We audited 8,074 UK law firm websites for AI search. One is ready.

When someone opens ChatGPT and asks for "a good solicitor near me" or "a conveyancing firm in Leeds that handles leasehold", the assistant does not hand back a page of ten blue links. It names two or three firms. To be one of them, a website has to clear the top band of our 0 to 100 AI visibility scale, which we call AI-Ready and which starts at 80. We pointed that audit at 8,074 UK law firms, drawn from the SRA-regulated population via public listings: independent high-street firms alongside the large national names. Here is the headline.

One firm cleared the bar. Connaught Law, a small independent, scored 80.9 and became the first AI-Ready business in any edition of this index. The cohort mean sits at 50.6 and the median at 52.3. Behind that single success, 1,757 firms (22%) score below 40, the point at which an AI can barely use the site at all. A further 48 firms sit at 75 to 79.9, one push from the line, and not one of them has taken it. No firm anywhere reaches Verified, our 90-plus mark.

That distribution is the story of law firm AI search in one table: a huge, undifferentiated middle, a fifth of the profession effectively unreadable, and a single firm over the line. But the interesting question is not who won. It is what separates the firms an AI can read from the firms it cannot. The answer turned out to have almost nothing to do with marketing.

The JavaScript trap

1,072 firms of the 8,074, one in eight, have a website built as a JavaScript app. That is a technical decision, usually made once, usually by an agency or a developer, and usually never revisited. It is also the single most damaging thing we found in the data.

Read the last column again. Four firms out of 1,072 built in JavaScript reach Strong. Among everyone else, 31.2% do. That is roughly 78 times less likely. There is no other variable in this dataset, not size, not budget, not city, not the age of the firm, that moves the odds anything like that far.

It gets worse when you look at the bottom of the table. 905 of the 1,757 effectively invisible firms are JavaScript-built. More than half of the law firms in Britain that AI simply cannot see share one root cause, and it is not the content.

Why it happens, in plain English

An AI reader is not a browser. It fetches the page and reads what comes back. That is the whole mechanism. A browser will happily download an empty page, run a few hundred kilobytes of JavaScript, and assemble your firm's name, your solicitors, your practice areas and your phone number on screen a second later. The AI reader does not wait for that. It takes what the server sent and moves on.

So if the words only appear after JavaScript runs, the machine sees an empty shell where the firm should be. No name, no location, no practice areas, nothing to quote. The legal work is identical. The lawyers are identical. The page just never renders for the reader that matters. This is why JavaScript SEO for law firms is not a niche developer concern: on this evidence it is the concern.

The uncomfortable part is that the site looks perfect to everyone who checks it. The partners see it. The marketing team sees it. The client on their phone sees it. The only visitor who sees nothing is the one deciding whether to recommend you, and it never files a complaint.

The one block of markup that decides it

If JavaScript decides whether the machine can read the page at all, JSON-LD decides whether it understands what it read. JSON-LD is a small block of structured data in the page source that states, in a fixed machine-readable format, what this organisation is: the name, the address, the practice areas, the people. It is invisible to clients and takes an afternoon to add.

Just over half of UK firms, 54%, publish it. They average 59.8. The 46% that do not average 39.7. A twenty-point gap on the presence or absence of one snippet.

The distribution is more striking than the average. 99.7% of Strong firms have JSON-LD. Just 1.3% of the weakest do. Of all 8,074 firms in the cohort, only 7 reach Strong without it. Seven. That is not a ranking factor among many, weighed against a dozen others. It behaves closer to a switch: on, and you are in the running; off, and almost nothing else you do rescues you.

Which makes the area scores painful reading. Structured data for law firms is the weakest of all eight areas we measure, at 30.3, while the plumbing that lets a crawler in is the strongest at 73.5. The profession has built the door and left the shelves bare.

Source: SearchScore SAVI UK Law Firms 2026. 8,074 firms scored across 250+ AI search visibility signals. "AI-Ready" is a score of 80 or higher; "Strong" is 60 or higher. Area figures are cohort averages out of 100.

The biggest firms are not the ones AI can find

If AI visibility were bought, the giants would own it. We tagged the 62 largest national and City firms inside the cohort. They do score above the field: 55.1 against 50.6. Their budgets buy a better website, and it shows.

Not one of the 62 is AI-Ready. The only firm over the line in Britain is a small independent.

The diagnosis is right there in the last two rows. On everything money buys, the giants win comfortably: technical quality 81.5 against 73.5, brand authority 56.6 against 37.6, depth of content 79.1 against 63.1. Then the pattern stops dead. On structured data they score 31.3 against the field's 30.3, which is to say no better than a two-partner firm above a shop. On AI citability they score 68.8 against 72.1, slightly worse.

They bought every signal a person notices and skipped the only one a machine needs. The engines do not read a balance sheet or a Chambers ranking. They read a website. A small firm that labels its work properly beats a national firm that does not, and in this cohort exactly that happened.

What else we found

Four findings that did not fit the pattern, and one that should worry the profession.

• 305 firms actively block ChatGPT's crawler. They disallow GPTBot in robots.txt. 309 block Google-Extended and 300 block ClaudeBot. They average 45.7. A few hundred UK law firms have opted out of AI search on purpose, which is a legitimate choice, but it is worth knowing whether anyone at the firm made it deliberately or a default template made it for them.
• London has no advantage at all. The capital averages 50.5 against the 50.6 national figure. Edinburgh and Liverpool lead the country on 52.4; Glasgow (48.2) and Birmingham (48.7) trail. The legal heartland is exactly as invisible as everywhere else.
• Conveyancing is the least AI-visible area of law, at 48.4, and it is the highest-volume consumer legal search there is. The work most people ask the internet about is the work the internet can least describe. Immigration leads on 54.5.
• LLPs average 55.6 against 48.7 for firms carrying no such marker, which is the closest thing to a structural advantage in the data, and it is still nowhere near the line.

What a firm should actually do

The fixes follow straight from the data, in order of leverage. None of them is a growth hack, and most are a developer ticket rather than a campaign.

• Make the page work without JavaScript. Render the content server-side, or pre-render it, so the words are in the HTML the server sends. If your site is one of the 1,072, this is not one improvement among several. It is the improvement, and everything below it is wasted until this is done. Ask your developer one question: does the firm's name, address and practice areas appear in the page source with JavaScript turned off?
• Add JSON-LD structured data. Mark the firm up as an Organisation or LegalService with its address and practice areas. 99.7% of Strong firms have it and only 7 firms in the country get to Strong without it. This is the switch.
• Name your solicitors, with credentials, in machine-readable form. Not just a photo grid on a team page a human can read. The structured version is what lets an assistant confirm that a real, regulated professional stands behind the advice.
• Publish the real questions clients ask and answer them plainly: what conveyancing costs, how long probate takes, what happens at a first hearing. That is the text an assistant lifts when someone asks it the same thing.
• Do not block the AI crawlers unless you have decided, deliberately, that you want no part of AI search. Check your robots.txt; 305 firms are blocking GPTBot and some of them did not choose to.
• Earn genuine reviews and citations on sources an AI already trusts, to lift brand authority off the floor at 37.6.

Notice what is not on that list. None of this is law firm SEO in the old sense: no keyword density, no link-buying, no monthly retainer for blog posts nobody reads. Getting a law firm into ChatGPT's answer is a question of whether the machine can read the page and identify the firm on it. That is closer to plumbing than to marketing, which is exactly why it keeps falling between the developer and the marketing team, with neither believing it is theirs.

The gap between the firms at the top and the rest of the profession is a few labels and a rendering decision, not the quality of the legal work. That is the whole finding. GEO for law firms in 2026 is not an arms race; it is a queue nobody has joined.

Frequently asked questions