Still Fox Free snapshot

Original data, published whatever it shows.

Still Fox, a Seattle practice, publishes original research on what AI assistants can read and say about businesses.

Entries

Research here means original data, a stated method, intervals, a claims section, a version and a date, published whatever it shows. We do not repackage other people's studies as research. Those appear as graded rows in the evidence ledger, with vendor studies marked. An entry that finds nothing is published as a null result.

In preparation

Seattle Law Firm AI Readiness Index, first edition

Sites checked 2026-10-11. Version 1 when published. Publication date set at publication.

On 2026-10-11 we checked the home pages of 22 Seattle law-firm websites on four things: whether robots.txt lets each AI crawler in, whether the firm's CDN or firewall turns away a request that carries a crawler's user agent, how much of the page's text exists without JavaScript, and whether the firm's name, address and phone appear on the page. The first edition reports counts only, until a firm consents to be named. The method is one fixed set of automated requests per site, sent with each crawler's user agent and compared with a plain browser request.

In preparation

Baseline: what the assistants say about Still Fox before anything changes

Design frozen 2026-10-11. Version 1. Any change after the first real run starts version 2.

We ask a fixed set of questions in three groups: what a partner agency would ask, what buyers ask in our first vertical, law firms, and broad control questions. Each question is asked many times on the official search APIs of OpenAI, Perplexity and Anthropic. Gemini is pending its terms. The surface is API only. The measure is the share of answers that name Still Fox, with 95% Wilson intervals, per cluster and per assistant. We froze the design and our expectation on 2026-10-11, before any real run, and the result will be published whatever it shows.

Live, as of 2026-10-11

The evidence ledger

Lives on the Method page.

Every tactic we use, sell or refuse is graded by the strength of the study behind it, from A for controlled tests to D for unsupported claims, and the tactics we refuse to sell are marked. Read it on the Method page.

Get the monthly research email

One message a month at most, listing what published. Nothing else. Unsubscribe in one click.

Rules for every entry

  • Every number sits beside its interval or count, the surface (official API or consumer app) and the run count. A screenshot is not a measurement.
  • Every entry states its method, carries a claims section, and shows a version and a date.
  • Every entry is published whatever it shows. A null result is published as a null.
  • No firm is named without its written consent. Until then, counts only.
  • We name the sponsor of any survey we quote, say when a source is a vendor or a test is dated, and promise no rankings, mentions or leads.

For a first reading of your own business, request a free snapshot.