AI answer engines are already quoting your client reviews back to prospects, and most firms have no idea which ones. A prospect types a question into ChatGPT or lands on a Google AI Overview, and the model stitches together a response from whatever text it can find under the firm’s name. A Google review from eighteen months ago. A one-star Yelp screed. A Reddit thread where somebody vented about a billing dispute. By the time that person opens the firm’s website, they’ve already read a sentence the firm never wrote and never approved.
That changes what a reputation actually is. It used to be an average score with some narrative underneath. Now it’s a sentence a machine may summarize on your behalf, and the factors that determine which sentence gets surfaced look very different from the reputation metrics firms are used to watching.
The Star Average Matters Less Than the Words Beneath It
The 4.7 at the top of the profile is comfort food. It’s not what gets quoted. Language models don’t average; they extract. When a prospect asks whether a firm is any good, the model reaches for a sentence that sounds like an answer: specific, recent, emotionally clear.
A measured four-star review saying “communication was fine, outcome was reasonable” gets passed over for a two-star review that names a paralegal, a missed deadline, and a dollar figure. That’s the pattern Search Engine Journal documented in AI Overviews: the surfaced reviews tend to be recent, specific enough to name features or people, published on platforms the model trusts, and echoed across more than one source.
A single detailed complaint can outrank forty generic compliments, because the complaint reads like information and the compliments read like noise. Agencies that handle law firm digital marketing increasingly spend their first week on an audit of what the models are already saying under a firm’s name.
The Review Record Extends Far Beyond Google
Firms pour most of their review energy into Google Business Profile. The models read wider than that. Reddit threads, Avvo pages, Yelp, Trustpilot, Better Business Bureau complaints, state bar discipline notices, old forum posts, local news comment sections. All of it is fair game for a summary.
The firm with 180 five-star Google reviews and a single ugly Reddit thread may find the Reddit thread quoted first, because the model treats a conversational, specific post on a community platform as higher-signal than a short compliment on a business listing. The practical move isn’t to chase every platform. It’s to know which ones exist under the firm’s name and to monitor them the way you’d monitor a court docket.
AI Summaries Are Moving Earlier in the First Impression
AI summaries are now the opening scene. Pew Research found a majority of users now encounter AI-generated content somewhere on a typical results page. For a legal prospect, someone already anxious and already comparing, the AI blurb is often the first full paragraph they read about a firm. The firm’s own homepage comes second, if it comes at all.
That reorders the marketing stack. The old sequence was ad, click, landing page, phone call. The new sequence often starts with a machine-written paragraph the firm never saw, and changing the inputs the models read is slower than most firms expect.
More Reviews Alone Won’t Change the Story
Volume helps, but not the way firms think. A review campaign that produces fifty short, generic five-star reviews moves the average and barely moves the summary. The models still reach for the specific, named, recent sentence, and if that sentence is negative, fifty “great firm, highly recommend” lines won’t crowd it out.
What actually shifts the summary is specificity on the positive side. A review that names the type of case, the attorney, and a concrete outcome gives the model something to quote that competes with the ugly sentence. A few practical moves:
- Ask for detail, not stars. Prompt the client to mention the practice area, who they worked with, and what the firm actually did. A line like “helped me through a difficult custody matter and explained every filing” is quotable in a way that “great firm” never will be.
- Respond to every negative review. A short, non-defensive reply gives the model a counter-sentence to pull from. Keep it factual, acknowledge the frustration, and skip the temptation to re-litigate the case in public.
- Watch Reddit and niche forums. These threads get weighted heavily by AI engines and are usually invisible to the firm. Set up alerts for the firm’s name on the platforms you don’t normally check.
- Clean up stale complaints where you can. Old BBB entries, resolved bar complaints, and dead forum posts sometimes allow updates or responses. Updating them changes what the model has to work with on its next crawl.
The Reputation Problem Starts Before the Review Is Written
It reads like a marketing problem, but it starts at intake. The reviews that end up quoted back are usually about the first phone call, the paralegal who didn’t follow up, the fee letter that wasn’t explained, or the week between the signed engagement and the first substantive update.
Marketing can shape how those stories get told online. It can’t manufacture the stories themselves. The firms that come out of the next two years with clean AI summaries will be the ones that treated the client experience as the source material, and the review profile as the edited version a machine is now reading aloud to the next prospect.
