Google AI Max proves its value when it adds conversions you would not have won with your existing keywords, at a cost per conversion or ROAS your business accepts. Measuring that requires a controlled experiment, reporting on which matches and assets came from AI Max, and checks on query relevance and landing page choice. Platform totals alone cannot show whether the feature added value or simply took credit.
What exactly are you measuring?
AI Max for Search campaigns bundles three behaviours: search term matching beyond your keywords, text customisation of headlines and descriptions, and final URL expansion to other pages on your site. Each can help or hurt independently. A good measurement plan looks at all three, not just the campaign total.
The core business question is incrementality. If AI Max shows 200 conversions but 150 of them would have come through your phrase match keywords anyway, the real gain is 50. The design of the test decides whether you can see that.
How should you design the experiment?
Google Ads lets you run campaign experiments that split traffic between the original campaign and a version with AI Max switched on. This is the cleanest method available inside the platform.
- Choose a campaign with steady volume; tiny campaigns will not reach significance in a reasonable time.
- Use a 50/50 split and keep budgets, bids and ads the same in both arms.
- Run for at least four weeks, and longer if conversions are delayed, as in B2B lead generation.
- Avoid launching during big seasonal changes, such as Black Friday, unless that is what you want to measure.
- Write down the success criteria before you start: for example, at least 10% more conversions at no more than 5% higher CPA.
Which KPIs should you track?
Outcome metrics
- Conversions and conversion value per arm.
- Cost per conversion and ROAS per arm.
- New customer share, if you report it through customer acquisition goals.
- Lead quality, measured with offline conversion imports from your CRM, for lead generation accounts.
Diagnostic metrics
- Share of conversions from AI Max matched queries versus keyword-matched queries.
- Conversion rate of expansion queries compared with keyword queries.
- Share of clicks sent to expanded final URLs and their conversion rate.
- Performance of customised text assets against your own headlines.
- Brand query share in the test arm, to catch brand traffic inflating results.
Google Ads reporting includes a match type or source view that shows which search terms came from AI Max. Use the search terms report with that segment, and the landing page report, to see where the extra volume goes.
How do you judge query quality?
Numbers can look fine while queries drift away from your business. Once a week during the test, export the AI Max search terms and tag them into three groups: relevant and new, relevant but already covered, and irrelevant. A simple sheet is enough.
For example, a language school advertising ‘English courses Istanbul’ might find AI Max matching ‘IELTS preparation Kadıköy’ (relevant and new), ‘English course Istanbul price’ (already covered) and ‘English teacher jobs’ (irrelevant). The irrelevant group becomes negative keywords; the new group may justify new ad groups with tailored landing pages.
How do you measure text customisation and URL expansion separately?
Once the main experiment has a result, a second round can isolate the parts. Keep search term matching on and run one period with text customisation switched off, then compare asset-level CTR and conversion rate with the earlier period. Do the same for final URL expansion. This tells you which component carries the gain. In some accounts matching drives most of the improvement while generated text adds little, which is useful to know if your brand team dislikes the generated copy.
What should the dashboard look like?
Build it in Looker Studio using the Google Ads connector, plus a sheet for your manual query tags and CRM data if relevant.
- Summary: conversions, CPA and ROAS for control versus AI Max, with the percentage difference.
- Source split: conversions and cost from AI Max matches versus keyword matches.
- Landing pages: top expanded URLs with clicks and conversion rate.
- Query quality: weekly counts of relevant-new, covered and irrelevant terms.
- Change log: negatives added, URL exclusions, text customisation on or off.
The change log matters more than it seems. Without it, a jump in week three cannot be explained.
What are common measurement mistakes?
The first is ending the test too early because the first week looks good or bad. Smart Bidding needs time to adjust to new traffic. The second is judging on platform ROAS alone when conversion value includes brand searchers who would have bought anyway. A third is changing other settings mid-test, such as budget or bid targets, which breaks the comparison.
Finally, remember conversion lag. If customers take two weeks to convert, the last two weeks of the test are incomplete. Wait before reading the final numbers.
How Alien Road applies AI Max measurement
In a typical engagement we agree success criteria with the client before any change, written in the test brief. We then check conversion tracking, including enhanced conversions and CRM imports, because a test on bad data produces a confident wrong answer.
During the experiment our team tags search terms weekly, adds negatives and URL exclusions as needed and records each change. We share a short weekly note and a full report at the end, comparing the arms against the agreed criteria.
The recommendation is one of three: apply AI Max to the campaign, apply it with text customisation or URL expansion switched off, or keep the original setup. Each option is backed by the dashboard pages above.
Where to start this week
Write a one-paragraph test brief for your largest non-brand Search campaign: the success criteria, the duration and what you will do if the test wins or loses. Share it with whoever approves budget.
Then set up the Looker Studio page with the summary and source split before launching. Having the dashboard ready on day one means you will read the results the same way at the end as you planned at the start.
Keep reading: Google Advertising · AI Advertising Optimization