AI Max Final URL Expansion: Audit, Test, and Control It

Learn how to audit whether Google AI Max's Final URL expansion beats your hand-picked landing pages, isolate the variable in tests, and steer AI routing without losing control.

AI Max Final URL Expansion: Audit, Test, and Control It

Google says advertisers who turn on the full AI Max feature set see an average of 7% more conversions or conversion value than with search term matching alone. That gap comes mostly from one thing: AI Max now picks where your traffic lands. Very few teams test that change on its own.

AI Max landing page selection, which Google calls Final URL expansion, adds a new conversion rate variable to every Search campaign. Here’s how to audit it, test it cleanly, and guide it without switching it off.

Find out which high-intent buyers convert on which pages, so AI routing works with your funnel instead of against it.

See intent data in action→

What Final URL expansion actually changes

For years, the landing page was the one part of a Search campaign you fully controlled. You wrote a final URL, tested it against another, and kept the winner. Final URL expansion ends that. According to Google’s documentation, it sends traffic to the most relevant URLs on your domain when the system predicts better performance, and it only picks URLs that match the query and fit your ad group’s theme.

Three details matter more than most people realize:

  • It’s on by default. When you opt into AI Max, Final URL expansion is checked unless you uncheck it yourself.
  • It rewrites your copy too. Final URL expansion needs text customization turned on, so whenever Google swaps the landing page it can also write a new headline and description to match.
  • It overrides pinned assets. Google’s help page says pinned RSA assets aren’t used when a more relevant URL is chosen. If your compliance team relies on pinned disclaimers, check this today.

The timing is about to get tighter. Google’s update from June 11 says campaigns using automatically created assets and campaign-level broad match will auto-upgrade to AI Max starting in September, while the Dynamic Search Ads sunset moves to February 2027. After the upgrade, former DSA campaigns run all three AI Max features, and Final URL expansion is one of them.

Why Google’s benchmarks can’t answer your question

The early adopter numbers look good. Google reports that advertisers who turn on AI Max typically see 14% more conversions or conversion value at a similar CPA/ROAS. For campaigns still using mostly exact and phrase keywords, Google puts the typical uplift even higher at 27%. In the same announcement, Google says L’Oréal saw a 2X higher conversion rate at a 31% lower cost-per-conversion, while MyConnect drove 16% more leads at a 13% lower cost-per-action.

So does AI routing work? Those numbers can’t tell you. Every one of them mixes three changes into a single before-and-after result: new queries, new copy, and new landing pages. Google notes that MyConnect’s 30% increase in conversions came specifically from net-new queries. A campaign can gain conversions from broader matching while Final URL expansion quietly drags down conversion rate on your best traffic. The blended result still looks positive.

Key Insight

A positive AI Max result doesn't prove AI routing helped. Broader query matching can hide a landing page problem for months, because new volume makes up for weaker conversion rates on your core traffic.

The closest benchmark is Google’s finding that the full feature suite, including final URL expansion, delivers an average of 7% more conversions or conversion value than search term matching alone. Even that bundles text customization in. Your account is the only place to get a real answer.

How to isolate the landing page variable

Google’s one-click AI Max experiments are the obvious place to start, and the wrong tool for this question. Google’s documentation describes the control arm as your campaign with the AI Max toggle turned off and the trial arm with it on. That tests AI Max as a whole package. You can deselect Final URL expansion in the treatment arm, but the control still runs without search term matching, so targeting stays mixed in with the result.

To isolate routing, keep everything else the same in both arms. Here’s the approach we recommend:

A reasonable objection: text customization behaves differently once Final URL expansion is on, because it writes copy to match the new page. That’s fine. In practice, page swap and copy rewrite come as a pair, so you’re testing the bundle you would actually deploy.

One technical check before launch. Google warns that tracking templates without a proper {lpurl} tag can send users to a static URL or a 404 instead of the dynamic page. Test a sample of AI-selected URLs through your template first, or your experiment will measure broken links.

1

Establish a stable baseline:

Run the campaign with AI Max on, search term matching and text customization enabled, and Final URL expansion off, long enough for Smart Bidding to settle.

2

Build a custom experiment:

Create a custom experiment from that campaign where the only change in the treatment arm is turning on Final URL expansion. Both arms then share the same query pool and copy engine.

3

Freeze outside changes:

Don’t touch bids, budgets, keywords, or site templates during the test. Changing a page in the middle of the test breaks your comparison.

4

Read the URL-level data:

Use the search terms report’s combined view of queries, headlines, and URLs, plus the landing pages report, to see which pages the AI chose and how each one converted.

5

Judge by segment, not just the total:

Split results by product category and intent type before deciding. A flat overall result often hides one segment that gained and another that lost.

Which segments benefit, and which need a human override

When you split the data, a pattern usually shows up. AI routing tends to win where your site has many specific pages and the query tells it exactly which one is right. It tends to lose where the “relevant” page and the “converting” page are different pages.

Where AI routing usually helps:

  • Large catalogs with clean, crawlable product and subcategory pages, where a query for a specific product should land on that product and not on a generic category page.
  • Long-tail, question-style queries. L’Oréal’s example query, a question about the best cream for dark spots, matches an informational page better than any keyword-mapped URL would.
  • Sites where every indexable page has a clear call to action and a working conversion path.

Where human override usually wins:

  • Lead gen funnels built around a purpose-made landing page with a short form. AI may pick a “more relevant” service page that has no form above the fold.
  • Pages built for compliance or pricing, where the legally approved page is the only acceptable destination.
  • Sites with thin or outdated content: old blog posts, discontinued products, or location pages that no longer match what you offer.
  • High-intent branded and bottom-funnel terms, where you’ve already tuned the page for conversion.

Here’s the rule of thumb for page structure: AI routing works best when your site’s relevance signals and your conversion paths point to the same pages. If your best-converting pages are noindexed, light on text, or kept separate from the main navigation (which is common for paid landing pages), the AI has little to read on them and will choose your content-heavy pages instead. This is the same issue we cover in our guide on intent targeting migration: the system follows signals, so the signals have to point at outcomes.

Build a feedback loop instead of a cage

Once you know which segments lose, there’s a temptation to turn Final URL expansion off everywhere. Don’t. You’d lose the segments where it wins, and in September Google will push many campaigns back to its defaults anyway. A better approach is to give the system structured feedback, one layer at a time.

URL exclusions at the campaign level. Google lets you exclude specific URLs or write rule-based exclusions. Start with the obvious ones: careers, blog archives, support docs, legal pages, and any URL that converted poorly in your experiment. Rules scale better than lists, for example excluding everything under /blog/ or /support/.

URL inclusions at the ad group level. Google describes inclusions as a way to add URLs Final URL expansion didn’t pick up on its own. Use them to make sure your purpose-built landing pages are in the candidate pool, especially if they sit outside the site’s main crawl paths.

Page feeds with custom labels. For large sites, page feeds labeled by category or margin tier let you exclude or include whole groups of pages at once. When a category loses in testing, you exclude one label and leave the rest of the site open.

Sitelinks as a guaranteed path. Even when the AI swaps the main final URL, your sitelinks still point to the URLs you set. Point them at your highest-converting pages so users always have a direct route there. Compare sitelink click-through and conversion against the main headline to see whether users prefer your page over the AI’s pick.

Page-level conversion tagging. Bidding learns from the conversions you report. If your purpose-built pages fire richer conversion events (qualified leads, not just form views), the system gets a clearer signal about which destinations produce value. This is the same discipline we recommend for setting attribution windows per segment, and it helps avoid the attribution blind spots that cause automated systems to optimize for the wrong outcome.

Review the loop monthly. Pull the landing pages report, flag any AI-selected URL with meaningful spend and a conversion rate below your manual baseline, then either exclude it or improve the page. Sometimes the AI found the right intent and the page just needs a stronger call to action. If you’re comparing autonomous ad tools more broadly, our agentic ad stack guide applies the same “steer, don’t cage” approach.

Your next step: before the September auto-upgrade, run one custom experiment where Final URL expansion is the only difference between arms, on your highest-spend Search campaign. Then use the segment-level results to write your first exclusion rules.

FAQs

What is Final URL expansion in Google AI Max?

Final URL expansion is the AI Max setting that lets Google replace your chosen final URL with another page on your domain that it predicts will perform better for a given query. It only selects pages that are relevant to the query and themed to the ad group, and it requires text customization to be enabled.

Is Final URL expansion turned on automatically?

Yes. It is enabled by default when you opt into AI Max for Search campaigns. You must uncheck it during setup or in campaign settings to turn it off.

Can Google’s one-click AI Max experiment isolate landing page performance?

Not on its own. The one-click experiment compares a control arm with AI Max off against a trial arm with AI Max on, so targeting changes are mixed into the result. To isolate landing pages, run a custom experiment where both arms use AI Max and only the treatment arm has Final URL expansion enabled.

How do I stop AI Max from sending traffic to certain pages?

Use campaign-level URL exclusions, either as a list of specific URLs or as rules that exclude whole site sections. For large sites, page feeds with custom labels let you exclude entire groups of pages at once.

Does Final URL expansion override pinned RSA assets?

Yes. When Final URL expansion picks a more relevant URL, pinned responsive search ad assets are not used. URL inclusions have the same effect. To always serve pinned assets, you must turn off both Final URL expansion and URL inclusions.

Which websites benefit most from AI landing page routing?

Sites with many specific, crawlable pages that each have a clear conversion path usually benefit most. Lead gen funnels with purpose-built forms, compliance-restricted pages, and sites with thin or outdated content usually need human override.

Point AI routing at buyers who are ready to convert

You now know how to test whether AI Max’s page choices beat your own and how to steer them with exclusions, inclusions, and conversion signals. Intercept adds buyer intent data on top of that, so your best pages reach people who are actively looking to buy.

Talk to an expert→