A global leader in SEO, content marketing, and data analytics, Anastasia Braitsik is our go-to Digital Marketing expert. With Google’s push towards automation, advertisers are facing a critical new tool: AI Max. It promises to find new conversions by looking beyond keywords, but as Anastasia cautions, it can be a huge money pit if not handled correctly. Today, we’re diving deep into her pre-test checklist, exploring the foundational work necessary for success. We’ll cover everything from the nuances of conversion tracking and budget allocation to the hands-on management of AI-generated assets and landing pages, providing a clear roadmap for anyone considering this powerful, yet potentially volatile, new feature.
An advertiser is considering AI Max but uses a ‘Maximize conversions’ bid strategy and has some unverified conversion tracking. What specific risks do they face with this setup, and what steps should they take with their bidding and tracking to prepare for a successful test?
That’s a combination that practically screams “money pit,” and it’s a scenario I’ve seen cause a lot of frustration. The biggest risk is that you’re essentially giving the AI a broken compass and telling it to find treasure. AI Max optimizes aggressively toward whatever you define as a conversion. If your tracking is inflated, duplicated, or simply not focused on real business outcomes, the system will learn from that bad data and make terrible decisions, chasing phantom goals while your budget evaporates. Compounding this is the ‘Maximize conversions’ bid strategy. It’s designed to get the most conversions possible, regardless of the cost. When you pair that with an AI that’s already expanding its reach, you can see CPAs skyrocket unexpectedly. The AI has no guardrails; it will keep spending to get the next conversion, even if that conversion costs five times what you’re comfortable with. Before even thinking about an AI Max test, the first, non-negotiable step is to fix the conversion tracking. It needs to be precise, deduplicated, and tied directly to what makes the business money. Once that’s solid, I strongly recommend switching to a target-based bid strategy like Target CPA or Target ROAS. My own experiments have consistently shown that AI Max is far more predictable and stable with these strategies because they provide the system with clear performance boundaries.
For a campaign that generates over 30 conversions a month but is often limited by budget, what is the likely outcome of enabling AI Max? Explain the interplay between budget allocation and AI Max’s experimentation, and detail how advertisers should adjust their budget strategy first.
In this situation, enabling AI Max will almost certainly lead to poorer performance. It’s a common mistake to think of it as a magic button to find more conversions within the same budget, but it doesn’t work that way. When a campaign is already losing impressions because the budget is too low, it means you aren’t even fully funding your handpicked, proven keywords. When you turn on AI Max, you’re introducing a massive new variable. The system will start experimenting with broader targeting, and to fund those experiments, it has to pull money from somewhere. That “somewhere” is your existing keywords. So, the budget for your top performers gets diluted even further, their impression share drops, and the core of your campaign weakens. The goal should be to spend as much as you can on what already works, and then use any leftover budget for AI Max to experiment. The first thing that advertiser needs to do is solve the budget problem. They need to ensure their core, handpicked keywords are fully funded and not losing out on impressions. Only after that foundation is secure should they even consider layering on AI Max to explore new pockets of growth.
AI Max treats keywords as broad match and then expands even further. For an account that has historically struggled with broad match performance, what foundational work in ad groups, ad copy, and landing pages is necessary before even considering an AI Max test?
This is a critical point because if you can’t make traditional broad match work, AI Max will be a disaster, plain and simple. It’s like trying to run a marathon when you haven’t even managed to jog around the block. AI Max takes the concept of broad match and puts it on steroids, so any existing weaknesses in your account structure will be magnified. Before even thinking about a test, you have to go back to basics and prove you can successfully run broad match. This often means a significant overhaul. You’ll need to look at reorganizing your ad groups to be more thematically tight, ensuring there’s a strong, logical connection between the keywords, ad copy, and landing page in each one. You’ll have to test new ad copy that speaks directly to the broader intent you’re targeting, not just the exact keyword. And you absolutely must optimize your landing pages to convert traffic that might have a less specific initial intent. Only after you’ve done that work and have seen consistently good results from standard broad match keywords should you earn the right to try AI Max. It’s a graduating step, not a starting point.
Many advertisers create dedicated landing pages for different geographies or campaigns. Given that AI Max has struggled to match the correct landing pages in these scenarios, what is the most effective way to manage the URL expansion feature, and what are the specific URL types to exclude?
The URL expansion feature is a perfect example of a great idea with a currently flawed execution, especially for advertisers with sophisticated setups. I’ve seen this go wrong many times. We had an account where the California campaign was suddenly sending all its traffic to landing pages built specifically for users in Texas—a complete mismatch and a total waste of money. In every single instance where an advertiser has carefully crafted dedicated landing pages for each campaign or ad group, I have yet to see AI Max find a better option on its own. The most effective way to manage this is to be extremely proactive with your exclusions. If you decide to test URL expansion, you must meticulously exclude every landing page that is not relevant to that specific campaign’s geography. Beyond geographic pages, there’s a whole list of URLs most advertisers should exclude right away: help files, support sections, FAQs, blog posts, old A/B test pages you forgot to take down, and any page not specifically designed for conversions or without tracking. You have to give the AI very strict boundaries; otherwise, it will make some baffling and costly choices. For many advertisers right now, the safest and most effective strategy is simply to turn URL expansion off.
Automatically created assets can be a powerful tool for scaling messaging but can also generate non-compliant or off-brand ad copy. Could you walk through the process of using text guidelines and term exclusions to control this feature and share what advertisers must regularly review?
This feature is the one I have the highest hopes for, but it’s also the one that requires the most babysitting. The potential to automatically create customized sitelinks and callouts at the ad group level is immense; it’s a level of granularity many advertisers just don’t have the time to do manually. However, the risk is huge. If you enable automatically created assets, Google can also create new headlines and descriptions, and its track record here is poor. I’ve seen it make promises the brand can’t keep or use messaging that would be non-compliant in regulated industries. To control this, you need to use the text guidelines and term exclusion features—though it’s worth noting they aren’t available in all accounts yet. In the guidelines, you can explicitly state how you want ads to be written, defining the tone, style, and key messages. With term exclusions, you can list words and phrases that should never be used. If you enable this, you must commit to a regular review process. You have to consistently check the new assets Google is creating on your behalf. Is the messaging on-brand? Are the promises accurate? Are the sitelinks pointing to relevant pages? You can’t just set it and forget it; it demands constant vigilance to prevent it from damaging your brand.
When setting up a first-time test, the process of enabling AI Max for just a few ad groups can be slow. Can you provide a step-by-step guide on how to use the Google Ads Editor to efficiently run a limited test on top-performing, non-brand ad groups?
Absolutely. Trying to do a limited ad group test through the web interface is painfully slow and a real headache. The interface forces you to enable AI Max at the campaign level first, and then you have to manually go into every single ad group you don’t want to test and turn it off one by one. It’s incredibly inefficient. The Google Ads Editor is the only sane way to do this. The process is straightforward. First, you open the editor and get recent changes for the campaign you want to test. Then, navigate to the ad group level within that campaign. From there, you can easily sort your ad groups by performance metrics like conversions or conversion volume to identify your top non-brand candidates—the ones with a strong history and, hopefully, proven success with broad match. Simply select only those specific ad groups you want to include in the test. In the settings for those selected ad groups, you’ll find the option to enable AI Max. You just switch it on for that small, selected batch. Then you post your changes back to the account. This approach is not only faster but also much less prone to error. You are actively opting in your best ad groups, rather than trying to opt out all the others.
What is your forecast for AI Max?
My forecast for AI Max is one of cautious optimism. Right now, it’s very much a new product, and like most Google Ads products at launch, it can perform poorly and feel half-baked. Many accounts have tried it and only found failure. However, Google has a long history of this exact pattern: they release a powerful but flawed automation tool, gather years of data, listen to advertiser feedback, add more granular controls, and eventually, it becomes an indispensable part of the toolkit. I believe AI Max will follow that same path. Its potential to save time and uncover new conversion paths is undeniable. The ability to move beyond a rigid, keyword-only structure is the clear direction search is heading. But we’re not there yet. For the foreseeable future, it will require advertisers to be incredibly diligent—to “babysit” it by constantly reviewing search terms, landing page placements, and auto-generated assets. It is not a solution for every account, and for many, the ad writing is still quite poor. But as the AI gets more refined and the controls get better, I predict it will become a highly successful and widely adopted feature. The key for advertisers is to remember it’s a tool, not a replacement for strategy, and to test it carefully and methodically, starting with the solid foundation we’ve discussed.
