Google's Undefined 'AI' Label Just Made Your AI Max Creative a CTR Variable

Google began testing an "AI" label on mobile search ads in late April 2026, with two design variants (a rounded pill background and an information icon) and neither of which surfaces a tooltip or explanation when tapped. The label is undefined to users, which means Google can ship a transparency feature without committing to what it discloses. Advertisers running AI Max, Performance Max, or Responsive Search Ads should ship a non-AI-assisted creative control group this quarter to measure the labeling CTR penalty before Google decides what the label means.

What Brodie Clark actually saw

The label was first spotted by Melbourne-based SERP researcher Brodie Clark and reported by Search Engine Roundtable on April 27, 2026. According to PPC Land's writeup, two visual treatments are running in parallel: one with the "AI" text inside a rounded pill background attached to the ad block, the other with an information icon next to the "AI" designation. Tapping either one currently does nothing. No tooltip. No disclosure panel. No definition of what the label is actually telling you.

That last detail matters more than the design. A label that does nothing on tap isn't a transparency feature, it's a placeholder. Google is testing the visual real estate before deciding what claim, if any, sits underneath it.

Why a placeholder label is the actual story

The obvious read is that Google is signalling AI-assisted creative, AI-assisted bidding, or AI-assisted matching through products like AI Max for Search and Performance Max. The less obvious read is that Google is buying optionality. If state-level disclosure rules tighten, Google can retroactively define the label as a creative-AI marker. If the FTC pushes on bidding transparency, the label can quietly shift to that. The visual stays the same. The legal meaning fills in later.

That's a reasonable hedge given where regulation is moving. New York's S.8420-A "synthetic performer" disclosure rule takes effect June 2026. California's AI labeling and metadata rule kicks in August 2, 2026 with a $5,000 per day penalty for stripping embedded provenance, per an NYU Compliance and Enforcement summary. The federal picture is messier. The FTC walked back its Rytr enforcement earlier this year, signalling a softer posture on generative AI broadly. So Google has a patchwork to manage, and a label that can be defined later is exactly how you manage one.

The other half of the story: sponsored grids are quietly everywhere now

The label test arrived during a noisy expansion of sponsored placements across Google's surfaces. The timeline, pieced together from PPC Land's reporting and its companion piece on the Shopping tab:

• April 2, 2026: Sponsored tiles appear inside main search results product grids, 1 to 4 per grid.
• April 20, 2026: Free listing grids on the Shopping tab get sponsored tiles mixed in.
• April 22, 2026: The "related products" grid below standard organic results becomes a sponsored surface.
• April 27, 2026: The undefined "AI" label test gets documented.

The labeling on these new grid surfaces is section-level "Sponsored" rather than per-tile. That's a softer signal than the per-tile markers used in Shopping or AI Mode, and it's the part advertisers shouldn't gloss over. If a single grid has 1 organic tile, 2 sponsored tiles, and 1 organic tile, the user sees four products and one "Sponsored" word above them. The intent reads as discovery, the inventory reads as ad. CPCs land on what would have been free clicks two months ago.

What this means for paid search teams this quarter

Three things changed in roughly six weeks: a new ad surface (related products grid), a new ad surface (Shopping tab free listings), and the visual scaffolding for an AI disclosure that Google hasn't committed to yet. The shopping shifts you can't opt out of. The labeling one you can prepare for.

From what I've seen in client accounts running AI Max alongside legacy Responsive Search Ads, the moment a new visual signal lands on the SERP, CTR moves before performance teams notice. The "Sponsored results" sticky header Google rolled out in October 2025 caused measurable CTR shifts on competitive head terms in the weeks after launch, which Search Engine Journal flagged at the time. An "AI" label, even an undefined one, is a candidate for the same kind of friction.

Personally, I wouldn't wait for Google to announce the label's meaning. The honest move is to set up the comparison now, while you can still get clean data.

The 60-minute control group setup

If you run Google Ads at meaningful scale, here's the audit I'd run this week:

1. Pull a creative inventory split. In Google Ads Editor, export every active ad. Tag each one as AI-assisted (Performance Max, AI Max for Search, RSA with autogenerated assets enabled) or human-only (RSA with autogenerated assets toggled off, manually written ETAs where still grandfathered). Most accounts are running 80%+ AI-assisted by default. That's the asymmetry.
2. Build a 10-15% control budget. On 2 to 3 of your highest-volume ad groups, duplicate the campaign with autogenerated assets disabled and lock the creative to human-written assets only. Match the budget split. Keywords, landing pages, audience signals, all identical.
3. Set the watch window. 21 days minimum. CTR delta is the headline metric. Conversion rate and CPA matter, but secondary, because the label hits the click decision, not the post-click experience.
4. Track impression share by surface. If you can isolate mobile-only impressions, do it. The label is mobile-first in testing. The CTR penalty, if there is one, will show up there before desktop catches it.

The benchmark to watch for: any CTR gap above 4% between AI-assisted and human-only creative on the same query set, on mobile, after the label expands beyond test. That's the rough threshold I'd treat as "this is real, not noise." If you see it, the calculus on AI Max defaults changes, at least for high-CTR-elasticity head terms.

The case the label never gets defined

It's worth saying out loud: this is a Google SERP test. Google kills SERP tests all the time. There's a real chance the "AI" label gets pulled in 60 days and never resurfaces. ALM Corp's analysis on April 28 made roughly the same point, that the test could be early exploratory work that never ships.

So why bother running the control group? Because the cost of running it is small (a duplicate campaign and three weeks) and the cost of not having pre-label CTR data when the label expands is large. You can't reconstruct the counterfactual after the fact. Either you have a clean before-and-after, or you're guessing.

And the related products grid expansion isn't a test, it's already shipping. So even if the AI label disappears tomorrow, the surfaces it would have appeared on are now permanent inventory you're competing for. That part isn't going back in the box.

The slightly off-script part

One thing nags me about the timing. Google rolled out the sticky "Sponsored" header in October 2025 specifically because the ad/organic distinction had gotten muddy. Six months later, they're testing a label that's even less defined than the one they just clarified. It's possible those two motions are coordinated, with the AI label slotting into a future taxonomy of disclosures. It's also possible they're not, and the AI label is a small team's experiment that hasn't been reconciled with the broader transparency push. We won't know which until Google ships or pulls it.

Either way, the related products grid is now an ad surface, and your AI-assisted creative is sitting under a label nobody has explained. Both of those facts changed in the last 30 days. The accounts that come out of Q3 with clean data on what the label costs them in CTR will be the accounts that didn't wait for an announcement.

For context on how much default-on AI behavior is already shaping Google Ads, see our prior reporting on AI Brief letting you preview AI Max output and on the PMax partner network exclusion alpha. The pattern across all three: Google is shipping AI-flavored controls fast, but the reporting and definitional layers are lagging. Build your own measurement before the labels arrive defined.

Notice Me Senpai Editorial