I ran AI Max for two weeks against a manual campaign. The lead-quality delta.

Quick Take

I ran a 14-day head-to-head on one client account: AI Max with Search Term Matching turned on in one campaign, manual control with the same budget and creative in the other. AI Max produced 41% more form fills. It also produced 43% fewer qualified leads. The volume looked great in the dashboard. The sales team had to filter through twice as much noise to find fewer real prospects. Test it. Don’t trust it without a sales-team feedback loop in place.

The setup

The account: a service-business client, $5K/mo Google Ads, lead-generation focused, a CRM that tracks both raw form fills and the “qualified lead” call disposition the sales team logs after a 5-minute discovery call. I had two months of clean baseline data on what “qualified” meant for this vertical before the test started.

The two campaigns:

• Control campaign: manual phrase and exact match, proven keyword set, my standard negative-keyword list, conversion bidding (tCPA at $87).
• Test campaign: AI Max enabled, Search Term Matching on, broad match expansion on, auto-generated headlines, same tCPA, same creative.

Same daily budget split. Same landing pages. Same conversion action wired the same way. The only variable was whether AI Max was on.

What AI Max actually does

When you flip AI Max on, you get a bundle of automations at once:

• Broad match expansion across the existing keyword set
• Search Term Matching that pulls signals from the landing page and matches queries against page content rather than the keyword list
• Auto-generated ad headlines and final URLs Google assembles from the assets
• Partial visibility: AI Max traffic gets its own match type, a Source column, and two footer rows in the keywords report, but low-volume queries stay hidden and the totals overlap with your exact and phrase keywords

Google’s published numbers for this bundle: up to 14% more conversions at a similar CPA, or up to 27% if the existing account is mostly exact/phrase. The pitch is that the AI learns from your best pages and lets you stop pruning the keyword list manually.

What do “Total: AI Max expanded matches” and “Total: AI Max landing page matches” mean in the keywords report?

Those two rows sit at the bottom of the Google Ads keywords report. Expanded matches is traffic from broad-match versions Google built from your keywords. Landing page matches is traffic Google matched from your site and assets with no keyword involved. Zeros mean AI Max matched nothing in that date range. Here’s how I read them.

Where the rows come from

Google added them in September 2025 as footer rows of the keywords report, not the search terms report. Google Ads Help defines them in About reporting in AI Max for Search campaigns: the expanded row counts the broad-match keywords AI Max built off your list, and the landing page row counts matches from your pages and assets, including clicks from Final URL expansion. The same traffic shows in the search terms report as Match type “AI Max” with a Source column you add by hand. The API calls the split AI_MAX_BROAD_MATCH and AI_MAX_KEYWORDLESS.

Reading the row you copied

That row came from the CSV download of the keywords table. A total row leaves the entity columns blank (match type, campaign, ad group, status, max CPC), which is the run of commas after the label.

total: ai max expanded matches,,,,,,usd,0,0, --,0,0.00,0,0,0.00%,0.00,0,0.00
total: ai max landing page matches,,,,,,usd,0,0, --,0,0.00,0,0,0.00%,0.00,0,0.00

Label, blank entity columns, currency code, 0 clicks, 0 impressions, then ”—” where CTR would go, because there’s no ratio on zero impressions, then cost and the conversion columns at zero. The tail depends on your column set.

Copied from the rendered table instead, the same rows lose their separators:

total: ai max expanded matches00 --gbp00.000.00%0.000.00
total: ai max landing page matches00 --gbp00.000.00%0.000.00

One row from an account where AI Max matched something:

total: ai max expanded matches,,,3.00,cad,75.03,,411,5.35%,225.09,,,,13.64%

3.00 conversions at 13.64% implies 22 clicks, and 22 clicks on 411 impressions is the 5.35% CTR. CAD 75.03 is the cost. The 225.09 is most likely conversion value. The rows appear in every UI language under a translated label.

Why yours is all zeros

AI Max is off, search term matching is off at the ad group level, the campaign is paused, or nothing has matched yet. In the exports I’ve seen, the rows appear even when AI Max matched nothing in the date range. A zero row needs no action.

Why the expanded-matches row isn’t extra traffic

Brad Geddes at Adalysis showed the expanded-matches total can include traffic your exact and phrase keywords were already winning: “Your AI Max Totals are not incremental gains.” (AI Max search term reporting). His fix is broad-match twins of each exact and phrase keyword so the match types can be read apart. Google adds its own caveat on the help page above: filtering the search terms report to Match type = AI Max leaves out the “Other search terms” bucket, so the footer rows can be larger than anything you can list out.

What a nonzero landing page row tells you about wasted spend

The rows don’t measure waste on their own. They tell you how much of your traffic Google matched without asking your keyword list, and that’s the traffic worth auditing first, because nobody chose it. In the account I tested, the AI Max campaign produced 41 percent more form fills than the manual control and cost 76 percent more per qualified lead, and almost all of the extra volume came from educational queries the landing page pulled in. A nonzero landing page matches row means Google is reading your site to decide what to buy. Open the “Search terms and landing pages from AI Max” view, sort by cost, and look at the zero-conversion terms. If your CRM tags qualified leads, compare cost per qualified lead for AI Max traffic against your own keywords.

Three checks that take ten minutes

1. Add the Source column to the search terms report.
2. Filter Match type to AI Max.
3. Open “Search terms and landing pages from AI Max”, sort by cost, and read the terms with zero conversions.

Then judge by cost per qualified lead in the CRM, not cost per form fill.

“Not eligible: Ad group paused; campaign paused” on a keyword row

A keyword row that reads “Enabled” in the first column and “Not eligible” under Status, with “Ad group paused; campaign paused” as the reason, has its own toggle on and both parents paused. The status column reports the parent. Max CPC and CTR show ”—” because nothing is set and nothing served. There’s nothing to fix at the keyword level; unpause the parents if you want it to serve.

The two-week deltas

After 14 days, same budget, same creative, same landing pages:

Metric	Control	AI Max	Delta
Impressions	21,400	38,200	+78%
Clicks	642	1,107	+72%
CPC	$3.89	$2.26	-42%
Form fills (raw conversions)	32	45	+41%
Qualified leads (CRM-marked)	14	8	-43%
Qualified-lead rate	44% of fills	18% of fills	-59%
Cost per qualified lead	$178	$313	+76%

The dashboard in Google Ads looked great. More impressions, more clicks, cheaper CPC, more form fills. The CRM told a different story.

Where the extra volume came from

I had reduced visibility on the AI Max side, but the search terms that did show up clustered around top-of-funnel and educational queries the keyword list would never have matched: “how does X work,” “is Y worth it,” “free Z template.” Page content was pulling in researchers, not buyers.

The form fills from those queries showed the same pattern: shorter discovery calls, less prepared, more “just looking” responses. The 45 form fills the test campaign delivered weren’t 45 versions of the 32 the control delivered. They were 14 of the same kind plus 31 of a much earlier-stage cohort.

For a service-business client where each qualified-lead-to-close cycle is 4-6 weeks and the sales team’s bandwidth is a hard constraint, the trade is bad. The team spent more time on 8 qualified leads (because they had to filter 45 fills) than on 14 qualified leads (filtered from 32).

When AI Max probably is the right call

This test was one account, one vertical, two weeks. The result isn’t a universal verdict. AI Max likely earns its keep when:

• The conversion window is long and the brand can afford to nurture top-of-funnel leads
• The product or service benefits from raw volume (high-CLV ecommerce with a strong post-purchase flow, for example)
• The sales team has spare capacity and a working lead-scoring layer that filters fills automatically before they hit the rep
• The existing keyword list is shallow and the landing pages are genuinely the better signal

For lean lead-gen accounts with a quality-constrained pipeline, the test above is the pattern I’d expect to see, not the exception.

How to test it cleanly on your own account

If you want to run the same head-to-head:

1. Pick a campaign with a stable baseline. Two months of consistent CPL and qualified-lead rate.
2. Duplicate it. Same creative, same landing pages, same conversion action, same bidding strategy and target.
3. Turn on AI Max in the duplicate. Leave the control alone.
4. Split the daily budget evenly.
5. Run for at least 14 days. Two full weeks of weekday/weekend cycles.
6. Compare the deltas at the CRM-qualified level, not the raw form-fill level. If your CRM doesn’t tag qualified vs unqualified, fix that before running the test.

Keep going

If this hit, the next two pieces in the same universe:

• Max Clicks, 0 conversions. The bidding strategy that floods accounts with bot traffic and how to read the Microsoft Clarity signal that proves it.
• Why Performance Max Gets Credit for Shopping Conversions. The attribution misread that gets Shopping campaigns killed.

Free PDF: The 25-page Google Ads Setup Audit. The STACK Audit checklist, expanded.

If the dashboard numbers and the CRM numbers don’t match, that’s the audit call. Talk to me.