Your Search campaigns are bringing in business at a cost you can live with. There are things you'd like to improve, but you aren't looking for a major change.
Then your Google rep recommends AI Max. They have a success story to share, and they want to know when you'll try it.
You ask around. One advertiser says it helped. Another spent months cleaning up irrelevant search terms. You're responsible for explaining the decision to the business owner, and neither experience tells you what will happen in your account.
I'd start by asking what you want this campaign to do better. A worthwhile test needs a question you care enough about to spend money answering.
What would AI Max improve in this campaign?
A working campaign can still have room to grow. Maybe the sales team has capacity for more enquiries, but your current targeting isn't bringing enough of them in at an acceptable cost.
That's a reason to investigate broader reach. You can describe the opportunity and decide what additional business would make the experiment worthwhile.
If you're already using the available budget to bring in the work the business can handle, the case for testing needs more thought. What would you gain, and where would the test money come from?
AI Max can expand matching beyond your existing keyword coverage. It also includes settings for generating ad text and choosing landing pages. Those capabilities are explained in Google's AI Max documentation. Whether any of them helps your campaign is the question you're evaluating.
Here's how I'd work through the recommendation before agreeing to a launch date.
How to decide whether an AI Max test is worth running
1. Give the proposed test a business question
Suppose you advertise a commercial cleaning service. The campaign produces regular enquiries, and the business wants more contracts for office cleaning. Enquiries about one-off domestic jobs have little value to it.
For this example, the question could be whether broader matching brings in additional office-cleaning opportunities at a cost the business can support.
That gives you something specific to discuss with the rep. Ask why they think AI Max could reach those customers and which settings they'd propose changing. If they share a case study, look at what the advertiser sold and what the reported conversion actually measured.
An increase in online purchases for a retailer offers limited evidence about enquiries for a local service. It might suggest an idea worth testing. You'll still need results from your own account before making a performance claim.
If you can't identify an improvement the business needs, you can leave the proposal open until there's a useful question to investigate.
2. Describe the configuration you'd actually be testing
Two advertisers can both say they tried AI Max while using different settings. Ask for enough detail to understand the proposed change.
Search term matching affects which searches can trigger your ads. Text customization affects the copy Google can generate. Final URL expansion can send someone to a different page on your website. Google provides controls for these features, so review their proposed states individually.
For the cleaning business, a destination change deserves attention if the site also promotes domestic cleaning. You'd want to understand the permitted pages before spending money on the test. Google's Final URL expansion guidance explains the destination controls and their interaction with ad text.
You might have a reason to test matching while retaining your existing copy and destinations. The operator should check whether the intended setup is available for that campaign. If several features change together, the result tells you about that combination. It won't establish which individual feature caused the difference.
Save the current settings so you have a record to compare against later.
3. Agree what would count as a worthwhile result
For our cleaning example, a submitted form is an early signal. Someone still needs to establish whether the enquiry concerns an office the business can serve.
Consider these invented figures. One comparison group spends $1,800 and produces nine suitable office-cleaning enquiries. The other spends the same amount and produces six. That's $200 per suitable enquiry compared with $300.
The second group could have more total forms and still produce fewer opportunities the business wants. Keeping the qualified count beside the spend makes that visible.
Before launching, agree how enquiries will be classified and when enough follow-up will have happened to compare them. If qualification takes two weeks, yesterday's forms will still have unanswered questions. Compare outcomes with enough follow-up in both groups.
Then discuss the cost the business can accept and the amount it's willing to spend learning. Those figures come from the business's economics. A generic percentage increase in CPA won't tell you whether an extra office-cleaning contract is worth acquiring.
The spending allowance also needs someone monitoring it. Writing an amount in your test plan doesn't create a platform hard cap. Decide who reviews spend and what circumstances would prompt them to stop the test early.
4. Check how you'd make the comparison
Google has an AI Max experiment flow that divides traffic within an existing campaign. Check the current eligibility rules before choosing that approach. Shared budgets and portfolio bidding, for example, can prevent use of the flow.
The setup enables matching, text customization and Final URL expansion by default, so inspect the selections. Also check the option that automatically applies favorable results if you want to review the decision yourself. Brand controls added during setup affect both comparison groups, which matters when interpreting the result.
If the intended experiment isn't available, ask the person managing the account to explain the alternative and its limits. Comparing this month with last month leaves room for changes in demand to influence the result. A simultaneous website redesign would add another explanation.
You can still learn from an imperfect comparison. Record what changed and be honest about what the result can establish.
What to tell the rep before committing
By now, you should be able to explain why a test is worth proposing, or what's missing before you can make that decision.
For the cleaning business, the next action might be to have sales check how reliably it classifies office-cleaning enquiries. Set a review date with that person before discussing a test start date with the rep.
If you want somewhere to work through your own campaign, Before You Say Yes to AI Max is my free guide with a reusable decision brief. Bring the completed brief to your next account review so you can discuss the actual proposal.