How to get your law firm mentioned in ChatGPT answers

What has to be true before ChatGPT will name your firm?

ChatGPT names a firm when its record is retrievable, corroborated and extractable. Retrievable means the firm appears in the sources the engine pulls from, which for legal queries are dominated by directories rather than firm websites. Corroborated means more than one independent source says the same thing about the same named entity. Extractable means a sentence exists that answers the user’s question and survives being quoted out of context. A firm can be excellent and fail all three.

The reason ordinary search work does not cover this is overlap. Martindale-Avvo reported in April 2026 that ChatGPT matches Google’s top ten results under 25% of the time for legal queries, while Perplexity and Claude sit near 75% and Gemini near 50%. Custom Legal, citing Semrush, puts the ChatGPT figure far lower still, at 6.82%. Neither publishes its query set, so treat the range rather than either endpoint as the finding. The reading is the same at both ends: a firm can hold page-one rankings in Google and still be absent from ChatGPT.

Step one: does your firm exist as one entity across the web?

Entity consistency is the first stage because every later stage depends on it. An answer engine has to decide that the Super Lawyers profile, the Google Business Profile, the state bar listing, the LinkedIn page and the firm website all describe one organisation. When the firm name is punctuated three ways, the suite number appears on two of five listings, and the practice focus reads as estate planning in one place and elder law in another, the corroboration the engine needs never accumulates. Nothing is penalised. The signals simply fail to merge.

Entity work is unglamorous and it is the cheapest thing on this list. It is name, address, phone and practice description made identical everywhere, one canonical firm description reused verbatim, a named-attorney page for every lawyer who appears on any directory, and markup that matches the text on the page. Google’s 2026 guidance on AI search citations, summarised by Just Legal Marketing, is that AI Overviews and AI Mode citations are earned through standard search fundamentals rather than any AI-specific file or schema.

• •One firm name string, one address format, one phone number, used on every profile without exception.
• •One practice-focus sentence, reused word for word, so the description itself becomes a matchable signal.
• •A page per attorney under a stable URL, with the same credentials, bar admissions and practice list that the directories carry.
• •Organisation, LegalService and Person markup that repeats what the visible page says rather than adding claims the page does not make.

Step two: which directory profiles actually carry your name into the answer?

Directories are the second stage because they are where the engines read. The 5WPR and Haute Lawyer Legal AI Visibility Index, published April 2026, found that seven directories own the citation layer for legal queries across eight practice areas, and that zero law-focused editorial sources appeared in top results. The same index reported that even Am Law 100 firms rank below directories for their own practice areas. A firm does not outrank the directory layer. It gets into it.

The work is profile completeness rather than profile quantity. A claimed profile with a blank practice breakdown, no attorney photographs and a three-line biography is a thin corroborating source. A completed one, carrying the entity strings from step one, is the record an engine reads back. Firm AEO treats the stage as done when every attorney appears on every eligible directory with matching text, and the paid listings are marked internally so the firm knows which citations it is renting.

The directory roster differs by engine, and Firm AEO has published that breakdown separately. For sequencing, the point is narrower: directory work is where a firm with no reviews and no content still has something to fix, and it pays back fastest because the profiles are already indexed.

Step three: how many reviews, and how recent?

Reviews are the third stage because they are the corroboration signal the engines weight most heavily for consumer services and the one that takes longest to build. SOCi’s 2026 Local Visibility Index, covering 2,751 brands and roughly 350,000 locations, found ChatGPT recommends only 1.2% of locations and that the locations it does recommend average 4.3 stars. That average is a floor, not a target. A firm sitting at 4.1 stars with eleven reviews, the newest of them from 2024, is below the line the engine appears to draw before it will name anyone at all.

Volume, recency and response rate all move together, and only volume can be rushed, which is exactly the one a law firm must not rush. The Federal Trade Commission’s rule banning fake and incentivised reviews took effect on October 21, 2024, with penalties reaching $51,744 per violation, and ABA Model Rule 7.1 independently forbids a firm from publishing anything false or misleading about its services. Any programme that pays for reviews, filters negative ones before they post, or has staff write them is an ethics problem before it is a marketing problem, and the firm’s own ethics counsel is the one who signs off on the request language.

What is left is slower and it works. A review request sent on every closed matter, at the same point each time, through a process nobody has to remember. A reply to every review, positive and negative, written by a person, disclosing nothing about the representation. A typical consumer firm starting from a few dozen reviews should expect six to twelve months before the profile looks materially different, which is illustrative rather than promised.

Step four: what content does ChatGPT actually pull a sentence from?

Content is the fourth stage because a page only earns retrieval once the entity behind it is resolvable. What gets pulled is specific, answer-shaped and narrow. Lexscale, writing about ChatGPT visibility for personal injury firms, puts the minimum footprint at five to ten pages per practice area, which is a coverage argument rather than a word-count argument: one page per question a client actually asks, not one long page covering the practice. Citorian’s five-engine study found engines named a specific firm 79% of the time for specific-injury prompts against 18% for situational ones, which is the same lesson from the query side.

Shape matters as much as coverage. The Princeton generative engine optimisation study by Aggarwal and colleagues, presented at KDD 2024, measured up to roughly 40% higher visibility for content carrying statistics, citations and quotations, with statistics alone accounting for a 22% to 41% lift. A jurisdiction-specific number with a source beats a paragraph of reassurance. At this stage the decision is what to publish, not how to phrase the lead sentence.

• •One page per real question, titled in the phrasing a client uses, not in the phrasing a practice group uses.
• •Sub-practice pages rather than one practice page, because engines decompose a prompt into narrower sub-questions before answering it.
• •A dated, sourced number in every page that can carry one, because unsourced reassurance is the part an engine summarises past.
• •An attributed author who exists elsewhere as an entity, so the page inherits the attorney’s corroboration rather than standing alone.

Step five: where does third-party corroboration come from?

Corroboration is anything about the firm that the firm did not write. Bar admission records, court records, local news coverage, podcast and panel appearances, law school and association pages, and the editorial write-ups attached to legitimate awards all qualify. Consultwebs states on its AI SEO page that 85% to 90% of AI citations come from earned media. That figure carries no published methodology or underlying study, so Firm AEO does not plan against it, and the directional claim it makes is still worth taking seriously: the sources an engine trusts most are the ones the firm has the least control over.

The one thing corroboration cannot be is manufactured. Placed content presented as independent coverage, reciprocal mention schemes and self-written directory editorial all read as the firm talking to itself once an engine deduplicates by entity. For a firm with no earned media, the accessible starting points are the ones that are simply records: a complete state bar profile, bar association programme pages, and court-record presence where matters produce published decisions.

Does firm size change the order of the work?

It changes the pace, not the order. The 5WPR index describes a visibility paradox in which large, well-known firms under-invest in the citation layer and rank below directories for their own practice areas, which leaves the layer contestable by firms that would never outrank them in Google. Combined with the low ChatGPT and Google overlap, that is the structural reason a four-attorney firm can appear in an answer where a two-hundred-attorney firm does not.

What size buys is throughput at step three and step five. A large firm closes more matters, so it sends more review requests, and it out-publishes a small firm. The small firm’s compensating advantage is narrowness: it can cover one sub-practice completely, in one metropolitan area, faster than a full-service firm covers twelve. The sequencing answer is that a smaller firm compresses steps one and two, picks one practice area for step four, and accepts that step three runs on its own clock regardless of headcount.

How long does each stage take to show up?

Each stage has a different lag between doing the work and seeing the engine change its answer, and the lags do not run in parallel. Entity and directory work can be visible inside a re-crawl cycle because the profiles are already indexed and the engine only has to read the corrected text. Review and earned-media work is bounded by how fast a firm closes matters and how often third parties choose to write about it. Lexscale, in the same personal injury guidance, describes a three to six month window before ChatGPT visibility work shows results, which is consistent with the pattern below.

What does the first 90 days look like in order?

The first ninety days should finish steps one and two, start step three, and open step four on a single practice area. Nothing here requires a new platform or a new website. The sequence below is what Firm AEO runs at the start of an engagement, and a firm that wants to do it without an agency can run the same list.

• •Days 1 to 10: run the firm’s own client questions through ChatGPT, Perplexity and Google AI Mode while logged out, three times each, and record which firms are named and which sources are cited. That is the baseline everything else is measured against.
• •Days 1 to 20: fix name, address, phone and practice description everywhere they appear, and write the one canonical firm description that every later profile reuses.
• •Days 15 to 45: claim and complete every eligible directory profile, attorney by attorney, using the strings from the previous step.
• •Days 20 to 30: stand up the review request process at a fixed point in the matter lifecycle, with request language the firm’s ethics counsel has approved.
• •Days 30 to 75: publish five to ten answer-shaped pages covering one practice area’s real questions, each carrying a sourced number and a named author.
• •Days 75 to 90: re-run the day-one prompt set under the same logged-out conditions and compare appearance rate, not rankings.

What this work cannot promise

No agency controls what an answer engine says, and any firm that is told otherwise should ask to see the mechanism. ChatGPT’s answer to the same prompt varies between runs, between accounts and, since location sharing rolled out on March 26, 2026, between devices. Directory citations can be bought in the sense that some profiles are paid placements, but the recommendation itself is not for sale, and Best Lawyers’ ChatGPT app, launched April 22, 2026, routes its answers through peer-review rankings rather than advertising spend.

What the work does is move the odds, by making the firm retrievable, corroborated and quotable where it was not. The measurable outcome is appearance rate across a fixed prompt set, run the same way each month. Firm AEO reports that number whether it moved or not.