Updated
Updated What changed on
Fact-checked Google's structured data and FAQ rich result guidance, the Ahrefs schema test, and the OpenAI, Anthropic and Perplexity crawler rules were re-checked against the 14 October 2026 source checks. Nothing else on the page was re-checked, including the study's own May 2026 figures.
- Removed the claim that firms without schema are invisible to AI search: Google says structured data is not required for its AI features, but it still helps pages qualify for rich results.
- Added Ahrefs' May 2026 test, where adding JSON-LD gave no citation lift in AI Mode or ChatGPT, and a small decline in AI Overviews.
- Corrected the FAQ schema section: Google stopped showing FAQ rich results on 7 May 2026, and says no special schema is needed to appear in its AI features.
- Corrected the crawler wording: the study tested GPTBot, which only controls OpenAI training. To appear in ChatGPT search, a site must allow OAI-SearchBot.
- Changed the crawler check to the search crawlers that matter: OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot.
- Added that Google's own May 2026 guide says optimising for AI Overviews is still SEO.
- Updated the OpenAI crawler link to its current address.
UK law firms are going missing from AI search
The 30-second version
I checked the homepage of 73 UK law firm websites. Nearly a third had no schema identifying them as a business or organisation: 22% had no structured data at all, and only 1 in 4 used schema that says they are a law firm. Google says structured data is not required for its AI features, but it still helps pages qualify for rich results. Six firms loaded fine in a browser but blocked OpenAI's GPTBot training crawler at the server, almost certainly without choosing to. Here is the full data, how I measured it, and how to check your own firm.
Why I ran this
Law firms have spent years getting their websites to rank in Google. That work is not wasted. But the way people find a solicitor is changing. They ask ChatGPT for the best employment lawyer in their city. They read Google's AI Overview instead of clicking a result. They let Perplexity shortlist three firms for them.
Those engines do not rank ten blue links. They read the web, decide which sources to trust, and name a few. If a firm's website is hard for them to read, it does not get named. I wanted to know how ready UK law firm websites actually are for that shift, so I measured it.
How the study was done
- Sample
- 73 UK law firm websites, taken from firms ranking on Google's first page for "solicitors" or "law firm" across 10 cities: London, Manchester, Birmingham, Leeds, Bristol, Liverpool, Sheffield, Newcastle, Nottingham and Glasgow. These are SEO-active firms, not a random draw. If visible firms struggle, the long tail is worse.
- What I checked
- Each firm's homepage. I read its JSON-LD structured data and recorded the schema types. I requested the homepage as a normal browser, as OpenAI's GPTBot and as PerplexityBot, and recorded what each got back.
- Date
- 22 May 2026.
- Limits
- This is a homepage-level snapshot, not a full-site crawl. 8 of the 73 sites returned a forbidden or authentication response to the crawler, usually aggressive bot protection, and were left out. All percentages below are from the 65 sites that served a readable homepage. The FAQ figure is homepage-specific.
Finding 1: one in five firms has no schema at all
Structured data, or schema, is the machine-readable layer of a website. It is how a page states plainly what it is: this is a law firm, here is its name, here is where it is, here are its people. Google says structured data is not required for its AI features, but it still helps pages qualify for rich results.
Of the 65 firm homepages I could read, 14 had no structured data at all. Not a single line. To an AI engine, those pages are a wall of text with no labels. One national firm with immigration and personal-injury practice areas had over 3,400 words on its homepage and zero schema. The words are there for a human. A machine has to infer everything.
The 25% figure is the most telling. Schema has a type built for exactly this: LegalService. Used well, alongside Organization, Person and a postal address, it gives search engines a clean set of connected facts. The firm, its solicitors, its location, all linked. Three quarters of the firms I checked did not use it. The best ones did, and it showed: their markup read like a tidy business card. Do not expect schema alone to win AI citations, though. In Ahrefs' May 2026 test, adding JSON-LD gave no citation lift in AI Mode or ChatGPT, and a small decline in AI Overviews.
Finding 2: almost nobody uses FAQ schema
AI engines pull answers. A question with a clear, self-contained answer is the easiest thing for them to lift and quote. Google stopped showing FAQ rich results on 7 May 2026, and says no special schema is needed to appear in its AI features. FAQPage remains valid schema.org markup.
Of 65 law firm homepages, exactly one carried FAQ schema. The markup matters less now. The questions still matter. Clients ask the same questions before they ever call: how much does a divorce cost, do I have a claim, what happens at a first meeting. A firm that answers those in plain language on the page gives clients the answer directly. Google says no special schema is needed to appear in its AI features.
Finding 3: some firms block OpenAI's crawler without knowing
This one surprised me. A firm can reasonably decide whether to let AI crawlers read its site. That is a real choice. What I found was firms not making the choice at all.
One firm used its robots.txt to formally tell AI crawlers to stay out. A clear, recorded decision.
Six firms loaded fine in a browser but their server or security layer dropped OpenAI's GPTBot crawler. Nothing in robots.txt said to. Almost certainly nobody chose this.
That second number is the quiet problem. These six firms have working, often well-built websites. A human visitor sees everything. But when GPTBot comes to read the page, the hosting or security layer kills the connection. GPTBot only controls whether OpenAI trains on the content. To appear in ChatGPT search, a site must allow OAI-SearchBot, which this study did not test. A silent block on either gives no warning and shows no error in any dashboard. It is the kind of fault you only find by testing for it directly.
This is not rare or exotic. I found the same fault on my own site while preparing this study: the host's security layer was silently turning away GPTBot. robots.txt said one thing, the server did another. The only way to know is to test what the server actually does.
What this costs a law firm
None of this shows up in a normal SEO report. A firm can rank on page one of Google, have a tidy site, and still be missing from the answer ChatGPT gives when someone asks it to recommend a solicitor. The two are measured differently and they are drifting apart.
For a law firm the stakes are specific. Legal work is high-value and high-trust. A client choosing a solicitor for a £400,000 house purchase, a contested probate or an employment dispute is exactly the kind of person who now asks an AI engine to shortlist for them. If a firm is unreadable to that engine, it is not on the shortlist. Adding schema alone will not change that. In Ahrefs' May 2026 test, adding JSON-LD gave no citation lift in AI Mode or ChatGPT, and a small decline in AI Overviews.
The encouraging part: most of this is fixable, and quickly. Schema is a one-off technical job. A crawler block is a hosting setting. Neither needs a website rebuild. Start with crawler access: OpenAI says sites that block OAI-SearchBot will not be shown in ChatGPT search answers.
How to check your own firm in 10 minutes
Check your structured data
Run your homepage and one practice-area page through a schema checker. Look for LegalService or Attorney, plus Organization and Person. If all you see is WebSite and WebPage, your schema does not say you are a law firm. Use the free AI schema validator.
Test what AI crawlers actually get
robots.txt is not enough. Test the live server response too. For AI search visibility, check OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot. GPTBot and ClaudeBot only control training. If your site loads for you but drops a search crawler, that engine can leave you out without any warning. Use the AI bot robots.txt tester.
Ask the engines a client question
Open ChatGPT and Perplexity. Ask what a client would: "best employment solicitors in [your city]" or "recommend a conveyancing solicitor near [town]". See whether your firm is named, and which firms are. That is your real AI search ranking.
Score the page
Run a practice-area page through an AI visibility check to see how extractable it is: schema, answer structure, author signals, freshness. Use the free AI search visibility tool.
If those four checks come back clean, the firm is ahead of three quarters of the profession. If they do not, the fixes are smaller than they look. Schema and crawler access are the groundwork of generative engine optimisation, and they are where I would start on any GEO audit for a legal client.
Common questions
Is page-one Google ranking enough to show up in AI search?
Should a law firm block AI crawlers?
How much work is it to fix a firm's AI visibility?
LegalService, Organization and Person schema is a one-off technical task, not a rebuild. Fixing a crawler block is a hosting configuration change. Answering common client questions on the page takes a little writing. None of it requires a new website. The slow part is usually finding out the problem exists.Will you check my firm's site?
Find out where your firm stands
I will run these checks on your firm's site, read how AI engines currently treat it, and send back a plain list of fixes in priority order. No retainer, no sales pitch.