When a prospective client asks ChatGPT, Perplexity or Google’s AI Mode to recommend a solicitor, the firms it names are chosen in the moment, from whatever it can find about them online. Client reviews are one of the main things it finds. This holds across practice areas. Whether the question is about a divorce, a commercial dispute, a house purchase or a clinical negligence claim, firms with a large, recent body of reviews on the platforms AI tools read are named more often, and firms with only a handful of reviews are mostly left out, regardless of the quality of their work.
BrightLocal’s 2026 Local Consumer Review Survey found that 45% of consumers had used an AI tool for a local business recommendation in the past year, up from 6% the year before. That figure is from a US survey, so UK adoption is likely lower. Legal services are the kind of high-stakes, unfamiliar purchase where people want a shortlist and a reason for each name on it, so the behaviour is worth planning for now.
We have written before about why review collection should be a never-off marketing activity and how to collect reviews across multiple platforms. This post covers the mechanism: what AI tools do with reviews when they recommend a law firm, which review signals the research shows actually make a difference, and what that means for how a firm should collect them.
1. How AI search builds a shortlist of solicitors
An AI tool does not hold a list of recommended law firms. When it is asked for the best solicitor for a particular problem (“who is the best solicitor for a contract dispute in Manchester?”, for example), it runs several web searches, reads the pages it finds, and writes an answer from them. The firms it names are the ones that appeared in those sources, and the reasons it gives are taken from what those sources say about them.
Whito’s August 2026 study asked ChatGPT, Gemini and Google AI Mode for the best solicitor for a business in six UK cities and kept the screenshots. ChatGPT opened each answer with a map and a set of business cards showing each firm’s star rating and review count, then wrote a ranked shortlist with citations to Chambers. Google AI Mode built a comparison table and cited ReviewSolicitors in its first paragraph, placing one Manchester firm top on the strength of a 95%+ satisfaction rating there. Gemini said its London shortlist was made up of firms highly rated by Chambers and The Legal 500.
The recommendation is therefore assembled from three layers: Google Business Profile data (star rating and review count), third-party directory and review pages, and the firm’s own website. Reviews sit in the first two layers, and they are often the stated reason a firm is placed where it is.
The answer is also generated fresh each time. In Whito’s test, 74% of the 61 firms named were named by only one of the three engines, and the authors note that a different day would produce a different list. That variability is a reason to be present in as many of the sources as possible rather than to optimise for one tool.
2. Where client reviews enter the answer
Reviews reach AI answers through three routes, and the routes differ by tool.
- Google Business Profile ratings and review counts. These feed the business cards ChatGPT shows and the local results inside Google AI Mode. Google states that review count and review score factor into local ranking, and Whito’s October 2026 follow-up, which compared AI Mode’s picks with Google’s local results across a range of service businesses, found that the businesses AI Mode named mostly sat near the top of the local results and had more reviews than those it did not name. A firm that ranks well in the Map Pack for “commercial litigation solicitor Manchester” is already most of the way to being named by AI Mode for the same question.
- Review platforms cited as sources. TendorAI’s July 2026 study ran 68 prompts across 17 UK cities on ChatGPT and Perplexity and classified all 12,279 citations the two tools produced. ReviewSolicitors was Perplexity’s single most-cited domain, at 682 citations, or 8.1% of everything it cited. Google AI Mode cited ReviewSolicitors directly in Whito’s screenshots. Trustpilot appears less often for legal queries; the sector-specific platform is the one AI tools reach for first.
- Forum threads where past clients recommend firms. Reddit was Perplexity’s second most-cited source in TendorAI’s data, at 359 citations, ahead of every directory except ReviewSolicitors and ahead of every law firm website. A thread where someone asks “can anyone recommend a solicitor in Leeds?” and gets named answers is the same kind of third-party evidence as a review, and Perplexity cites it heavily.
The tools do not agree on sources. In TendorAI’s run, ChatGPT cited no directories or review platforms at all and built its answers from firm websites and Google search results, whereas a month later Whito saw it citing Chambers and showing Google review counts in its answer cards. Which review source feeds which tool changes between model versions and between weeks, which is the main reason relying on a single review platform is a risk.
3. Which review signals actually make a difference
Across the studies published this year, the pattern is consistent: review volume separates the firms that get named from the firms that do not, and star rating does not.
- Volume. Insites’ 2026 AI Visibility Report analysed 10,000 local businesses against ChatGPT and Perplexity. Businesses recommended by both tools averaged 133 Google reviews; those invisible to both averaged 11. Google review volume, Google Business Profile completeness and Local Pack presence were the three strongest correlating signals in the whole study. An 8ight study of 2,976 businesses found ChatGPT named 3.4% of businesses with fewer than 25 Google reviews and 20.9% of those with 250 to 999. In Whito’s legal study, where an engine quoted review data alongside a recommendation, the median review count was 113.
- Rating as a threshold rather than a differentiator. In the 8ight data the median star rating was 4.9 for named and unnamed businesses alike. Whito’s quoted ratings ran from 4.0 to 5.0. A firm needs a rating that does not raise a flag; above that level, more reviews does more than a higher average.
- Recency. AI tools read current pages, and recent reviews are the evidence that a firm is still doing the work it is being recommended for. A profile whose last review is 14 months old does less for the firm than a smaller profile with reviews from the last quarter.
- The specifics clients mention. The reasons AI tools give are taken from the sources. Google AI Mode’s stated reason for its top Manchester pick was the firm’s satisfaction rating on ReviewSolicitors. When reviews name the type of matter (“handled our shareholder dispute”), the solicitor (“Sarah in the commercial team”) and the location, the tool has the words it needs to match the firm to a question about that practice area, that person or that city. Generic praise (“great service, would recommend”) adds to the count and gives the model nothing to match against or quote.
- Named solicitors. ReviewSolicitors publishes profiles for individual solicitors as well as firms, and questions such as “who is the best solicitor for a contract dispute?” are often answered with people rather than firms. A review that names the solicitor builds that person’s profile, which is a separate route into the answer from the firm’s own profile.
4. What this means for how your firm collects reviews
The practical changes follow from the signals above.
- Treat volume as the target. Across the studies above, the more reviews a firm has, the more often it is named in the answers AI tools write about solicitors. The way to build that volume is a review request at the close of every matter, from every fee earner, with a follow-up a fortnight later. We set the process out in our post on never-off review collection.
- Spread reviews across the platforms the tools read. Google Business Profile feeds ChatGPT’s answer cards and AI Mode’s local results. ReviewSolicitors feeds Perplexity and AI Mode’s citations. Trustpilot carries general consumer recognition, and Legal 500 and Chambers client feedback matters where the firm’s positioning supports it. Our multi-platform guide covers how to run this without the process drifting.
- Ask for specifics. The request template should prompt the client to mention what the matter was, who acted for them and where. Fee earners should not script the review. A line such as “it would help us if you could mention the type of matter and who you worked with” is a reasonable ask, and it changes what ends up on the page.
- Keep the Google Business Profile complete. Insites found profile completeness was one of the three strongest signals alongside review volume. Categories, services, opening hours, a working website link and a description that names the practice areas all affect whether the profile is pulled into an AI answer, and whether the answer describes the firm accurately.
- Respond to every review. Google says helpful replies help a business stand out, and a visible response pattern is read by any tool that opens the page. Responses need to stay within client confidentiality, so a short, considered reply to each review is enough.
- Notice forum mentions. A firm should not seed recommendations on Reddit or similar forums. What it can do is keep an eye on where it is mentioned, and make sure the Google and ReviewSolicitors profiles someone finds afterwards back up what was said.
How does your firm compare?
If you'd like to find out which firms AI tools recommend in your practice areas, and how your firm's reviews compare, speak to our team. Call us on 0161 402 3170 or get in touch using our online form.
