When ChatGPT, Claude, Perplexity, or Google AI Mode names a law firm in an answer, the citation rarely comes from the firm’s own website. It comes from a review aggregator: Avvo, Martindale-Hubbell, Justia, Super Lawyers, Best Lawyers, FindLaw, or a state bar directory. A 5WPR and Haute Lawyer report published in April 2026 put a number on it: seven directories account for nearly every AI citation across the major legal query categories. If your firm is not present, consistent, and structurally clean inside those seven, you are not in the answer.
Why review aggregators dominate AI citation for legal queries
LLMs are not crawlers in the Google sense. They are pattern matchers that have been trained or grounded on a fixed pool of sources their operators decided to trust for a given vertical. For legal queries, that pool is small and remarkably stable across engines. Super Lawyers, Avvo, Martindale-Hubbell, and FindLaw show up in the largest share of ChatGPT responses to “best [practice area] lawyer in [city]” prompts. Perplexity adds Justia and state bar sites. Google AI Mode pulls from the same set plus the firm’s own Google Business Profile.
The reason these seven win is structural, not commercial. Each one provides the same data fields for every attorney: name, firm, jurisdiction, practice areas, years licensed, peer ratings, client reviews, disciplinary history, and contact information. That uniformity is what an LLM needs to do entity resolution. When the model is asked “who handles construction defect cases in Mount Pleasant SC,” it is not generating an answer from raw web text. It is matching the query to entities it can identify with high confidence, and entities defined identically across multiple authoritative directories have the highest confidence.
A firm with a beautiful website but no Avvo profile, no Martindale rating, and inconsistent listings across Justia and FindLaw is, from the model’s perspective, an entity it cannot resolve. The firm next door with a 4.8 Avvo rating, an AV Preeminent Martindale badge, and a verified Justia profile gets named instead.
What signal does each aggregator actually pass to an LLM?
Different aggregators carry different weight depending on the query. The pattern that has emerged across 2025 and 2026 citation studies looks like this.
Avvo passes attorney-level signals: the 1.0 to 10.0 Avvo rating, peer endorsements, client reviews, and disciplinary record. It is the strongest single source for queries that name an individual attorney or ask “who is the best lawyer for X.” Avvo profiles are also the easiest for an LLM to parse because the schema is consistent across every attorney listing on the site.
Martindale-Hubbell passes peer-validation signals: AV Preeminent ratings, BV Distinguished ratings, and the Lawyer Ratings & Client Reviews data set. The AV Preeminent rating in particular acts as a trust shortcut. When an LLM sees a firm cited as AV Preeminent across multiple sources, it weights that firm higher in recommendation answers.
Justia passes structural signals: practice area taxonomies, jurisdiction tags, and free-text attorney bios that are unusually clean for an LLM to extract. Justia also surfaces in Perplexity citations at roughly twice the rate it shows up in ChatGPT, because Perplexity’s grounding model gives more weight to open-web directories than commercial platforms.
FindLaw passes consumer-trust signals: the Thomson Reuters editorial backing, the Super Lawyers integration, and the geographic directory structure. Google AI Mode favors FindLaw because the platform’s local-attorney pages map cleanly to Google’s own local entity graph.
Super Lawyers and Best Lawyers pass recognition signals: editorial selection by peers and methodology that LLMs can describe in citation text. Both engines are explicitly mentioned by name in Perplexity and ChatGPT answers when they back a recommendation, which is a citation-quality multiplier most firms underestimate.
State bar directories pass authority-of-record signals: license status, jurisdiction, bar number, and good-standing confirmation. These are the lowest-volume citations but the highest-trust. An LLM that cannot verify a firm against the relevant state bar will often refuse to recommend it at all.
The structured-data signal: what aggregator pages send to AI engines
Every one of the seven directories ships its attorney pages with extensive schema markup. Martindale-Hubbell pages carry LegalService, Attorney, and Person schema with worksFor links between them. Avvo wraps every profile in Attorney schema with aggregateRating, review, and areaServed fields populated. Justia uses LawProfessional schema with practiceArea taxonomy. FindLaw layers LocalBusiness, Attorney, and Review schema on every directory page.
That structured data is what an LLM extracts. The natural-language bio matters too, but it matters less than most firms think. The schema fields are deterministic. They produce identical output every time the page is processed. The bio is probabilistic. Two crawls a month apart can yield different summaries from the same paragraph.
This is why a sparsely written but schema-complete profile usually outperforms a richly written but schema-light one. The aggregator with the most populated structured-data fields for your firm is the one the LLM will quote when asked about you. Stackmatix’s 2026 schema study found sites with structured data on attorney pages get cited 3.2 times more often than sites without it. That multiplier applies to your aggregator profiles, not just your own site.
The practical implication is uncomfortable for firms that focus their content investment on their own website. Your homepage hero, your beautifully written About page, your custom-illustrated practice area pages: the model probably is not reading any of them when it answers a recommendation query. It is reading the seven directories. Your job is to make sure those seven directories are saying what you want them to say, with the structural completeness that lets the model quote them with confidence.
How LLMs do entity matching across aggregators
Entity resolution is the layer that decides whether your firm is one entity or many. The same physical firm can appear differently across Avvo, Martindale, Justia, and Google Business Profile: minor variations in name, address format, phone number, attorney roster, and practice area taxonomy. Every variation costs you. The LLM either picks the version it has the most confidence in and ignores the others, or, worse, treats them as separate entities and dilutes the citation share across all of them.
The fix is mechanical: pick one canonical name, address, phone format, and practice area list and enforce it everywhere. Use the exact name on your state bar registration. Use a USPS-validated address. Use a single phone number per office. Use the same practice area labels across every aggregator profile, ideally matching the labels each aggregator already uses (do not invent custom labels like “white-glove litigation” when “Civil Litigation” is the standard).
Cross-link where you can. Every aggregator allows a website URL field. Use it. Every profile should link to the same canonical firm URL, ideally the homepage. That single cross-reference signal compounds across all seven directories and gives the LLM a confident graph of “this firm is one entity that exists in all of these places.”
Add Wikidata. A firm with a Wikidata entry tied to its website, address, attorneys, and practice areas gives the LLM a citation-grade anchor that exists outside any single commercial platform. The April 2026 study showed firms cited in more than 70% of ChatGPT responses had 15 or more entity citations across Wikidata and directories. Wikidata is the single highest-leverage entry on that list because it is the one most firms have not yet created.
How to feed your own site’s structured data into the same signal
Aggregators do most of the citation work, but a complete LegalService schema on your own site closes the loop. The model uses your structured data to verify what it pulled from the directories. A mismatch between your site’s schema and the aggregators’ schema reduces citation confidence on both sides.
The minimum useful stack on your own site is LegalService for the firm, Attorney or Person schema for each attorney with worksFor pointing to the firm, FAQPage on every practice area page, and Review schema aggregating the same review counts and average ratings you show on Avvo and Google. The numbers must match. If your site claims 250 reviews at 4.9 stars and your Google Business Profile shows 47 reviews at 4.7 stars, the LLM picks Google and downgrades your site.
FAQPage schema on practice area pages is the highest-return addition for AI citation. February 2026 testing across legal queries found FAQPage schema appearing in 67 percent of AI answers to relevant questions. The pattern works because it gives the model a clean question-and-answer extraction target that matches the shape of the queries users actually type into ChatGPT and Perplexity.
How review counts and ratings translate into LLM behavior
LLMs do not have a hidden rating threshold, but in practice the behavior maps to roughly these brackets across the citation studies published in late 2025 and early 2026.
Under 10 reviews per directory: rarely cited unless the firm is the only result in a narrow geographic or practice area niche. The model treats sparse review counts as low-confidence signal.
10 to 50 reviews per directory, 4.5+ average: cited intermittently. The model will name the firm in answers when the query is specific enough that the firm matches cleanly, but loses to more-reviewed competitors on broader queries.
50 to 200 reviews per directory, 4.7+ average: consistently cited. This is the threshold where the firm starts to appear in three-to-five-firm recommendation lists across multiple AI engines for its primary practice areas.
200+ reviews per directory, 4.8+ average: dominant in citation share. The firm shows up in the top one or two recommendations across most engines, and often gets named in answer text rather than just in source lists.
The numbers are per directory, not per firm. A firm with 400 Google reviews and 8 Avvo reviews is a firm with 8 Avvo reviews from the LLM’s perspective when it pulls the Avvo entity. Spread the review velocity across the seven aggregators that matter, not just Google.
Why the citation share is consolidating, not fragmenting
A common assumption is that as more AI engines launch, citation share will spread out. The opposite has happened. The seven legal directories that own the citation layer in 2026 own a larger share than they did in 2025, not a smaller one, because every new engine that enters the market grounds itself against the same authoritative corpus. Perplexity, ChatGPT, Claude, Google AI Mode, and the various ChatGPT clones built into vertical tools all converge on the same source set for legal queries.
This is good news for firms with strong aggregator presence and terrible news for firms that have ignored the directories. The window to catch up is closing because the engines are starting to weight long-standing aggregator presence as a trust signal in itself. A profile that has existed and accumulated reviews for five years carries more weight than a profile created last month with the same review count.
The fastest path forward for a firm starting late is to focus the first 90 days on Avvo, Martindale, Justia, and Google Business Profile in that order, build to 50+ verified reviews on each, and only then expand to the second-tier directories. Spreading thin across all seven simultaneously produces 10 reviews each and 10 reviews each gets you nowhere.
FAQ
Do LLMs cite my law firm’s own website or just aggregators?
Both, but aggregators dominate for recommendation queries. Your own site gets cited more for explanatory queries (what is the difference between Chapter 7 and Chapter 13) and FAQ-style queries where your content has the cleanest answer. For “best lawyer in” recommendation queries, the citation almost always comes from an aggregator first, with your site cited as a corroborating source.
Which aggregator should I focus on first?
Avvo if you are an individual attorney optimizing personal citation. Martindale-Hubbell if you are a firm optimizing firm-level recognition. Justia if you are competing in jurisdictions where Perplexity citations matter. Google Business Profile in every case, because it is the only one that also feeds Google AI Mode and traditional local search at the same time.
How long does it take for a new aggregator profile to start passing citation signal?
Roughly 60 to 120 days after the profile is verified and has accumulated 10+ reviews. The lag is partly indexing and partly trust-building. Engines that ground on a corpus do not see new entities the day they are created. They see them at the next training or grounding refresh, and the bar for inclusion is higher for entities with thin history.
Does paying for premium directory tiers improve AI citation?
Indirectly. The premium tiers (Avvo Pro, Martindale ProVantage, Justia Premium) give you more complete profile fields, photos, video, content blocks, and review-collection tools. The structured-data completeness that produces is what improves citation, not the badge or the paid placement itself. The badge has no direct citation value to an LLM.
What is the single fastest fix if I want to be cited more in AI answers next quarter?
Audit name, address, phone, and practice areas across the seven directories and Google Business Profile, then enforce one canonical version everywhere. Most firms have at least three variations live across their footprint. Consolidating to one within 30 days is the single highest-return fix because it lets the LLM resolve your firm to one confident entity instead of three uncertain ones.
What to do next
If you do not know what your firm looks like across the seven aggregators right now, the answer is “probably worse than you think.” Most firms have a strong Google Business Profile, an under-claimed Avvo profile, a Martindale listing nobody has touched since 2019, and a Justia stub. That distribution loses to a competitor with even coverage across all four.
Run the audit, fix the inconsistencies, set a review-velocity target for each directory, and check back on AI citation share in 90 days. If you want a side-by-side of how your firm currently looks to ChatGPT, Perplexity, and Google AI Mode versus your top three competitors, book a free AI visibility audit or estimate the revenue impact with the AI SEO ROI calculator.
Tagged