Answer Engine Optimization (AEO) for law firms is the practice of making a firm and its lawyers retrievable, verifiable, and recommendable inside AI-generated answers on ChatGPT, Gemini, Claude, Perplexity, Copilot, and Google’s AI Overviews. It works by strengthening the evidence AI engines use to justify a recommendation: entity clarity, independent corroboration, and technical retrievability.
Legal buys the most expensive clicks on the internet. The average legal click runs around $9, and high-intent terms like “car accident lawyer” clear $100 in competitive markets. ChatGPT hands out the same recommendation for free, and most firms have no idea whether they’re in it. Meanwhile, the clicks themselves are shrinking: Pew Research Center’s July 2025 analysis of real browsing behavior found that when Google shows an AI summary, users click a traditional result on just 8% of visits, versus 15% without one. Traffic from generative AI sources to consumer sites grew roughly 1,200% between July 2024 and February 2025. And the profession itself has already crossed over: 79% of legal professionals now use AI in some capacity, up from 19% a year earlier.
The prospective client got there first. This guide covers what they see, how the engines decide, and what your firm should do about it.
AI Search Moved in Front of Your Intake
A general counsel gets an SEC subpoena involving token sales. She doesn’t call her network first. She opens ChatGPT and asks what kind of outside counsel she needs, then which firms have handled similar investigations, then which partners in New York combine fintech and enforcement experience. Twenty minutes later she has a shortlist and a set of questions to test it.
A personal injury prospect runs the same play at a different altitude: “I was hit by a truck in Phoenix. Do I need a lawyer? Who handles cases like this?”
That session compresses research, referral, and validation into one conversation. The engine defines the problem, decides which evidence counts, assembles a shortlist, and often tells the prospect what to do next. By the time anyone lands on your website, they arrive with a theory of the case and a list you didn’t know you were competing on.
The interesting part is that none of this shows up as “AI traffic” in your analytics. The prospect closes ChatGPT, types your firm’s name into Google, and walks in the front door looking like branded search or direct. The influence happened upstream, in a channel you weren’t watching.
How AI Search Engines Decide Which Law Firms to Recommend
AEO for law firms is a burden-of-proof problem. An answer engine behaves like a skeptical judge: it assembles a case file about your firm from whatever the public record offers, weighs the evidence, and recommends the firms whose qualifications it can verify with confidence. Firms it can’t verify don’t get argued against. They get omitted.
Forget rankings. That framing matters because it kills the most common mistake in legal marketing right now, which is treating AI visibility like a rankings game. There is no fixed number one in ChatGPT. The answer shifts with the wording of the prompt, the user’s location, the model, the retrieval layer, the sources available that day, and the context built up in the conversation. Your firm can appear in one run and vanish in the next. A favorable screenshot is evidence of possibility, nothing more.
Track the pattern, not the snapshot. The right unit of analysis is repeated inclusion across a portfolio of commercially important questions:
- Which firms show up in 7 out of 10 runs?
- Which show up once?
- Which never appear at all?
That distribution is your actual market position.
Cited isn’t the same as recommended. The burden runs deeper than mentions. An engine can cite your firm’s article as a source for legal education while never recommending you as counsel. Those are different verdicts built on different evidence. Being cited proves your content is useful. Being recommended proves the machine could establish, across independent sources, that your firm belongs on the shortlist. This guide is mostly about the second one.
Every AI Engine Answers the Same Question Differently
Optimizing for “AI search” as one channel is like buying one ad and expecting it to run everywhere. Each engine retrieves from a different corpus, trusts different sources, and reasons differently. In Goodie’s citation data, ChatGPT and Perplexity citation overlap runs under 1%. They’re effectively answering from different internets.
| Engine | Retrieval backbone | What moves the needle |
| ChatGPT | Bing’s index plus OpenAI’s own crawlers | Bing indexation, Bing Places, directories, entity consistency |
| Gemini and AI Mode | Google’s index, Maps, Google Business Profile | GBP completeness, review volume and recency, local pack signals |
| AI Overviews | Google’s standard search index | Classic organic strength plus extractable, answer-first content |
| Perplexity | Its own crawl and index | Fact-dense pages, recency, clear citations and sources |
| Copilot | Bing | Same levers as ChatGPT’s search layer |
| Claude | Anthropic’s search and retrieval crawlers | Crawlable, well-structured pages Anthropic’s bots can reach |
For a firm, the practical read: your ten-year neglect of Bing just became expensive, because Bing feeds both ChatGPT and Copilot. And for any consumer practice with local intent, your Google Business Profile now works two jobs: it drives the map pack humans see and the recommendations Gemini generates.
Where AI Answers About Lawyers Come From
Most firms assume the fix for weak AI visibility is publishing more content on their own site. The citation data says otherwise. Full disclosure before the numbers: Goodie is our platform, so this is our dataset. Read it knowing that.
The Numbers
We analyzed 1,856,259 social citations across 10 AI surfaces and 29 social domains between January 19 and August 25, 2026 (Goodie, AI Search x Social, Volume 3). Across the categories we track, the citation mix breaks down like this:
- Earned media: roughly 81%
- Social: about 8%, and climbing
- Owned content: 0.07%
Social content draws 2.31-4.17X more AI citations than owned content, and the gap is widening. YouTube alone accounts for about 46% of social citations.
Sit with that 0.07% for a second. The firm blog you’ve been feeding for a decade is a rounding error in the retrieval corpus. The engines build their case from what everyone else says about you.
Consumer Legal: The Directory Wall
For consumer legal, that means Justia, Avvo, FindLaw, Nolo, Super Lawyers, Expertise.com, state bar profiles, Reddit threads, YouTube explainers, and local news. Ask ChatGPT for a personal injury lawyer in Phoenix, and you’re mostly watching it synthesize directories and reviews, with the firm’s own site playing a supporting role.
Big Law: The Credibility Stack
For sophisticated corporate work, the source map shifts to Chambers, Legal 500, Best Lawyers, Law360, Reuters, Bloomberg Law, court records, regulatory filings, conference bios, and deal announcements. A CFIUS recommendation and a scaffolding-accident recommendation are different evidence problems, and there is no universal list of sites the machines trust. Map the sources per practice, per market. That map is the real content strategy.
Consumer Legal and Big Law Are Playing Two Different Games
Consumer Legal is a Saturation Game
The prospect’s question is local, urgent, and comparative, and the engines lean on directories, reviews, and maps because those sources are structured and independently verifiable. The winning moves are unglamorous:
- Complete and consistent directory profiles
- A Google Business Profile treated like a product
- Steady review velocity with real responses
- Local press
- YouTube content that answers the questions people actually ask at midnight
Name, address, phone, and practice descriptions must match everywhere; every inconsistency is reasonable doubt.
Big Law is an Entity-Resolution Game
Nobody asks Gemini for “a law firm.” They ask which partners combine SEC enforcement work with digital assets experience, or which firms regularly defend Medicare Advantage companies in False Claims Act matters. To answer, the engine has to connect a lawyer to a subpractice to specific matters to independent recognition, and most firm websites make that connection nearly impossible. “Government Enforcement” as a page title tells the machine almost nothing. Bios written as adjectives (“a seasoned litigator with a national practice”) give it nothing to verify.
The failure mode is the same in both markets: the firm has the expertise but the public record can’t prove it. Established firms rarely have an expertise shortage. They have a representation problem.
The Highest-Value AEO Asset Is the Lawyer, Not the Blog
An answer engine recommends people it can resolve. So the core work of legal AEO is turning your site from a collection of pages into an expertise graph: lawyers connected to practices, practices connected to matters, matters connected to industries, regulators, courts, jurisdictions, and outcomes.
Three moves carry most of the weight.
- Rebuild priority bios as authority pages. A bio should let a machine (and a skeptical GC) establish the lawyer’s practice, subpractices, industries, bar admissions, courts, former government roles, representative matters, publications, and independent recognition, all stated concretely. “Advises fintech clients on regulatory matters” is unusable. “Represented a U.S. fintech platform in an SEC investigation concerning digital asset offerings” gives the engine a specific, checkable relationship. Where confidentiality permits, specificity is the whole game.
- Make practice pages answer problems instead of describing departments. The page should connect the practice to the actual questions clients bring: which enforcement regimes, which regulators, which recurring disputes, which lawyers handle each, with links that make the relationships explicit.
- Enforce entity consistency everywhere the lawyer exists: the bio, the bar profile, Chambers, LinkedIn, conference pages, directory listings. Add Person and LegalService structured data with sameAs links tying the profiles together. The engines cross-examine your sources; make sure your witnesses agree.
Technical AEO: Can the Machines Even Read Your Law Firm?
Everything above assumes the engines can retrieve your pages. Plenty of firms fail here without ever deciding to, because a blanket “block the AI bots” call made by IT in 2023, or a CDN default nobody reviewed, quietly removed the firm from the corpus. Cloudflare began blocking AI crawlers by default for new domains in July 2025, which means some firms are invisible by infrastructure, not by choice.
The critical nuance: retrieval and training are separate decisions, controlled by separate crawlers. You can keep your content out of model training while staying fully visible in AI search. Most firms should.
| Crawler | Runs it | Controls | Blocking it means |
| GPTBot | OpenAI | Model training | Opting out of training; search visibility unaffected |
| OAI-SearchBot | OpenAI | ChatGPT search inclusion | Your pages stop surfacing and linking in ChatGPT search |
| ChatGPT-User | OpenAI | Live, user-requested fetches | ChatGPT can’t open your page when a user asks it to |
| ClaudeBot | Anthropic | Training and general crawl | Opting out of Anthropic’s crawl |
| Claude-SearchBot | Anthropic | Claude’s search citations | Reduced visibility in Claude’s web-connected answers |
| Claude-User | Anthropic | User-requested fetches | Claude can’t retrieve your page on demand |
| Google-Extended | Gemini training only | No effect on Search or AI Overviews, which follow normal indexing | |
| Bingbot | Microsoft | Bing’s index | Falling out of the index that feeds ChatGPT search and Copilot |
Then audit the full retrieval chain, not just robots.txt: CDN and WAF behavior, HTTP responses, canonicalization, JavaScript-dependent content, sitemaps, and internal links. And fix the PDF habit. Firms love publishing client alerts, rankings submissions, and matter lists as PDFs; every one of them is evidence the engines struggle to lift. If it proves expertise, it belongs in crawlable HTML.
What Doesn’t Work (Skip the Snake Oil)
The AEO vendor market is young and loud, so let’s clear three things that keep showing up in pitches to law firm CMOs.
llms.txt does nothing today. It makes sense as a framework, but there’s no proof yet. Google’s John Mueller said plainly in 2025 that no AI system currently uses llms.txt, comparing it to the long-dead keywords meta tag. Adding the file is harmless. Paying for it as a deliverable is not.
FAQ schema stopped producing rich results for almost everyone in August 2023, when Google restricted them to authoritative government and health sites, and Google retired the feature entirely in May 2026, even for those remaining sites. FAQs are still valuable on the page because they answer the query fan-out; just don’t expect markup magic.
Schema in general is hygiene, not authority. Use LegalService for the firm rather than the weaker Attorney type; mark up lawyers as Person with sameAs; keep it accurate. Structured data reduces ambiguity about who you are. It cannot manufacture a reputation the public record doesn’t support. Anyone promising to make you “rank #1 in ChatGPT” is selling a position that doesn’t exist.
The Bar Rules Follow You Into AI Search
This is the chapter most AEO guides skip, and for this audience it’s the one that matters. (Yes, I’m about to tell a page full of lawyers that none of this is legal advice. You’d do the same.)
Model Rule 7.1 prohibits false or misleading communications about a lawyer’s services, including statements that create unjustified expectations. Rule 7.2 governs advertising, paying for recommendations, and specialist claims. Rule 7.3 restricts solicitation. States layer their own regimes on top; Florida, Texas, and New York run some of the strictest, with filing and review requirements and tight rules on superlatives and testimonials. And the ABA’s Formal Opinion 512 (July 29, 2024) makes clear that a lawyer’s professional duties travel with generative AI use. If your marketing team uses AI to draft the content that engines will later quote, the accuracy obligation under 7.1 doesn’t relax. It compounds.
Two implications:
- Watch your superlatives. AI-optimized copy loves “best” and “leading.” In many states, those claims need substantiation, and an engine may repeat your language verbatim in front of a regulator’s constituents. Write bios and practice pages you’d defend in a bar inquiry, because functionally you’re publishing them into one giant quotation machine.
- Build a correction protocol for AI misdescriptions. Engines get firms wrong: hallucinated specializations, wrong offices, invented case results. You can’t send a cease-and-desist to a probability distribution. What you can do is saturate the sources the engines retrieve with accurate, consistent, specific information, fix conflicting directory data, and use each platform’s feedback channels for material errors. Accuracy saturation is the only durable remedy, and it’s the same work as the rest of this guide.
How to Measure Legal AEO Without Fooling Yourself
One aggregate “AI visibility score” is convenient on a dashboard and useless for diagnosis. Track outcomes separately, because they have different fixes:
- Mention rate (the firm appears at all)
- Recommendation rate (the firm is suggested as counsel)
- Citation rate (the firm’s content is used as a source)
- Source coverage (your footprint across the domains the engines rely on)
- Accuracy (the engine describes you correctly)
Segment all of it by practice, market, and intent. A firm can dominate legal-education citations and be absent from every hiring recommendation. Two problems, two playbooks.
Build the prompt panel around decisions, not keyword variants. For consumer practices: problem recognition, urgency, eligibility, process, cost, lawyer selection, local comparison, branded validation. For corporate practices, write prompts the way an in-house lawyer asks: “Which firms have substantial CFIUS experience in semiconductor deals?” Run the same panel repeatedly, across engines, over months. Measure the distribution. Ignore the anecdote.
Then account for the zero-click reality. Most AI-influenced demand arrives as branded search and direct traffic, so watch branded query volume alongside your panels, and add one line to intake: “an AI assistant (ChatGPT, Gemini, etc.)” as a how-did-you-hear option. Firms that added it are consistently surprised by the answer.
Full disclosure again: Goodie’s Visibility Monitoring tracks exactly this across 12+ AI surfaces, so I have a stake in you taking measurement seriously. The methodology above works whether or not you ever use our platform.
The Law Firm AEO Playbook
Run it in this order. The sequence matters more than the calendar.
- Pick the decisions worth winning. Not keywords: client situations where appearing on an AI shortlist changes revenue. Ten to fifty of them, per practice.
- Baseline your footprint. Run those prompts across ChatGPT, Gemini, Claude, Perplexity, and AI Overviews. Record who’s recommended, who’s cited, which sources shape each answer, and where you’re described wrong.
- Diagnose the failure mode. Absent from sources? Unretrievable? Unresolvable as an entity? Each has a different fix, and most programs fail by fixing the wrong one.
- Repair retrievability. Crawler policy (allow search bots, decide on training bots deliberately), CDN and WAF behavior, JavaScript rendering, PDFs into HTML.
- Rebuild the lawyer and practice graph. Authority bios, problem-first practice pages, matter specificity, entity consistency, structured data.
- Build the external record. Close directory gaps, drive review velocity, earn coverage in the publications your source map surfaced, and put real answers on YouTube. The 81/8/0.07 split says this is where the citations live.
- Publish evidence, not volume. Original research, enforcement trackers, settlement data, state-by-state guides: assets that make the firm a primary source other sites cite.
- Measure recommendation rate and rerun. Same panel, every month. The goal is a rising probability of inclusion, compounding quarter over quarter.
Boutiques should run this across their whole practice; focus is their advantage, and a specialist firm can make most of its expertise graph legible in a quarter or two. Am Law firms should pick three to five practices where the economics justify the work, win those, and expand. A firmwide “AI transformation” announcement is usually a substitute for doing this properly anywhere.
Know Where You Stand Before You Optimize
Every day a prospect asks ChatGPT, Gemini, or Perplexity who handles cases like theirs, the engine is building a case file, whether your firm is in it or not. Firms that treat this as a rankings problem chase screenshots. Firms that treat it as a burden-of-proof problem build the evidence: retrievable pages, resolvable entities, and a public record that holds up. That’s the work this guide covers, in the order it pays off.
See Where Your Firm Stands in AI Search
You can’t fix what you can’t see. Goodie’s Visibility Monitoring shows which firms get recommended, cited, or skipped across ChatGPT, Gemini, Claude, and Perplexity.
AEO for Law Firms: FAQ
AEO (Answer Engine Optimization) for law firms is the work of making a firm visible, accurately described, and recommended in AI-generated answers on platforms like ChatGPT, Gemini, Claude, Perplexity, and Google’s AI Overviews. It spans technical retrievability, entity clarity, third-party evidence building, and measurement of recommendation rates.
SEO earns a ranked position on a results page; AEO earns inclusion in a synthesized answer. SEO optimizes pages for keywords. AEO optimizes the firm’s full public record, owned, earned, and social, so engines can verify and recommend it. Strong SEO still helps, especially for AI Overviews, but it no longer covers the channel.
Make sure Bing indexes your site and OpenAI’s search crawlers aren’t blocked, then strengthen what ChatGPT retrieves: complete directory profiles, consistent lawyer entities, specific matter evidence, reviews, and earned coverage. Mentions follow verifiable third-party evidence far more than they follow your own blog output.
Decide training and search separately. Blocking GPTBot or Google-Extended opts you out of model training without hurting search visibility. Blocking OAI-SearchBot, Claude-SearchBot, or the user-fetch agents removes you from AI answers. For a firm that wants clients, allowing search retrieval is the default; check that your CDN isn’t blocking it for you.
Decide training and search separately. Blocking GPTBot or Google-Extended opts you out of model training without hurting search visibility. Blocking OAI-SearchBot, Claude-SearchBot, or the user-fetch agents removes you from AI answers. For a firm that wants clients, allowing search retrieval is the default; check that your CDN isn’t blocking it for you.
Modestly, as hygiene. LegalService and Person markup with sameAs links helps engines resolve who you are and reduces misdescription risk. There’s no evidence markup alone earns recommendations, and FAQ rich results have been retired from Google Search since May 2026. Do it for clarity, not as a strategy.
