Be the name
the assistants give.
Search is turning into an answer. Google answers a growing share of searches itself, Bing has Copilot, ChatGPT has search, and Perplexity never had a results page. When someone asks an AI assistant for a solicitor, a clinic or a tradesperson, the businesses named in that answer are being chosen now, and the names stick. LLMs Bureau gets your website read and cited in those answers, then shows you each one as it is served. We have done it three times from scratch, and the readings are below.
LLMs Bureau is a UK answer engine optimisation service for small and medium businesses and professional practices: it gets a website read and cited by AI assistants such as ChatGPT, Claude and Perplexity, then shows the owner every answer the site was used in, with the log line to prove it.
ChatGPT answers from live fetches so far: it read 12 different pages in week one, and one answer drew on five of them. Bing has barely started, which is the gap we are closing.
- ChatGPT/The brand, or "private investigator London" / "private investigator UK"
- ChatGPT/guides/how-to-find-someones-address-ukFinding someone's address in the UK
- ChatGPT/guidesWhat the firm publishes
- ChatGPT/services/people-tracingTracing a person
- ChatGPT/guides/how-to-find-a-lost-friend-ukFinding a lost friend
- ChatGPT/services/covert-surveillanceWhat a UK investigator may lawfully do
- ChatGPT/guides/can-i-hire-a-private-investigator-to-follow-someoneCan an investigator follow someone
- ChatGPT/guides/is-private-investigator-evidence-admissible-in-courtWhether the evidence stands up in court
- ChatGPT/services/cyber-investigationsWho is behind an account
- ChatGPT/guides/who-is-behind-an-anonymous-accountWho is behind an anonymous account
- ChatGPT/termsChecking the firm out
- ChatGPT/technologyWhat tools the firm uses
No readings yet. We will read your site the way ChatGPT does: the home page as a browser and as the named AI crawlers, plus robots.txt, llms.txt and the sitemap. A score out of 100 and the findings come back on one page.
Reading your site as a browser, as ChatGPT-User, as GPTBot, as ClaudeBot and as PerplexityBot. Ten seconds or so.
Free. No sign-up to see the score. Five checks an hour per visitor. We fetch from our server with each reader's user agent set; rules that check a crawler's address may treat the real reader differently.
Pages in the sitemap fetched by each index's crawler, to 14 Oct 2026. Answering people with our pages: ChatGPT, DuckDuckGo AI and Perplexity. Not yet: Claude, Gemini, Copilot.
- ChatGPT/The brand, or the category: "address tracing service UK", "tracing agent UK"
- ChatGPT/pricingWhat a tracing agent costs
- ChatGPT/use-cases/trace-a-debtorFinding a debtor's address
- ChatGPT/use-cases/reconnect-with-someone + /aboutFinding a lost person, and who is behind the service
- ChatGPT/for-solicitorsTracing for a law firm
- ChatGPT/developersA tracing API
- ChatGPT/how-it-worksHow an address trace works
- DuckDuckGo AI/guides/finding-missing-beneficiariesHow to find missing beneficiaries
- DuckDuckGo AI/how-it-worksHow an address trace works
- Perplexity/The brand, or the category
There is no better time to start, because the names are being chosen now.
The query is moving to the assistants.
Traditional search engine volume will fall 25% by 2026, with search marketing losing share to AI chatbots and other virtual agents.
Gartner, press release, 19 February 2024
68% of US Google searches ended without a click in January to April 2026, up from 60% in 2024; AI Overviews now appear on more than a fifth of searches, and when they do, clicks fall by nearly 60%.
SparkToro, Similarweb clickstream data, 2026
An assistant reaches for what it has already used.
Every answer served is a page read again. Once our address-tracing site's home page had been used, it carried 75% of the answers that followed: people asking for the category were given the same name each time. The sites in the answers this year are the ones the assistants keep returning to. The ones that are not have to displace them.
Our address-tracing site / server logs / 29 Sep to 14 Oct 2026
Three sites from scratch, three in the answers.
A UK address-tracing service, cited inside its first fortnight. A private investigation firm, 98% of its pages read by Google within nine days. And Bureau Nine, a one-page holding site, invite-only, nothing to sell: read by Claude the day after it went up and by ChatGPT, to answer people, since 08 October. No backlinks, no paid links, no history. The method is ours and we do not publish it; what it does is on this page.
Server logs of the three sites / to 14 Oct 2026
Three new sites. No backlinks. No paid links. Read by the assistants inside a fortnight.
We launched a UK address-tracing service on 29 September 2026. Once ChatGPT's search crawler had taken it in, the answers it served a day rose 10x, and on 14 October ChatGPT started sending visitors, one of them straight to the instruction form. Investigation Bureau went live on 5 October; within nine days Google had read 98% of its pages, Anthropic 94% and OpenAI 91%. A London investigation firm that we show unnamed saw the visitors ChatGPT sent it rise 207% from August to September 2026, in its own analytics, and one of those visits became a £3,300 instruction. Even Bureau Nine's holding site, one page with nothing to sell, is fetched by ChatGPT and Claude to answer people who ask about it.
Sources: our server logs (live fetchers such as ChatGPT-User, and index crawlers such as GPTBot and Googlebot read against each site's sitemap) and the firm's own analytics, never mixed. Every percentage names the periods it compares; the raw figures are in the case files.
The day after ChatGPT's search crawler took the site in, the run started.
Answers served a day rose 10x once ChatGPT had indexed our address-tracing site, week two was up 300% on week one, and 89% of everything ChatGPT has answered from the site came after the crawl. The home page carries 75% of those answers, which means people are asking for the brand or the category by name. On 14 October the first visitors arrived from ChatGPT, one of them straight to the instruction form.
DuckDuckGo AI cited /guides/finding-missing-beneficiaries and /how-it-works; Perplexity cited the home page.
Questions shown as likely are inferred from the page fetched; the question text itself is only visible in Bing Webmaster and Search Console AI reports. Open case file 01.
Each assistant reads a different index, and none of them runs JavaScript.
Google's AI reads Google. Copilot reads Bing. ChatGPT mixes Bing, its own index and live fetches. Claude reads Brave. Perplexity runs its own crawler. No AI crawler we have logged runs JavaScript, so a site that draws its text with code is a blank page to every one of them.
Also reading it: Amazon Alexa 100%, Meta AI 97%, Apple 69%, Anthropic 17%. The share of the sitemap each index's crawler has fetched, from our logs; a log shows fetches, not index membership.
ChatGPT search: not yet; ChatGPT answers come from live fetches Bing: 1% of the site
Showing our own gaps is the point.
A page built this way scores under 30 in the free check. Server-render or pre-render the pages and the right-hand view fills in.
We do not need the keys to your site. One script line and a Monday email.
Start with nothing installed: on a paid plan we ask the assistants the questions your buyers ask three times a week and record who gets named. Add one line of script and we see the visitors the assistants send and what they enquire about, without a single cookie. Add a small snippet at your edge and we see every answer served, per assistant, per page, as on the Bench above. Then we write the exact changes, as numbered change orders your developer applies in an afternoon. On the top plan we apply them ourselves.
The Monitor
Three times a week on a paid plan we put the questions your buyers ask to ChatGPT, Claude, Perplexity and Gemini and record who gets named, in what position, and who is named instead of you. A free account gets one round, once its site is connected.
Named or not, per question, per assistant, with history from day one.
The Tag
A script served from our domain, no cookies, no browser storage. It sees the visitors an assistant sends, the page they land on, and the phone and email taps and form sends made on that page.
Visitors from assistants, by assistant and landing page, with the contacts made on the page they landed on.
The Edge Tag
Crawlers do not run scripts, so a page tag never sees them. A small snippet at your edge reports each crawler hit as user agent, path and status, never an IP.
Answers served per assistant per page, and which index crawlers have fetched which pages. WordPress plugin and PHP middleware: coming.
Change orders
We read your site the way an assistant does, score it, and write each fix as a numbered copy-paste block for your developer. The method stays ours; the result goes on your site.
Open, applied or verified, with the reading each order should move.
Subject: Readings to 14 Oct 2026: answers served up 300% week on week READINGS answers served a day up 10x since ChatGPT indexed the site PAGES CITED home page 75% of answers; pricing, trace a debtor, reconnect with someone, for solicitors, developers, how it works COVERAGE Amazon Alexa 100%, Meta AI 97%, Bing 84% of the site read CLICK-THROUGHS the first from ChatGPT on 14 Oct, one to the instruction form NEXT one change order, with the exact text to paste
Every figure in the brief is the client's own, with its date range. Inferred questions are marked likely. How it works in full.
Type your domain. We read it the way ChatGPT does.
We fetch your robots.txt, llms.txt, sitemap.xml and home page live, as a browser, as ChatGPT-User and as GPTBot, and we check whether ClaudeBot and PerplexityBot are let in. If your edge blocks a crawler, it shows. You get a score out of 100 and the findings in plain words. No sign-up to see the score. Five checks an hour per visitor; results stay at a link you can send to your developer.
Reading your site as a browser, as ChatGPT-User, as GPTBot, as ClaudeBot and as PerplexityBot. Ten seconds or so.
We fetch from our server with the reader's user agent set; rules that check a crawler's address may treat the real crawler differently. Rate limits are per visitor address, hashed daily. More about the check.
Free to start. Three plans from £99 a month.
A free account shows you every answer served, who reads your site and who the assistants send, and once your site is connected it puts 25 of your buyers' questions to the assistants once, free. Watch adds whether the assistants name you, asked three times a week. Fix also writes the changes. Run applies them and writes one article a month. We never promise a position. We show you the readings.
Connect up to 3 sites with the Tag, the Edge Tag or an uploaded log, and see the weekly audit (up to 25 pages), answers served, coverage, readers, visitors from assistants and the Monday brief, at no cost; and once a site is connected, one free baseline round of the Monitor: 25 of your buyers' questions put to ChatGPT, Claude, Perplexity and Gemini, once per account.
or £990 a year, two months free
A business that wants to know whether the assistants name it when its buyers ask, and who they name instead.
- Everything in the free account, plus the Monitor: 25 of your buyers' questions put to ChatGPT, Claude, Perplexity and Gemini three times a week
- The questions built with you by us, from what you sell, where and to whom
- Named or not, in what position, and who is named instead, per question and per assistant, with history from day one
or £4,950 a year, two months free
A business that wants the changes written down exactly, so its developer or agency can apply them in an afternoon.
- Everything in Watch, with 50 questions asked three times a week and the weekly audit read to 250 pages
- Up to 5 numbered change orders a month, written by us for your developer: the exact text, where it goes, why, and which reading it should move
- Each order re-checked by us once your developer marks it applied, and shown as open, applied or verified in the dashboard
3-month minimum, billed monthly
A business or practice that wants the bench run for it, including the writing, and where one new client is worth more than a month of Run. One site of up to 50 pages; larger sites by quote.
- Everything in Fix, with 100 questions asked three times a week and the weekly audit read to 500 pages
- We apply the change orders ourselves through a named editor login on your CMS or as pull requests to your repository; you approve each change before it publishes and every change is listed in the dashboard
- One dated, answer-first article a month, written for your approval, published with FAQ schema and a Markdown twin
Up to 50 pages read the way the assistants read them (no JavaScript, as a browser, as ChatGPT-User and as GPTBot) and scored out of 100 across Reachable, Readable and Declared; a baseline of 300 answers (25 of your buyers' questions put to ChatGPT, Claude, Perplexity and Gemini in three rounds); up to 10 numbered change orders a developer can apply; delivered within five working days as a dated report page with a 30-minute call, and credited in full against the first month of Fix or Run started within 30 days.
The questions people ask the assistants about this. Answered first, dated, quotable.
Every guide opens with the answer and carries the date it was written. That is how we write for clients and it is how we write for ourselves.
What is answer engine optimisation, and how is it different from SEO?
SEO wins a position on a results page; answer engine optimisation wins a mention inside the answer, where there is often no link to click at all.
Why does ChatGPT not mention my business?
ChatGPT leaves a business out because it cannot read the site, because the site does not answer the question asked, or because nobody else on the web mentions it.
Do AI assistants read JavaScript websites?
An AI assistant reads the HTML your server sends and nothing more: if your words only appear after JavaScript runs, to the assistant your page is blank.
How do I know if ChatGPT has read my website?
Your analytics tool cannot see ChatGPT reading your website; the only record is a single line in the server access log, written at the moment a person asked a question.
How long does it take to get cited by an AI assistant?
Both of our new sites were fetched by an AI assistant to answer a real question inside a fortnight of going live (server logs, September and October 2026).
Which AI crawlers should I allow in robots.txt?
A robots.txt that blocks GPTBot but allows OAI-SearchBot keeps you eligible for ChatGPT's search results today and out of the model's memory tomorrow.
What is llms.txt and does it help?
llms.txt is a free, harmless map of which page answers which question; publish one, but do not expect it to get you cited on its own.
All guides. Also served to the assistants in llms-full.txt.
Put your site on the bench.
Run the free check, then open a free account: once your site is connected, the assistants are asked 25 of your buyers' questions on us. Readings arrive every Monday. Plans start at £99 a month.
Last reading 14 Oct 2026 09:30
Agencies: ask us about running Fix for your clients, [email protected].