The test.
Wave 7 put two DAX consumer-goods brands under one question each: does the brand's own page survive an agent's verification of the claim the purchase depends on, and where does the buy actually close? One frozen task per brand, public, locked before measurement. 18 datapoints per brand, across OpenAI, Perplexity, and Anthropic, in German [S2]. The task slate is set by the public Task Selection rule [S4].
- Beiersdorf — verify Eucerin Sun Sensitive Protect Face SPF 50+, the 50 ml SKU: SPF 50+, parfümfrei, UVA protection, made for sensitive skin.
- Henkel — verify Persil Color Megaperls, the 40-wash Beutel: Color means bleach-free, 20 to 60 degrees, Megaperls pearl format.
- Adidas — the Wave 3 anchor rides along as the sector's control case: a sportswear add-to-cart that never got past the front door [S3].
One caveat, stated once for all three. Single task per brand, three-provider track, and the Wave 7 fleet pass is partial: the HTTP and coding agents ran, the browser and autonomous-operator passes land in the next consumer pass. AI Usability is derived from that partial profile. The read is directional, Confidence C [S1][S2].
The claim is on the brand's page. The buy is on the shelf.
Beiersdorf: the claim survives every reader, the checkout was never there.
Eucerin Sun Sensitive Protect Face SPF 50+ is a derma-cosmetic with a four-part locked claim: SPF 50+, parfümfrei, UVA protection, made for sensitive facial skin. Most buyers reach it through dm, Rossmann, Müller, or a pharmacy platform, not through eucerin.de. AI Visibility 34.4, mid-pack [S2].
The page survives the reading test. The HTTP agent reached the locked SKU page and confirmed the full claim in raw HTML: LSF 50+, parfümfrei, the sensitive-skin positioning, UVA and UVB protection, down to the pharmacy code PZN 00802461 [S1]. An agent asked to verify the sunscreen for sensitive skin gets the truth from the brand's own page, no JavaScript required.
One number is missing. The exact UVA protection factor is not on the page. The EU UVA logo implies a factor of at least one third of the SPF, but an agent reading the raw page never sees the number, so it infers where it should quote [S1].
And there is nothing to buy. The brand-own locator routes to partner pharmacies; the transaction completes at dm, Rossmann, Müller, Shop-Apotheke, DocMorris, or Amazon [S1][S2]. The page informs. The shelf transacts.
The claim survives down to the pharmacy code. The buy was never on the page.
Score: AI Visibility 34.4 · AI Usability ~39 (partial fleet) · Composite 4.1 / 10 [S1][S2].
Henkel: a buy button on the page, the dosage missing beside it.
Persil sells three detergent lines in five formats. Getting Universal where Color was asked for is a confident wrong answer, not a near miss. That is the spec-fidelity test Wave 7 froze: Color means bleach-free, 20 to 60 degrees, Megaperls pearl format, the 40-wash Beutel.
The brand surface handles the hard part. The fleet pass reached the Persil Color Megaperls page on persil.de and confirmed the Color line, the Megaperls format, and the clean separation from Universal [S1]. An agent asked for a color-care detergent gets the right chemistry from the brand's own page.
persil.de even carries a JETZT KAUFEN button, the only buy call-to-action in this wave.
What the locked SKU page does not carry: the 40-wash, 1.4-kilo pack detail and the 75-millilitre dosage row. An agent confirming pack size or dosage gets a partial answer from the brand surface [S1]. And the buying path itself runs through the FMCG shelf: dm, Rossmann, Edeka, REWE, Amazon, Kaufland [S2].
The chemistry is on the page. The dosage is not.
Score: AI Visibility 35.3 · AI Usability ~39 (partial fleet) · Composite 4.1 / 10 [S1][S2].
Adidas: the third pathology, the door that does not open.
Adidas sits in the same sector grid and on a different pathology entirely.
Beiersdorf's and Henkel's pages open to agents and verify cleanly. Adidas does not open. Every measured agent class hit an Akamai 403 at homepage entry: the browser runs, the coding agent, the autonomous agent. Before search, before the product page, before any cart [S3].
The anchor marks the third surface mode in consumer retail. Not "the page opens but the close is at retail", which is Beiersdorf and Henkel. Not "the marketplace owns the close", which is the Scout24 and Zalando story. But "the door does not open at all."
Sometimes the question is not where the buy closes. It is whether the agent gets in at all.
Score: AI Visibility 35.0 · Usability not scored (sub-pilot; no agent class got past the entry) [S3].
The shelf owns the close: what the two tell us.
Two brands, two locked claims, one pattern. Both pages survive the verification test. Neither page owns the transaction.
Beiersdorf. The cleanest verification story in the wave: the full claim in raw HTML, pharmacy code included, one numeric short of complete. The buy was never on the page, by design of the German derma-cosmetic retail mode.
Henkel. The only buy button in the wave, and a dosage row missing from the locked SKU page. The brand solved the chemistry disambiguation agents get wrong, then left the pack-size answer to the shelf.
Adidas. The control case. A rich brand surface that no measured agent class ever saw, because the wall came before the page.
The common thread is the surface. Each one was built for a human who already knows the shelf: a parent who recognizes the Eucerin tube at dm, a household that re-buys the green Persil Beutel at REWE. The agent does not know the shelf. It reads what the page hands back and recommends from that. When the page hands back a verified claim and no next step, the recommendation closes wherever a checkout exists [S2].
Brand presence and agent success are not the same thing. The index measures the second.
So the work is on the surface, and the buyer changed. A claim an agent can verify and a next step it can take, on the brand's own page, separate the brand an agent recommends with evidence from the one it hands to the shelf: to dm, to REWE, to the Amazon listing.
