How to test AI Max on a B2B Search campaign

Google’s AI Max for Search is a set of targeting and creative features that you turn on inside an existing Search campaign. It is not a separate campaign type. Google describes three core features: search term matching, which expands reach with broad match and keywordless technology; text customization, which writes ad copy tailored to the search; and final URL expansion, which can send visitors to a different page on your site when Google expects it to perform better.

For a B2B advertiser, each of those can help or hurt. Extra reach is only worth paying for if it produces leads that sales accepts, at a cost the business can afford. This guide shows how to set up a test that answers that question with your own numbers, and how to read the result without fooling yourself.

Check the basics before you test

A test on a weak foundation tells you very little. Confirm these first:

• The campaign is not limited by budget. Google’s help page warns that AI Max “won’t be effective if campaigns are limited by budget,” and it shows an alert when a budget-limited campaign turns it on.
• Conversions reflect the outcome you want. If the campaign optimizes for form fills, AI Max will find more form fills. If you can import qualified leads, it can learn from those instead. The guide on measuring what a qualified lead costs covers how.
• Dynamic landing pages work with your tracking. Google advises checking that dynamically chosen URLs work with your tracking templates, and using compatible {lpurl} patterns so visitors do not land on 404 pages.
• You have a baseline. Pull recent spend, submitted leads, qualified leads and the usual time sales takes to qualify them. Without a baseline, you cannot say whether anything improved.

Start where you can read the result

Pick a campaign with a clearly defined offer, a focused landing page and enough history to establish a baseline. Avoid a campaign that is already unstable, such as one with recent budget swings, a new landing page or a tracking change.

Check volume honestly. If only a handful of qualified leads arrive each month, two test arms will not separate in a few weeks. In that case, set a longer evaluation window and treat the result as directional, or use an earlier outcome that happens more often, such as a meeting booked, as a secondary measure.

Decide what each control should do

AI Max is a bundle, and the controls decide how much freedom each part gets. Write down your choice for each one before launch so that the test is repeatable.

Keep the test focused

1. Write one hypothesis, such as “expanded search matching can find new enterprise buyers at a comparable cost per qualified lead.”
2. Use a control and a treatment where available. Keep budget allocation, geography and conversion definitions comparable, and record every setting that differs between the two arms.
3. Before launch, review brand and location controls, negative keywords, URL restrictions and which copy or pages automation can use. For an offer with distinct consumer and enterprise pages, review the landing page controls first.
4. Decide the evaluation window in advance. Google’s experiments guidance recommends running a test for at least 4 to 6 weeks, longer when conversions are delayed, and waiting one to two conversion cycles. B2B sales cycles are often longer than that, so plan for it.
5. Avoid other changes to the campaign during the test. If you must change something, log the date and the reason.

What to review while it runs

Do not wait until the end to look. Check weekly for problems that would justify stopping early:

• Search terms. Google’s search terms report shows “AI Max” as a match type for incremental terms, with a source column that distinguishes broad match expansion from keywordless matching. Review spend on irrelevant terms.
• Landing pages. The landing pages report has a “Selected by” column that shows how pages chosen by AI Max perform. Look for traffic going to pages that do not match the offer.
• Lead quality. Compare how qualification differs by query theme or geography, once sales has reviewed enough leads.
• Spend pacing. Check that both arms spend at a similar pace, otherwise the comparison is unfair.

Use the same decision rule for both arms

Compare qualified leads, cost per qualified lead and the share of leads accepted by sales. Give the most recent leads time to mature before judging. If the experiment brings more qualified leads but at a higher cost, compare the marginal cost with what the business can pay for incremental pipeline.

Here is an illustrative example. Both arms spend $6,000.

Lead volume rose by a quarter, but qualified leads rose by one. The cost per qualified lead is slightly better, and the qualification rate is lower. With numbers this small, the honest reading is “not clearly better.” Extend the test, tighten the controls or look at which query themes produced the extra qualified lead before scaling.

Common mistakes

• Judging on lead volume or cost per lead, which AI Max can improve while lead quality falls.
• Letting a budget-limited campaign run the test.
• Changing the landing page, bidding or budget in the middle of the test.
• Declaring a result before sales has reviewed the leads.
• Leaving final URL expansion unrestricted when the site has pages that should not receive paid traffic.

A simple rollout plan

1. Before launch: confirm tracking, choose one campaign, record the baseline and write the decision rule.
2. Weeks 1 to 6 or longer: run the test, review search terms and landing pages weekly, and log any changes.
3. Review: compare qualified leads and cost per qualified lead after sales has reviewed the leads.
4. Scale in steps: if the result holds, extend to one more campaign rather than switching everything on at once.

Platform references are listed below. Google’s features and controls change often, so check the current help pages before you launch.

Next step: Calculate your current cost per qualified lead.