Google’s AI Max for Search: What 30 days of testing reveal
Every few months, Google launches a new feature – and reactions usually split into two extremes: either it’s a welcome but overdue correction or a sign of PPC end times.
AI Max is no exception. Some advertisers love what it represents and how it’s delivered, while others are ready to dismiss it altogether.
As with most things, the reality falls somewhere in between. Instead of speculating on what-ifs, I’ll share what I’ve seen after testing AI Max for 30+ days across multiple accounts.
AI Max is an optional toggle for your Search campaigns that opens you up to what Google calls “the full potential of Google Search.”
It allows Google to go beyond keyword targeting and show your ads to people that its machine learning algorithms feel might soon need what you’re offering.
Think of how broad match keywords go beyond what people are searching for in the moment to what they might search for in the future.
Now expand that principle further past query-based targeting entirely to cover:
Plus, AI Max will also show your ads for keywords that it determines might be a fit based on the content of your website and landing pages.
In case you missed it, Google has moved brand inclusions/exclusions for Search to AI Max. This means that you can only use these features if you are using AI Max.
I don’t see a problem here. This isn’t Performance Max, where everything is combined into one.
You should already be running a standalone Search campaign for your branded terms.
The ability to run your Search campaigns with or without AI Max enabled separates it from Performance Max – which is its own campaign type covering a variety of ad inventories, including Shopping, Display, YouTube, and Gmail.
Still, they’re both algorithm-driven, so I think about them in similar ways.
While tools like AI Max can amplify the work and results of an experienced PPC manager, not every account is in the right position to benefit from them.
Here are some examples of scenarios where it may or may not be wise to use AI Max.
A good place to start is sitting down with your CMO or client to discuss goals, figure out whether you really need to be testing AI Max at this time, and go over what needs to be done for that testing environment to be successful.
This is a core part of how our team operates and leaves little room for unwanted surprises down the road.
Dig deeper: Why Performance Max for lead generation often fails and how to make it work
Get the newsletter search marketers rely on.
See terms.
In this digital camera account, where Search drives much of the activity, you can clearly see the revenue coming in.
In this client account, which sells digital cameras and relies heavily on Search, you can clearly see revenue values coming in.
While I’m not 100% certain, I’m confident that AI Max (both expanded match and landing page match) uses keywords and your landing page to find new opportunities.
But it doesn’t spend a lot from what I’ve seen: in this case, it’s ~$3,800 across 30 days (including expanded match) from a total spend of ~$51,000.
But it did generate 23 additional orders worth $52,486 – a very solid return of roughly 14:1! While there’s not a lot of volume, performance is better than conventional keyword targeting.
This is why I say it’s a no-brainer to test for ecommerce.
My client now has an extra $52,000 in revenue that we may or may not have generated without AI Max.
To expand that thinking to Google’s algorithmic campaigns in general, that incremental boost may have come from Performance Max if I hadn’t run Search.
To think of AI Max as either a PPC apocalypse or search on autopilot is completely the wrong mindset.
First, it’s a 100% optional setting and doesn’t absorb existing campaigns like Performance Max did. So, if you’re really set against AI Max, just don’t use it.
Second, worrying that AI Max is another step toward fully automated Google Ads doesn’t benefit anyone.
Don’t panic yourself out of a job, because human marketers are still needed to:
Dig deeper: AI Max experiments arrive in Google Ads: Here’s how they work
It’s still too soon to definitively judge AI Max, but the early signs are positive – possibly more so than they were for Performance Max.
The main things this 30+ day test taught me are:
AI Max is new and still in beta, so it will be a while before we have a product that works the way Google intended.
As always, reserve a portion of your budget for experimentation and test cautiously so that you don’t overextend and find yourself having a difficult conversation with clients and bosses.
Continue reading...
Every few months, Google launches a new feature – and reactions usually split into two extremes: either it’s a welcome but overdue correction or a sign of PPC end times.
AI Max is no exception. Some advertisers love what it represents and how it’s delivered, while others are ready to dismiss it altogether.
As with most things, the reality falls somewhere in between. Instead of speculating on what-ifs, I’ll share what I’ve seen after testing AI Max for 30+ days across multiple accounts.
Google’s new AI Max feature explained
AI Max is an optional toggle for your Search campaigns that opens you up to what Google calls “the full potential of Google Search.”
It allows Google to go beyond keyword targeting and show your ads to people that its machine learning algorithms feel might soon need what you’re offering.
Think of how broad match keywords go beyond what people are searching for in the moment to what they might search for in the future.
Now expand that principle further past query-based targeting entirely to cover:
- Past search history.
- Demographic and behavioral attributes.
- Historical search intent.
- All the other user measurements Google has access to, but that we can’t see.
Plus, AI Max will also show your ads for keywords that it determines might be a fit based on the content of your website and landing pages.
Reminder: Brand controls for Search are exclusive to AI Max
In case you missed it, Google has moved brand inclusions/exclusions for Search to AI Max. This means that you can only use these features if you are using AI Max.
I don’t see a problem here. This isn’t Performance Max, where everything is combined into one.
You should already be running a standalone Search campaign for your branded terms.
Key differences between AI Max and Performance Max
The ability to run your Search campaigns with or without AI Max enabled separates it from Performance Max – which is its own campaign type covering a variety of ad inventories, including Shopping, Display, YouTube, and Gmail.
Still, they’re both algorithm-driven, so I think about them in similar ways.
- Spam is likely to be heavy when using form submissions unless there’s a filter to minimize that, so I prefer to use them for calls only. Conversely, this is also part of what makes AI Max a no-brainer for ecommerce, which is less prone to spam.
- They’re both prone to matching with some random queries that may not be great for lead quality, so even legitimate inquiries from real people might be unqualified and looking for something else due to the nature of query matching.
- Overall, I don’t really want to use them for lead-gen unless I have qualified data coming through and passing back into Google through offline conversion imports.
- Unlike DSAs, which pull ad copy directly from your landing pages, AI Max will come up with new copy using its algorithm, similar to how Performance Max auto-generates creative assets.
- You can also exclude Search Partners and Display entirely, unlike Performance Max. This added bit of control is very welcome.
When to use and avoid AI Max in your Search campaigns
While tools like AI Max can amplify the work and results of an experienced PPC manager, not every account is in the right position to benefit from them.
Here are some examples of scenarios where it may or may not be wise to use AI Max.
- For ecommerce, AI Max is worth testing as a source of truly incremental revenue. At worst, you may spend a bit extra during the first month of testing and get no tangible results, but you won’t lose resources beyond those wasted CPCs.
- For lead gen, the only time I’d consider AI Max is if you’re passing back CRM data on qualified or closed leads. In this scenario, where you’re keeping a close eye on the two-way exchange of user data, it can also be quite helpful.
- Mainstream niches and anything with large search volumes (e.g., home decor, consumer goods) might benefit from AI Max targeting expanded opportunities.
- New accounts without historical conversion data should avoid AI Max until they’ve met Google’s manual conversion threshold (30 conversions a month) for at least a few months.
- Small accounts spending up to a few thousand dollars a month should stabilize manual demand capture before turning on any algorithmic features.
- Regulated industries like pharmaceuticals might need to play it safe with who sees their ads and how those ads show up. When in doubt, consult your legal team or client – this is not legal advice.
- Anywhere you would use dynamic search or Performance Max, AI Max is worth considering.
A good place to start is sitting down with your CMO or client to discuss goals, figure out whether you really need to be testing AI Max at this time, and go over what needs to be done for that testing environment to be successful.
This is a core part of how our team operates and leaves little room for unwanted surprises down the road.
Dig deeper: Why Performance Max for lead generation often fails and how to make it work
Get the newsletter search marketers rely on.
See terms.
Ecommerce example: The case for ‘why not AI Max?’
In this digital camera account, where Search drives much of the activity, you can clearly see the revenue coming in.
In this client account, which sells digital cameras and relies heavily on Search, you can clearly see revenue values coming in.
While I’m not 100% certain, I’m confident that AI Max (both expanded match and landing page match) uses keywords and your landing page to find new opportunities.
But it doesn’t spend a lot from what I’ve seen: in this case, it’s ~$3,800 across 30 days (including expanded match) from a total spend of ~$51,000.
But it did generate 23 additional orders worth $52,486 – a very solid return of roughly 14:1! While there’s not a lot of volume, performance is better than conventional keyword targeting.
This is why I say it’s a no-brainer to test for ecommerce.
My client now has an extra $52,000 in revenue that we may or may not have generated without AI Max.
To expand that thinking to Google’s algorithmic campaigns in general, that incremental boost may have come from Performance Max if I hadn’t run Search.
Gaps in AI Max: Where does the human touch play a role
To think of AI Max as either a PPC apocalypse or search on autopilot is completely the wrong mindset.
First, it’s a 100% optional setting and doesn’t absorb existing campaigns like Performance Max did. So, if you’re really set against AI Max, just don’t use it.
Second, worrying that AI Max is another step toward fully automated Google Ads doesn’t benefit anyone.
Don’t panic yourself out of a job, because human marketers are still needed to:
- Make smart decisions around fundamentals like strategy, audiences, and messaging.
- Write RSAs that earn clicks by meeting search intent and showcasing value.
- Build landing and product pages with good UX that get people to convert.
- Create experience loops using email, social media, and other channels (because PPC doesn’t exist in a vacuum).
- Troubleshooting, tweaking, and a thousand other things that AI is far from capable of doing autonomously.
Dig deeper: AI Max experiments arrive in Google Ads: Here’s how they work
Early lessons from AI Max and what to do next
It’s still too soon to definitively judge AI Max, but the early signs are positive – possibly more so than they were for Performance Max.
The main things this 30+ day test taught me are:
- I’d love to see more reporting and analytics for AI Max, like why certain segments of users were targeted. Google is scaling back individual user-level reporting due to privacy concerns, but there’s no reason we shouldn’t see something beyond top-level performance.
- Since AI Max is primarily focused on the Search network and, therefore, works with the most user data, it seems to yield better-quality output than other algorithmic features. But it also doesn’t spend as aggressively as I would have thought.
- The Performance Max mindset definitely helps push you in the right direction with AI Max, and while all account-level data is helpful, there’s no concrete proof (yet) that getting results with one means you’ll succeed with the other.
AI Max is new and still in beta, so it will be a while before we have a product that works the way Google intended.
As always, reserve a portion of your budget for experimentation and test cautiously so that you don’t overextend and find yourself having a difficult conversation with clients and bosses.
Continue reading...