AI Max Holdouts Kept CPCs Flat. Their DSAs Auto-Upgrade in September.

Six agencies tracked Google's AI Max for Search through its first year and reported CPC increases of 10 to 25 percent depending on adoption depth, alongside year-over-year search spend up 7 to 15 percent. Holdouts who stayed on Dynamic Search Ads kept their search CPCs roughly flat. That window closes in September, when Google auto-upgrades all remaining DSA, automatically created assets, and campaign-level broad match Search campaigns into AI Max.

The agency numbers do not look like a 7 percent conversion lift

Digiday's year-one survey pulls together data from six agencies that have actually run AI Max in production accounts. Mediaplus told Digiday its clients saw average CPC rises of 10 to 15 percent. Collective Measures reported a similar 10 percent typical range with some clients hitting 25 percent, alongside 7 to 10 percent year-over-year spend growth on accounts using AI Max as a growth lever. Go Fish Digital reported a 15 percent year-over-year search spend increase. Ashley Fletcher, CMO at Adthena, summed it up in one quote: "The CPC pain is real."

The auction-side data lines up with what advertisers feel. Adthena estimates a 35 percent year-over-year increase in advertisers participating in search auctions. So you have more bidders, more AI-driven bid pressure, and more spend chasing the same query inventory. The CPC chart goes one direction.

An independent analysis from Smarter Ecommerce (SMEC) of more than 250 Search campaigns running AI Max found a median revenue lift of 13 percent paired with a median CPA increase of 16 percent. ROAS results stretched from 42 percent above baseline to 35 percent below, and only 22 percent of campaigns hit their original ROAS target. The headline number Google leans on is "7 percent more conversions or conversion value at a similar CPA/ROAS." The headline number practitioners are seeing is "more revenue, but the CPA tax is bigger than the lift on most accounts."

Why holdouts have been getting away with flat CPCs

The contrarian read here is the one most paid social and search managers I talk to keep missing. Accounts that did not adopt AI Max, kept DSA running, or stuck with manual broad match did not eat the same 10 to 15 percent CPC increase, because they were not bidding into the AI-priced inventory the same way. The auction floor went up because adopters added incremental budget and let the system optimize toward the maximum each advertiser could afford. Holdouts coasted on whatever auction position their existing keyword stack still earned.

This is also why year-over-year CPC reports look so different account to account. The number you see in your interface depends almost entirely on whether you flipped the AI Max switch and when. From what I have seen across paid search reviews this quarter, accounts that migrated tier-1 campaigns in the spring of 2026 tend to be on the high end of the CPC range. Accounts that ignored the upgrade prompt and kept DSAs warm tend to be flat. The gap is real, and it is about to close.

What September actually upgrades, and what it bills

Google's official announcement is unambiguous about scope. Starting in September, three categories of legacy Search settings auto-upgrade to AI Max: Dynamic Search Ads campaigns with dynamic ad groups, Automatically Created Assets, and campaign-level broad match. After the upgrade window closes, you cannot create new DSA campaigns through the Google Ads interface, Google Ads Editor, or the Google Ads API. Search Engine Land and Search Engine Journal both confirm there is no opt-out path for September; the only options are migrate now on your terms, or migrate then on Google's.

The 7 percent conversion claim that Google cites comes with footnotes most slide decks strip out. It applies to non-Retail advertisers, requires the full feature suite (search term matching, text customization, and final URL expansion), and is sourced as "Google internal data, 2026." That is the comparison against using search term matching alone, not against your existing manual broad match or DSA setup. If your starting point is a tightly managed exact and phrase match keyword stack, the 7 percent does not necessarily apply to your delta.

Read the 7 percent claim like a paid media operator

The cleanest cross-check on Google's claim is the Brainlabs analysis of 23 tests across 16 mature advertiser accounts, which found campaigns enabling all three AI Max features had a 40 percent higher success rate than baseline search term matching. That is more flattering than SMEC's median, but it is also a comparison against the AI Max baseline, not against the legacy DSA or manual broad match setups that most accounts will actually be migrating from in September.

So the honest framing is: AI Max with full feature suite likely beats AI Max with feature suite turned off. AI Max with full feature suite versus a well-tuned legacy account is the question almost nobody has tested at scale, and the SMEC data suggests the answer for the median account is "more revenue, worse CPA." That should change how you set your post-September ROAS target, not your conversion expectation.

The 90-day audit to run before August

Here is the audit sequence that actually de-risks September. None of this requires new tooling.

For every Search campaign currently using DSA, ACA, or campaign-level broad match, log a 90-day baseline now: CPC, impression volume, conversion volume, CPA, and ROAS. The reason for 90 days is that auction conditions in May 2026 are noisier than usual because of the wave of voluntary migrations Google's account managers have been pushing. A shorter window distorts your before-and-after comparison.

Migrate one tier-2 campaign manually using Google's voluntary upgrade tool, which ports historical settings into the new standard ad group structure. Run it for 30 days. Track CPC delta, conversion volume delta, search term match rate, and ROAS against the original target, not against a softened target someone reset for AI Max. If your observed CPA increase tracks above 15 percent (the SMEC median), that becomes your tier-1 forecast.

For tier-1 spend, do not rely on the 7 percent conversion claim as a budget input. Reset your tier-1 budget assumption based on the CPC delta you measured on tier-2, plus the SMEC CPA distribution. The accounts that look caught flat-footed in October will be the ones who took the 7 percent number into a Q4 planning meeting.

One related angle: Google's AI Brief feature now lets you preview AI Max output in plain language before launch, which is at least a partial answer to the "what is the algorithm doing with my budget" complaint. Use it during the tier-2 test, not after the September flip.

The CFO conversation that should happen before August

My honest read is that most accounts will find their tier-1 campaigns absorbed more of the 10 to 15 percent CPC bump than the agency averages suggest, because the six agencies Digiday surveyed are mature and have already moved their tier-1 spend. The accounts at the most exposure in September are the long tail of mid-market advertisers who left DSA running on autopilot and assumed the upgrade was opt-in. It is not. And the conversion lift number is not a budget guarantee, especially against a non-AI-Max baseline.

Pencil in the 90-day audit window now. If you wait until your account manager calls you in August, you are negotiating inside someone else's calendar, not yours. The accounts that walk into Q4 with their own observed CPA delta in hand are the ones who will hold the line on ROAS targets when AI Max becomes the only option.

Notice Me Senpai Editorial