For fifteen years, the conventional wisdom on legal directories has been that they are a tax — a thing your firm shows up in because someone else linked to it, with marginal returns at best. That conventional wisdom was correct, for Google.
It is wrong for the AI engines, and increasingly so. In our Q1 panel of fifty New York personal injury firms (No. 001), four directory domains accounted for 67% of every citation the engines returned. Firms' own websites? 11% on average, 3% on Perplexity. The shift is not marginal.
The reason is mechanical. Large language models prefer structured data over prose, and they prefer authoritative third-party structured data over self-described prose. A law-firm "About Us" page reads, to the model, like marketing — to be quoted from cautiously, if at all. An Avvo profile reads like a record — practice areas, bar admissions, years licensed, peer endorsements, in fixed fields, attached to a domain the model has been trained to trust. Given a question like "who's a good Bronx personal injury lawyer?", the model will reach for the record before the marketing every time.
Which means: your directory profiles are not the tax. They are the primary marketing surface. They are where your firm gets described to prospects who never see your website.
§ 02 — The six directories that matter, rankedNot every directory is equal — and the engines weight them differently.
Across the panel, the AI engines reached for the same six legal directories with overwhelming consistency. They are not interchangeable. Here is which engine weights which directory, scored on what we observe in actual citations:
| Directory | ChatGPT | Claude | Perplexity | Gemini |
|---|---|---|---|---|
| Avvo | H | H | H | M |
| Justia | H | H | M | M |
| FindLaw | H | M | M | H |
| Lawyers.com | M | M | M | H |
| Super Lawyers | H | M | L | H |
| Expertise | M | L | M | M |
What jumps out: Avvo is the closest thing to a universal currency. Super Lawyers is a ChatGPT favorite — Thomson Reuters' control of the domain and its long history mean the model trusts it heavily — but Perplexity barely sees it. Lawyers.com is overweighted by Gemini for reasons we suspect relate to Google's first-party indexing of the property.
The takeaway is not "fix everything everywhere." It is: fix Avvo and Justia first, because they are the two listings that pay off across all four engines, and they are the two most likely to be unclaimed or stale.
§ 03 — What to actually fix, directory by directoryThe AI engines parse Avvo's structured fields aggressively. The ones that show up in citations: practice areas (use specific sub-areas, not "Personal Injury" alone), bar admissions and dates, peer endorsements, and the "About" narrative (the model will quote from this if it's specific and concrete).
The single biggest unclaimed-listing mistake we see: an old solo profile from 2014, never claimed, still pointing to a dead address. The model cites it. The prospect sees it. The phone never rings.
Justia is unusual among directories: it is free, it is comprehensive, and it is parsed with enthusiasm by ChatGPT and Claude. The fields the model values: jurisdiction-specific practice descriptions, education, professional associations, and (crucially) publications. A firm with three blog posts attached to a Justia profile gets cited more than a firm with none.
The unclaimed problem is less severe here than on Avvo, but the "filled out 30%" problem is rampant. Half-finished Justia profiles are the lowest-hanging fruit in the entire vertical.
FindLaw's "Lawyer Directory" remains heavily weighted on ChatGPT and Gemini. The catch: FindLaw monetizes premium placements, and the unpaid profile is genuinely thinner than the paid one. We do not recommend the paid placement on first pass — start with the free profile, complete it fully, and revisit the paid tier only if the audit shows FindLaw is still a gap after the free fix.
The "areas of practice" taxonomy on FindLaw is its own beast. Use the most specific terms the form will accept; the engines treat "Construction Accidents" as more authoritative than "Personal Injury - General."
Lawyers.com is the directory most directly indexed by Gemini in our data, which makes it disproportionately important if your prospects use Google's AI Overview. Same parent company as Martindale-Hubbell, and the rating system carries over — the AV-rated firms in our panel were cited 2.4× more often on Gemini than the unrated ones.
The peer-review system is where the work is. Solicit reviews from co-counsel and opposing counsel; the engines weight peer opinions more heavily than client testimonials.
Super Lawyers is the only directory on this list that you cannot fully control by claiming a profile — listing requires nomination and selection. But for firms that are listed, the citation lift is striking: ChatGPT cited Super-Lawyers-listed firms in our panel 1.8× more than non-listed peers on equivalent queries.
If you're listed, the work is making sure your profile bio is complete and that the listing is linked from your firm website with a permanent URL (not just a logo). If you're not listed, the nomination cycle is annual; ask co-counsel to nominate you next round.
Expertise has crept up on us. It is younger than the other five and was barely cited a year ago. Its share has climbed every quarter since, and on Perplexity it is now nearly co-equal with FindLaw. The structured-data quality is high and the model treats it as authoritative.
Watch this one. If you have time, claim and complete; if you don't, mark the calendar for the next quarterly review.
The unclaimed Avvo profile is the single most expensive thing in legal marketing.
If your paralegal has one afternoon, this is the order.
This is the script we hand a firm's paralegal when an audit surfaces directory gaps and the firm wants to move quickly. Total time, on average: 90 minutes for the highest-leverage 80% of the work, plus a long tail of peer-review requests that take care of themselves over the following weeks.
- Minute 0–10 — Inventory. Search every directory by your firm name and managing partner name. Note which profiles exist, which are claimed, which are stale.
- Minute 10–55 — Claim and update Avvo. Address, phone, practice areas (specific sub-areas), bar admissions, peer endorsements, and the "About" narrative. Be concrete; the model quotes specifics.
- Minute 55–85 — Claim and update Justia. Fill every field. Education, associations, publications. Cross-link from your firm site.
- Minute 85–90 — Send three peer-endorsement requests on Avvo and three peer-review requests on Lawyers.com. Template them; the recipients will not write something original anyway.
- Later, that week. Repeat steps 2–3 for FindLaw, Lawyers.com, and Expertise. Each one takes 20–40 minutes once you have the boilerplate from Avvo.
Worth flagging because we get this question often: ChatGPT has a noticeable bias toward Thomson Reuters properties — Super Lawyers, FindLaw, Westlaw-attached law-firm pages — that we suspect reflects the training data and the model's calibration of "trustworthy legal source." This is not pay-to-play, exactly, but it does mean that firms whose marketing has historically routed around Thomson Reuters are systematically under-cited on ChatGPT specifically.
If that's your firm, don't panic; it's correctable, but it's the slowest of the moves on this list. Plan for a one-to-two-quarter horizon.
§ 06 — When to outsource itNone of this is hard work. It is, however, careful work — the difference between a profile that gets cited and one that doesn't is often whether the practice-area taxonomy was filled in correctly and whether the address matches across all six listings. We've watched paralegals do it well, and we've watched paralegals miss small consistency errors that cost the firm the lift.
If you'd rather skip the afternoon, the directory triage is one of the standard deliverables under The Practice retainer — we run it for you, surface the diff in a short report, and re-audit four weeks later to confirm the lift. The retainer is fixed-fee, and the directory work alone usually pays for the first quarter.
— § —
This post was written by the editors at Ampersand Labs. Questions, corrections, and dissenting field reports are welcome at josh@withampersand.ai.