Carets

We asked four web-grounded AI assistants the same shopping question — "what are the best barefoot shoes for the office?" — phrased ten different neutral ways, using no brand names, and repeated each. We recorded which products each assistant recommended and how often. The signal is how frequently a product appears, not its position in any list.

Carets appeared in 77% of office answers, in all four engines — far above the rate this category's candidate set would produce by chance. (Across ten neutral phrasings the rate ranged from 50% to 100% — Carets is surfaced more for some office phrasings than others; the 77% is the honest average, and this phrasing sensitivity is why the record is Observed, not Verified.) The result held when the question was re-worded and under a conservative bootstrap interval clustered by prompt and engine.

The assistants' recommendations cited a diverse set of sources — not a single affiliate or the brand's own site alone. Domains observed citing Carets for this need include:

Every figure on this record links to the underlying, timestamped observations on the evidence page, where each answer, engine, date and citation is auditable.

• It meansAI shopping assistants consistently put Carets in the answer when people ask for office-appropriate barefoot shoes.
• It doesn't meanthat Carets is "the best," or anything about sales, quality, or fit — only that it is frequently recommended by AI for this need.
• ScopeMeasured in English for the US market. A claim for another market or language requires measuring there.
• Earned, not boughtThis signal cannot be purchased. It reflects what the AI engines actually recommend, and is re-checked quarterly.

On the merchant's product page, the record shows as a small, calm, monochrome badge that adapts to the store's theme. It always carries the category, the denominator, the engines, the date, and a link back to this evidence — never "ChatGPT recommends this."