A general counsel asks ChatGPT which firms handle cross-border data privacy disputes in Texas. A startup founder asks Perplexity which lawyers do Series A work in Austin. Those answers are assembled from pages the model can read, parse and quote. If your site does not answer the question in plain language, the model picks a firm that does.
This playbook covers what to fix on your site, in what order, and how to check whether it worked. It applies to firms of any size, though solo and small-firm sites usually see the fastest movement because there is less conflicting content to clean up.
How AI systems decide which firm to name
Three crawler behaviors matter most, and they differ by vendor.
OpenAI. OAI-SearchBot surfaces websites in ChatGPT search. Sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers. GPTBot is a separate crawler that collects content to improve foundation models, and blocking it does not affect ChatGPT search. ChatGPT-User handles user-initiated actions in ChatGPT and custom GPTs; because a user triggered it, robots.txt rules may not apply, and it is not used to decide whether your content appears in search.
Anthropic. ClaudeBot collects content that may be used for training. Claude-User fetches pages when someone asks Claude a question, so blocking it prevents retrieval for those queries. Claude-SearchBot indexes content to improve search results; blocking it prevents indexing for search. Anthropic says its bots honor robots.txt, support Crawl-delay, and publishes its IP ranges.
Perplexity. PerplexityBot surfaces and links sites in Perplexity search results and is not used to crawl content for foundation models. Perplexity-User fetches pages for user questions and generally ignores robots.txt because the user requested it.
Google. AI Overviews are part of Google Search, so normal Googlebot crawling and indexing govern them. Google-Extended controls whether content trains future Gemini models or is used for grounding in Gemini apps and the Vertex AI Gemini API. It does not affect inclusion in Google Search and is not a ranking signal.
One practical implication: if your robots.txt blocks Claude-User or OAI-SearchBot, you are invisible to a slice of queries regardless of how good your pages are. Check the file before rewriting anything. Robots.txt changes can take about 24 hours to propagate.
Start with practice-area pages that answer real questions
Most firm practice-area pages describe the firm. AI systems need pages that answer the client's question.
Take one practice area and list the eight to twelve questions a prospective client actually asks in the first call. Not "What is commercial litigation?" but "Can I sue a vendor in another state over a failed software implementation?" Write a page section for each, with a direct two-to-four sentence answer at the top, then the detail.
Useful patterns:
- Question as heading, answer in the first paragraph. Models extract passages; a heading that matches a real query is easier to lift.
- Name the jurisdiction explicitly. "We handle employment disputes in California state and federal courts" is retrievable. "We serve clients nationwide" is not.
- State the procedural posture. Whether a matter is pre-litigation, in arbitration, or on appeal changes who is a fit.
- Give the timeline and cost structure in general terms. "Most single-plaintiff discrimination cases in this state resolve in twelve to eighteen months" is the kind of sentence that gets quoted. Present ranges as your own experience, not as a survey.
- Avoid client-specific claims. Bar rules on testimonials and results vary by state; if you cannot substantiate a result, do not put it in a passage a model may quote out of context.
If your firm operates in multiple states, build a page per jurisdiction rather than one national page. The same logic that applies to local businesses and AI assistants applies here: the model is answering a question with a location in it.
Jurisdiction clarity is a ranking factor in practice
AI answers often include a qualifier: "for matters in [state], firms such as X and Y." If your site never states the states where your attorneys are admitted, you will not appear in those lists. Put admissions in the attorney bio and repeat the jurisdiction on the practice-area page. Two mentions, two different page types, same fact.
Attorney bios that models can quote
Bios are the most underused page type on most firm sites. A bio that says "passionate advocate" gives a model nothing. A bio that says "represents employers in wage-and-hour class actions in the Ninth Circuit; admitted in California and Nevada" gives it two retrievable facts.
For each attorney, include:
- Bar admissions by state, with year
- Courts admitted to practice before
- Two or three matter types described generically ("trade secret disputes involving departing employees"), not client names
- Published articles or speaking, if you have them
- A clear statement of who they do not represent, if that is accurate and useful
Keep bios in HTML text, not inside images or PDFs. Models read text.
Bar-rule compliance on claims
Every state bar has rules on advertising and claims about results. AI answers can quote a sentence without its surrounding caveat, so write defensively:
- Avoid superlatives you cannot document ("best," "top-rated") unless a verifiable third-party rating supports them and you cite it.
- Do not state or imply a guaranteed outcome. This applies to page copy and to any FAQ markup.
- Keep disclaimers adjacent to the claim, not in the footer.
- If your state requires specific advertising language, put it on the page, not only in the site terms.
This is not just a compliance issue. Models are more likely to quote hedged, specific language than promotional language, because it reads as factual.
Directories still matter
AI systems draw on third-party sources, not only your site. Legal directories, bar association listings, and reputable local media are frequently cited in answers. Make sure your directory profiles match your site on the facts that matter: practice areas, jurisdictions, office locations, attorney names and admissions. Inconsistencies between your site and a directory give a model a reason to prefer a competitor whose facts line up.
You do not need to be in every directory. You need the profiles you have to be accurate and current, especially the ones that already rank for your practice area and city.
Check whether any of this is working
You can test your visibility without paying for a rank tracker. Our free AI visibility check takes your brand and category, generates three buyer questions, asks Perplexity with web search, and shows whether your firm is named, which competitors are named, and which sources are cited. It runs three checks per day and does not track results over time, so treat it as a snapshot: run it after each round of page changes and note what moved.
The cited-sources list is the most actionable part. If competitors are being named from directory profiles and bar listings rather than their own sites, that tells you where to spend the next hour.
A sequence that fits a busy firm
- Audit robots.txt. Confirm OAI-SearchBot, ClaudeBot, Claude-User, Claude-SearchBot and PerplexityBot are not blocked. Fix and wait about a day.
- Pick your two highest-value practice areas. Rewrite each with eight to twelve real client questions, jurisdiction named in the first paragraph of every section.
- Rewrite the bios for the attorneys who work in those practice areas. Admissions and courts first.
- Update directory profiles to match the site exactly.
- Run the visibility check, record the named competitors and cited sources, and fix the gaps they reveal.
Start with step one today. A blocked crawler makes every other effort invisible, and it takes about ten minutes to confirm.