Singapore
Google Ads AI Max audit: what to check first

Google Ads AI Max audit: what to check first

Run this audit in the order the decisions were handed over: what each campaign switched on, what the matching layer now matches, the copy Google writes, the landing pages it chooses, and whether lead quality improved. As of November 2026, the useful question is which of those decisions your team has quietly stopped approving. Teams weighing AI automation for business plans meet the same boundary: the automation stops at the edge of the ad account.

  • Every AI Max feature has a matching report, so a check without a report is a guess.
  • Google's headline figures describe averages; a campaign sample shows what the extra volume costs.
  • The first fixes are the ones that stop money leaving the account today.

Who should run this audit, and who can skip it

Run it if a campaign was auto-upgraded, if someone enabled AI Max features without a written record, if leads are judged by sales rather than form fills, or if the ads carry mandatory wording. The checklist suits accounts where a lead or a sale can be traced back to a search query. It does not suit brand campaigns judged on impression share, or accounts too small for one week of data to separate a losing term from noise.

Stage one: What did the upgrade already switch on?

Brandon Ervin, Director of Product Management for Google Ads, wrote in a post published in April 2026 and updated in June 2026 that the Dynamic Search Ads sunset and auto-upgrade now begin in February 2027, while campaigns using automatically created assets and campaign-level broad match keep auto-upgrading from September 2026. The mapping matters more than the dates. Automatically created assets arrive with search term matching and text customisation on; campaign-level broad match arrives with search term matching on; Dynamic Search Ads brings all three features and keeps its old URL controls.

Two campaigns can show AI Max as enabled and run different levels of automation, so the check is per campaign. Open the AI Max settings panel at campaign level, then at ad group level. Search term matching can be set at ad group level; text customisation and final URL expansion are campaign settings.

Pass rule: for every campaign you can name the features that are on and who turned them on. If nobody can answer the second half, that campaign goes on the fix list first.

A marketer points at campaign settings on a monitor while a colleague notes who enabled each feature.

Which AI Max settings to audit first, and how to verify each one

Google's help page on reporting in AI Max for Search campaigns lists the report that proves each feature. Work through the four controls in this order, because each depends on what the previous one exposed.

Search term matching

Open the search terms report, add the Source column and filter Match type for AI Max. The "Search terms and ad combinations" view adds the headline and landing page that served. The AI Max filter can show lower numbers than expected because it leaves out the "Other search terms" bucket. Pass rule: every term that spends money traces to a query you would bid on.

Text customisation

Open the campaign-level asset report and the "Expanded final URL assets" tab, then the ad-level asset report. Assets built by customisation carry "Google AI" in the Added by column. Removal works asset by asset in the interface; bulk removal needs the API or Google Ads Editor. Pass rule: no generated line makes a claim a compliance reviewer would reject.

Final URL expansion

Open the landing pages report and read the "Selected by" column, which shows whether AI Max chose the URL. The keywords report adds an aggregate row, "Total: AI Max landing page matches". Pass rule: no paid visit lands on a page that cannot convert.

Brand and URL controls

Brand lists live under Tools, Shared library, Brand lists, and apply inside the AI Max settings panel at campaign or ad group level. URL inclusions and exclusions sit in their own tabs on the Keywords page. Pass rule: brand and competitor boundaries are written into a control, not left to the algorithm.

Stage two: Are the searches still the ones you bid for?

Mike Ryan of Smarter Ecommerce analysed one million AI Max impressions and found a clean 80/20 split: AI Max expands mostly on exact and phrase keywords, and has little to expand on broad match because those are already broad. Broad match expansion and keywordless matching often shared a roughly 50/50 contribution. An exact match keyword no longer limits the traffic it brings.

His team also documented accounts where competitor queries took most of the Search impressions. One account is not a forecast, but it shows how far the mix can drift. AI automation for business tools export and cluster queries; the decision about which ones you would bid on stays with your team.

Google's best practice notes give the rhythm: review search terms and item group performance every 1-2 weeks, and add a negative keyword only when a term consistently underperforms across all campaigns. After enabling AI Max, wait at least 2 weeks before changing negatives, because the learning period makes early readings unreliable.

Pass rule: you can point to a report from the last two weeks and name the queries that would have failed your old keyword plan.

Stage three: Is the copy Google writes safe to serve?

Text customisation builds headlines and descriptions from your ads, keywords and landing page content. Ask of each generated asset: is it accurate about the product, does it fit the ad group's audience, and is it earning spend? The asset report answers the third with performance data.

Pinning needs its own check. Google's AI Max documentation confirms pinning is not always respected when final URL expansion is enabled, and ignored when URL inclusions are set, because Google may write copy for a different destination. If a disclaimer must appear, treat pinning as a preference. Pass rule: a named person has read every Google AI asset that served this month, and campaigns with mandatory wording have a written workaround.

Stage four: Where does final URL expansion send traffic?

The landing pages report shows which pages took money and which AI Max selected. Look for pages that exist for other reasons: help articles, documentation, pages for other services, old offers that still convert.

Compare each page with the ad that sent the visitor there. A page can have a form and still break the path, for example when it asks for details the ad never mentioned. Lead quality belongs here too: a page can add form fills while adding unqualified enquiries, and the campaign report will not show the difference. Teams buying AI automation for business to police page rules start here. Pass rule: every URL taking spend has a conversion path that matches the ad.

A visitor holds a phone showing an ad while looking at a laptop form asking for unmentioned details.

Stage five: Do the extra conversions pay for themselves?

Google's claim is specific. In its June 2026 update, the company says campaigns using the full feature suite see an average of 7% more conversions or conversion value at a similar CPA or ROAS compared with search term matching alone. Search Engine Journal's report in November 2026 notes that the figure comes from non-retail data and is not an expectation for every account.

Google gave Smarter Ecommerce a larger pair: 14% more conversions or conversion value for advertisers that activate AI Max, and 27% for campaigns still mostly on exact and phrase keywords. Smarter Ecommerce tested the claim on its own book. Across more than 250 Search campaigns, the median uplift was 13% more conversion value, and the incremental conversions cost a median 16% more than the business-as-usual campaign. Its analysis could not rule out that some volume moved from other campaigns rather than being added.

That is a reason to judge cost per qualified lead instead of conversion count, not a reason to switch AI Max off. Comparing that number needs the CRM record of what happened after the form, which is why teams asking about ERP integration services arrive at this stage. In a sample of 600 active accounts, roughly 16% of advertisers were testing AI Max, so a controlled test is a normal position.

Pass rule: the campaign has an experiment or a clean before-and-after period, and the number you judge it on includes qualification, not form fills.

Stage six: What do Dynamic Search Ads advertisers still owe themselves before February 2027?

The date is fixed. Dynamic Search Ads sunset and auto-upgrade begin in February 2027, and unsupported legacy targets may become read-only when the migration runs. Advertisers still running dynamic ad groups can decide which pages and services stay eligible, and clean the targets that made sense when they were created.

Review the dynamic ad targets, URL exclusions and search term history before the upgrade. Waiting for the automatic mapping preserves the legacy setup, including the parts you would not choose today. Page eligibility comes before any rebuild around ERP integration services and a shared customer record, because the data work depends on knowing which pages may sell.

Pass rule: you can list the pages that should stay eligible, and the ones that should not, without opening the account.

What to fix first, ranked by impact

  1. Stop paid traffic to pages that cannot convert. Add URL exclusions for support pages, old offers and other-service pages, then recheck the report after a week. The spend continues every day the page stays open.
  2. Cut competitor and brand leakage in non-brand campaigns. Add negatives or brand exclusions where the search terms report shows queries outside the campaign's purpose.
  3. Review Google AI copy in regulated or partner-led categories, and record which campaigns depend on pinning.
  4. Correct settings that no longer match the campaign's job, such as a feature switched on by default.
  5. Put qualification into the measurement before you scale, because cost per form fill flatters a campaign that produces poor leads.

The first two cannot wait: they cost money while you decide.

A marketer shuts a valve on a leaking pipe above a tray of coins as a colleague watches.

Printable summary: twelve yes or no questions

  1. Have you listed which AI Max features are on in every campaign?
  2. Can you name who enabled each feature?
  3. Have you checked the ad group settings as well as the campaign settings?
  4. Can every spending search term be traced to a query you would bid on?
  5. Have you reviewed search terms in the last 1-2 weeks?
  6. Did you wait at least 2 weeks after enabling before changing negative keywords?
  7. Has a named person read every Google AI asset that served this month?
  8. Do campaigns with mandatory wording have a written workaround for pinning?
  9. Does every landing page taking spend have a conversion path?
  10. Are the pages that should never take paid traffic in the URL exclusions?
  11. Do brand and URL controls state your boundaries explicitly?
  12. Do you judge the campaign on qualified leads rather than form fills?

Where Topkee fits when the waste sits outside the platform's report

Topkee's ad spend waste audit starts from the client's own CDP, pixel and CRM data rather than the ad platform's back office, which is where the checks above stop. The audit ships with every report instead of being an add-on, and the free diagnosis names blank search terms and fatigued creatives cell by cell. What comes back is a numbered, trackable action list rather than a set of recommendations, with read-only access that takes 2 minutes to grant and can be revoked later.

The Campaign Intelligence Report adds the measurement layer the last stage needs: a six-layer funnel from impressions to signed deal, with ROAS reported each period as signed deal value divided by ad spend. Topkee does not ask clients to switch Google's automation off; the point is to feed better conversion signals back and audit the waste the account cannot see.

What should you do next?

Start with the stage that matches the next budget decision. If the account was upgraded this quarter, run the first two stages this week and take the rest next cycle. For the wider picture, read how AI Max changes matching in Google Ads and common Google Ads AI mistakes. Teams that want the audit to run on live CRM data should treat ERP integration services as the step after those two.

Sources

  1. "Google Ads AI Max Is Making More Decisions For You – Here's What To Audit Now", Search Engine Journal, 18 November 2026. https://www.searchenginejournal.com/google-ads-ai-max-is-making-more-decisions-for-you-heres-what-to-audit-now/596972/
  2. "We're upgrading Dynamic Search Ads to AI Max", Google (Google Ads blog), 15 April 2026, updated 11 June 2026. https://blog.google/products/ads-commerce/dsa-upgrade-to-ai-max-2026/
  3. "About reporting in AI Max for Search campaigns", Google Ads Help, accessed 19 November 2026. https://support.google.com/google-ads/answer/16470459
  4. "The Ultimate Guide to AI Max for Google Search", Smarter Ecommerce, updated 23 March 2026. https://smarter-ecommerce.com/blog/en/google-ads/the-ultimate-guide-to-ai-max-for-google-search/
Share to:
Date: 2026-11-19
Wing Yau

Article Author

Wing Yau

Marketing Manager