- Legal, medical, and financial advertisers often can’t use ad customisation features - which limits the impact of AI Max.
- You have zero appetite for risk
This isn’t the place to play it safe. Let it test, or don’t run it.
Future predictions: is AI Max the new PMax?
So, will AI Max take over Google Search just like Performance Max did for ecommerce?
Scott doesn’t think so:
“Not unless Google kills keywords entirely. But the direction is clear - more automation, less manual control.”
Kat’s advice echoed how many marketers felt about PMax a few years ago: Like it or not - embrace it, or fall behind.
“If we go keywordless, we go keywordless. The best advertisers adapt. That’s what keeps us all in jobs.”
Diane made the strongest call:
“AI Max is the prerequisite for showing up in AI Overviews. If you’re not testing it now, you’ll be late to the party.”
Her suggestion was essentially: don't wait. develop multiple testing strategies now.
Q&A: live audience questions answered
We closed the session with a rapid-fire Q&A, featuring questions submitted by the live audience. Here’s a quick summary of the key points, but make sure to watch the full session for complete context and additional insights:
Q: Do we need to aggressively exclude URLs? What about blogs?
Diane: If you’ve never run ads to blog pages before, exclude them. Otherwise, build a controlled page feed with the URLs you know work.
Q: What industries see the best results?
Scott: Retail and ecommerce - especially if you have lots of products and SKUs. That’s where the auto-copy features can really shine.
Q: Is AI Max going to replace DSAs?
Kat: I haven’t used DSAs in years. With PMAX and AI Max, I don’t see the need anymore.
Diane: I asked Google directly - they said “no plans to phase out DSAs.” But that probably means “not right now.”
Q: Does Broad Match still help if AI Max is on?
Scott: If you’re already running Broad, AI Max won’t add much. It reaches similarly broad terms.
Q: Can AI Max show irrelevant terms, like “dentist ads for tire shops”?
Kat: That hasn’t happened in our tests. But if it does, either exclude those terms or switch it off.
Q: Should we start tight and add exclusions over time? Or start loose?
Scott: If you’re using AI Max, let it explore. That’s the point. But do set some basic guardrails on brand and irrelevant URLs.
Final thoughts
AI Max isn’t magic. It won’t rescue broken campaigns. But when used with the right setup, guardrails, and expectations, it can unlock new performance insights - and maybe even a few surprises.
If you’re ready to test it, test it smart. Watch your queries. Watch your invalid traffic levels. And lean into the automation where it makes sense.
And if you’re seeing suspicious spikes in spend or traffic quality?
Run a traffic audit before your CFO runs the numbers.
Additional AI Max resources:
- Kat’s AI Max experiment results
- Scott’s AI Max experiment results
- Kat - The impact on AI Max on match type
- Scott - The impact of AI Max on match type
- AI Max & Competitor terms - Mike Ryan
- Sophie Logan - AI Max experiment results
- Jyll Saskin Gales - Optimising AI Max
