The UK AI Visibility Report for Solicitors — July 2026

US research says AI answers about lawyers are dominated by legal directories. We measured 12,279 citations across UK solicitor searches. The UK looks very different — and, for individual firms, harder to win.

Update, 25 August 2026: the follow-up study — Most UK Solicitors Are Never Recommended by AI, Report TAI-R-2026-002 — is now published.

When a prospective client opens ChatGPT or Perplexity and asks "best conveyancing solicitor in Cardiff" — or Leeds, or Bristol — whose websites does the AI actually read to build its answer?

In July 2026 we stopped guessing and measured it. We ran a fixed panel of 68 prompts across 17 UK cities — four prompt types per city (best-in-city, purchase intent, reputation, and a practice-area specialism), with the specialism prompts spanning 7 practice areas (personal injury, criminal law, litigation, conveyancing, immigration, family law, and wills & probate) — against both Perplexity and ChatGPT — 1,360 answer runs, tracking over 1,200 SRA-regulated firms in the measured cities. We recorded every source each answer cited — 12,279 citations in total — and classified each one.

The prompt panel was fixed before any testing began and contained no firm names, so results could not be biased towards any individual practice. Every run returned citations — 0 of 1,360 answers were built from nothing.

74% of AI citations went to individual law firm websites — scattered across 596 different firms. The typical firm appeared in roughly 3% of the answers it could have appeared in.

Finding 1: The UK is not a "directory wall"

A widely shared US study this month reported that 48% of AI citations for "best lawyer" searches went to legal directories, with directories the top source in most queries.

Our UK data says otherwise:

Source type	Share of 12,279 citations
Individual firm websites	74.1%
Legal directories & review platforms	17.2%
Forums & social media	4.6%
Search (google.com)	4.1%
Media & reference	0.02%

Three out of four citations went to law firms' own websites. The US finding and ours aren't directly comparable — different countries, engines and query panels, and our sample is roughly forty times larger — but the headline conclusion doesn't transfer: in the UK, AI systems are reading firm websites, not routing around them.

That last row deserves a second look, too. Newspapers, the Law Society, the SRA and gov.uk received 3 citations between them — out of 12,279. Whatever gets a firm cited by AI, it isn't press coverage.

Finding 2: The visibility exists — but almost nobody owns it

If firm websites get 74% of citations, why doesn't it feel that way to any individual firm?

Because it's shattered across 596 different firm websites. The single most-cited firm appeared in just 26 of our 68 prompts. The typical SRA-regulated firm in our tracked set appeared in roughly 3% of the answers where it plausibly could have.

That is the honest picture. AI isn't recommending your competitors instead of you — for most firms, it's recommending almost nobody consistently. The opportunity isn't locked behind an unbreakable directory wall; it's lying on the table, thinly spread, waiting for firms that give AI systems something solid to cite.

Finding 3: ChatGPT and Perplexity are different worlds

The two engines don't just differ at the margins — they cite from almost entirely different webs.

Perplexity runs a mixed economy: mostly firm websites (68.3%), with a substantial directory layer (25.0%) — reviewsolicitors.co.uk was its single most-cited domain (682 citations, 8.1%) — plus forums and social (6.6%).

Domain	Perplexity citations
reviewsolicitors.co.uk	682
reddit.com	359
solicitors.com	226
duncanlewis.co.uk	194
gardnerchampion.co.uk	178
solicitorsup.co.uk	168
solicitor.info	149
lawyersolicitor.co.uk	124
legalrank.uk	113
samconveyancing.co.uk	101

ChatGPT cited a legal directory zero times in 3,812 citations. And forums zero times. Its answer diet was firm websites (86.8%) plus google.com (13.2%) — it resolves the question through search, then reads firms' own pages directly.

The practical implication: there is no single tactic that covers both. A firm can be visible on one engine and absent from the other, because the two systems trust different parts of the web.

Finding 4: Reddit is the UK's second-biggest source for solicitor recommendations on Perplexity

Anonymous forum threads — real people asking "can anyone recommend a solicitor in [city]?" — were Perplexity's second most-cited source (359 citations), ahead of every directory except reviewsolicitors.co.uk and ahead of every law firm in the country. What past clients say in public threads is shaping AI answers more than most firms' marketing is.

Finding 5: One two-office firm out-cited every large practice in the country

The most-cited individual firm in our entire dataset — 258 citations, appearing in 26 of 68 prompts, on both engines — is not a national brand. It's Gardner Champion, a two-office SRA-regulated firm based in Staffordshire and West Yorkshire.

How? We can't prove causation from a snapshot, but the pattern is visible on their website: they've built individual location pages for cities across the UK, each structured around the services someone in that city would ask about. When an AI system looks for a page that directly answers "solicitor in [city]", they have one and most firms don't.

Whatever else is true, this settles one question: the fragmentation in Finding 2 is not a law of nature. A small firm with the right structure can be cited more than firms fifty times its size.

Why AI behaves like this

The two engines embody two retrieval philosophies. Perplexity leans partly on aggregators — sites that compare many firms and carry reviews are easy for it to validate and quote. ChatGPT resolves the question the way a person would: search, then read the firms' own pages. In both cases the winning pages share a trait: they directly, specifically answer the question asked — a named service, in a named place, with concrete detail an AI can lift and attribute.

What surprised us

• We expected the US "directory wall" narrative to hold. It didn't — directories took 17% of UK citations, not 48%.
• We did not expect ChatGPT to cite legal directories literally zero times in 3,812 citations. It did.
• We did not expect Reddit to be the second most-cited source for UK solicitor recommendations on Perplexity. It was.
• We expected media coverage to matter. Three citations out of 12,279 says it doesn't.

What this is and isn't

A note on method, because solicitors are sceptical professionals and rightly so:

• This is a measured snapshot, not a causal study. It tells you who AI cites today, not what makes AI cite anyone.
• It covers two engines (Perplexity and ChatGPT), one profession (SRA-regulated solicitors), 17 cities, four prompt types per city, 7 practice areas among the specialism prompts, one time window (July 2026).
• Every figure above is a counted share of our sample, not an estimate or a statistic borrowed from elsewhere. Domain classification (firm vs directory vs forum vs media vs search) was reviewed by hand for every high-volume domain.
• We counted ourselves. tendorai.com publishes profile pages for UK regulated firms, so it was classified as a directory rather than excluded. Its citations are included in the 17.2% directory share.
• practice_area in the published prompt file is derived from the prompt text and template type, not a recorded field. Prompts are reproduced exactly as issued, including a grammatical error in the immigration prompts.

We're not going to claim this proves any particular tactic delivers "3x more enquiries". It doesn't. What it proves is where the attention actually goes.

This research was conducted by TendorAI, a UK platform that measures how visible regulated firms are to AI assistants. We'll re-run this same panel quarterly and publish the changes — including when the data says a popular tactic doesn't work.

What a firm can sensibly do with this

Three things follow directly from the data, none of them a quick hack:

1. Your own website is the main asset — if it's structured to be cited. Three-quarters of citations go to firm websites, but overwhelmingly to pages that answer a specific question in a specific place. Generic brochure pages don't get cited; specific service-and-location pages do.
2. Your review footprint decides your Perplexity visibility. reviewsolicitors.co.uk is the single most-cited domain in UK solicitor answers, and Reddit threads outrank every directory bar one. What clients say about you in public carries further than what you say about yourself.
3. Measure both engines, because they disagree completely. ChatGPT and Perplexity cite from different webs. A firm checking only one has half a picture — possibly the wrong half.

---

Correction, 02/08/2026: an earlier version of this report described the panel as "68 prompts across 17 UK cities, covering 7 practice areas". The 7 practice areas apply to the specialism prompts (17 of the 68), not to the full panel. The panel size, run count and all citation figures are unchanged.

Limitations

• Two engines. ChatGPT and Perplexity only. Google AI Overviews reaches more UK users than either and is not represented here. These findings should not be generalised to AI search as a whole.
• One snapshot. A single collection date. AI citation behaviour changes; this is a point-in-time measurement, not a trend.
• One jurisdiction, one profession. SRA-regulated solicitors in England and Wales. Whether the same pattern holds for accountants, mortgage advisers or estate agents is untested.
• Constructed prompts. These are plausible client queries, not observed ones. We do not have access to what UK consumers actually type into AI assistants.
• Classification is imperfect. Distinguishing a small firm's website from a small directory is a judgement call at the margins. The most-cited domains were reviewed by hand; the long tail was not.

Data availability

Everything needed to check this study is published alongside it:

• The deviations log — every departure from the pre-registered plan, including corrections to our own errors
• The full 68-prompt panel (CSV)
• The city and firm-count panel (CSV)
• The domain classification list (CSV)

Raw AI responses and the full set of cited URLs from all 1,360 runs are retained and can be interrogated for audit purposes. They are not published in full, as they contain a large volume of third-party firm data collected without those firms' involvement.

How to cite this report

Reproduction

This report is published ungated, with no registration required. Reproduction is permitted with attribution to TendorAI and a link to this page. Figures, findings and charts may be quoted in full.

Free distribution is deliberate. Gated research cannot be cited.

Does TendorAI have a commercial interest in this result?

Yes, and it is worth stating plainly. TendorAI sells AI visibility software to UK regulated firms. The central finding here — that firm websites, not directories, dominate the citations we measured — was not the result we expected when we designed the study, and it cuts against a widely repeated narrative. It is published because it is what the data showed. Readers should weigh it accordingly, and the published prompt panel and classification list exist so that they can.

---

TendorAI measures how visible UK regulated firms are to AI assistants, identifies why they're not being recommended, and tracks whether it improves. See where your firm stands: /ai-visibility-report