15 questions people ask AI before hiring a law firm, and what answers them
Before a client ever opens a law firm's website, they are increasingly asking an AI assistant to shortlist one for them. I checked whether 18 real Indian law firm homepages are built to answer the 15 questions that shortlisting actually runs on.
Founders, general counsel and individuals looking for a lawyer are starting the search with an AI assistant, not a search engine results page. They ask a specific question, ChatGPT or Perplexity or Google's AI Overview reads what it can find, and it names a small handful of firms in the answer. A firm that never gets read never gets named, no matter how good the actual legal work is.
I took 15 real questions a client might type into an AI assistant while looking for legal help, and checked what a firm's website would need to say, and in what format, for an engine to answer with that firm's name. Then I scanned 18 real Indian law firm homepages to see how close they come.
The 15 questions, and the fix for each
List your practice areas by name on the homepage, not buried in a PDF brochure, and describe fundraising work in plain sentences an engine can lift.
Say it in words, not a PDF.
Name your M&A partners on a page that states the city and the deal types you have closed, so location and expertise sit together.
Pair the city with the practice, on the same page.
Spell out regulatory practice areas by their actual names (SEBI, competition law) instead of a vague 'corporate advisory' label.
Name the regulator, not just corporate law.
Publish a clear list of practice areas as structured data, so an engine reads the list as data rather than guessing from a paragraph.
Turn your practice list into machine readable data.
State billing structure plainly on the site, even a range, since most firms leave this blank and AI treats blank as unknown.
Blank fee pages mean AI has nothing to quote.
Give every partner a bio page with named practice areas, so a specific query maps to a specific person.
Name the person, not just the department.
Describe cross-border matters in plain language on the practice page, including which countries, since AI cannot infer this from a row of flag icons.
A flag icon is not a fact an AI can cite.
State the consultation policy directly on the contact page instead of requiring a call to find out.
Don't make the AI call you to find out.
List the specific courts and jurisdictions the firm appears before, not just a claim of pan India presence.
Name the court, not just the country.
Separate litigation and advisory work into clearly labelled sections so the distinction is explicit, not implied.
Don't make a reader infer what you actually do.
State headcount and office cities plainly, since vague language like 'a growing team' gives an engine nothing to extract.
Vague size claims give AI nothing to quote.
Publish real matter summaries or case studies with enough specifics that an engine can match them to a similar query.
A results page needs specifics, not just a client logo grid.
Name IP and trademark work explicitly as a practice area rather than folding it into a general commercial bucket.
Bucket terms hide the practice areas clients search for.
Put a named contact with a direct email next to each practice area instead of one generic inbox for the whole firm.
One generic inbox is a dead end for AI and clients.
State the confidentiality and privilege policy plainly on the site so it's not a question a prospective client has to ask first.
Say it up front, don't make it a first question.
What I found scanning 18 law firm sites
Scanning the 18 sites against my AI readability engine, the median score came out at 59.5 out of 100, with a hard split by firm size: large national firms scored a median of 83, the mid-size and boutique firms scored well below that.
Only 7 of the 18 firms publish any structured data at all, the format that lets an engine read facts directly instead of guessing from prose. Not one of the 18, zero, has FAQ schema, the format built specifically to answer a client's question in a way an engine can lift and quote.
That gap shows up directly against the 15 questions above. Most firms describe their practice areas somewhere on the site, in prose. Almost none state fees, jurisdiction, or consultation policy in a format an engine can extract with confidence.
The pattern under all 15
The pattern across all 15 questions is the same one. The expertise described on these sites is often real, a firm really does handle SEBI matters or cross-border deals, but it usually lives in a paragraph of prose or a PDF brochure, not in a machine-readable block an AI engine can extract cleanly.
And with FAQ schema sitting at zero across the entire sample, no firm in this vertical is currently answering client questions in the one format built for exactly that job. The firm that fixes this first has open ground, not a queue of competitors to catch up to.
Frequently asked questions
Do AI assistants really recommend specific law firms by name?
Yes, when they can find enough on a firm's site to be confident about who it is and what it does. If a page does not state facts clearly, in a format the engine can extract, the engine tends to skip it in favour of a firm whose site makes the answer obvious.
Which of these 15 questions matters most to fix first?
Start with whatever your site currently answers worst. But across the 18 firms I scanned, the biggest single gap was structured data describing practice areas, since it affects several of these questions at once rather than just one.
Is this only relevant to firms chasing new clients through AI search?
It matters even for firms that get most of their work through referrals, because a prospective client or a referring contact will often check an AI assistant's opinion before making the introduction. A weak result there is a quiet cost even when it is not the primary channel.
How is this different from the law firm scan study on this site?
The scan study reports the aggregate numbers: median score, the size split, structured data and FAQ schema rates across 18 firms. This page takes the same scan and reframes it around 15 specific questions a real client would ask, and what a site would need to say to answer each one.