Sixty-eight visitors from assistants. One instruction worth £3,300.
A London private investigation firm, shown here with its name removed, counted 68 visitors sent by AI assistants in its own analytics between 24 July to 7 October 2026: 146 page views, 107 of them in the UK. 8 tapped the phone number from the home page, 3 sent the form, and one became a £3,300 instruction. Every number in this file is the firm's own analytics, never our logs, and it is labelled so throughout.
QuotableVisitors sent by ChatGPT went 2, 14, 43 by month. The jump came in the month a new service page, a technology page and an llms.txt were published, and after ChatGPT's search crawler had taken the site in.
* July from the 24th; October the first seven days. Copilot sent 4 in August. The firm's own analytics, 24 July to 7 October 2026.
The home page first, then the one page that answered a specific question.
Most landed on the home page, which means the assistant had named the firm in answer to a general question (a private investigator in London or the UK) rather than citing a guide. The next most common landing was a service page about open-source intelligence published on 6 September 2026: 10 ChatGPT landings in its first month made it the second most cited page on the site within a month, behind the home page. Then a cost guide, a corporate page, a tracing page.
- named senior staff
- fixed quotes
- ICO registration
- 23 reviews
- a company three years old
- a London address
Three years of existence means the firm was in the models' training data as well as their search index. A new site does not have that, which is why our own two files above show the method from zero.
It is a firm's own analytics export (visitors by referrer, by month, by landing page) and its own enquiry records, read as aggregates and shown here unnamed. It is not our server-log reading of answers served; that measurement did not exist on this site. The two kinds of number are never mixed on this site, and this file carries only the firm's own.