# Edition 1 correction: two crawler labels, 2026-10-02

**What this is.** A dated correction to how two Edition 1 (July 2026) crawler labels are presented on the Index page. No share, rank or conviction figure changes. The CSVs at this URL are unchanged, and so are the labels themselves. The two rows now carry a footnote on the page, and the sentences that relied on them were rewritten. Nothing was edited silently.

## What was wrong

Edition 1 reported two tracked brands as blocking all three AI crawlers it tested:

| Vertical | Brand | Published label |
|---|---|---|
| Property | Rawson Property Group | blocks GPTBot + ClaudeBot + PerplexityBot (403) |
| Legal | Bowmans | blocks GPTBot + ClaudeBot + PerplexityBot (403) |

The probe also fetched each homepage as a normal browser. For these two sites the stored browser fetch was refused as well, with a 403 on both runs. A refusal that a normal browser also gets is the site turning away our requests in general, so it says nothing about AI crawlers in particular. Edition 1 did not use the browser fetch as a control, and it should have.

Stored codes, two fetches each (`benchmarks/property-2026-06/ssr/ssr-results.json`, `benchmarks/legal-2026-07/ssr/ssr-results.json`):

| Brand | Browser | GPTBot | ClaudeBot | PerplexityBot |
|---|---|---|---|---|
| Rawson Property Group | 403 / 403 | 403 / 403 | 403 / 403 | 403 / 403 |
| Bowmans | 403 / 403 | 403 / 403 | 403 / 403 | 403 / 403 |

Every other Edition 1 block label had the browser served on both runs, so it stands as published. That covers RE/MAX (GPTBot), PwC (all three), Adams & Adams, DSC Attorneys, STBB and The Beancounter (ClaudeBot).

## What changed

- The two labels carry a footnote on the Index page: a normal browser was refused too, so this 403 is not specific to AI crawlers. The label text is kept as first published so that Edition 1 and Edition 2 stay comparable.
- The property and legal commentary no longer describes either site as blocking AI crawlers. The accounting commentary no longer compares PwC's block to Bowmans'.
- The methodology now uses the browser fetch as a control. Edition 2 (pre-registered to publish on 6 October 2026; published 9 October 2026, three days late) adds a search-crawler verdict, which reports a crawler 403 the browser also gets as "not tested". Its Edition 1-method column will carry the same footnote wherever the browser was refused.

Found 2026-10-01 in an internal review of our crawler method. Corrected 2026-10-02: footnoted rather than restated, to keep the editions comparable. A test (`src/content/research/index/edition-2026-07.test.ts`) keeps the footnoted rows equal to the stored probe's browser-refused blocks.
