How AI Engines Choose Lawyers and What Your Firm Controls

Short answer: ChatGPT, Claude, Perplexity and Google AI Overviews build their answers from sources they can retrieve and trust: firm websites, legal directories, review profiles, news and local pages. They tend to recommend firms that are described consistently across those sources, that answer the question clearly on their own pages, and that other sites mention. You cannot buy a place inside the organic answer, though Google and OpenAI now sell labeled ads shown alongside AI answers. What you can work on is most of the inputs the engines read.

Want a measured baseline instead of a guess? See the FirmCited audit and FirmCited Monitor, covered near the end. If you would rather run the first pass yourself, start with is my law firm cited by ChatGPT?

Two kinds of AI answers

It helps to separate two cases.

Answers from training data. When an assistant answers without searching, it relies on what it learned during training. Those answers can be out of date and you have little direct control over them, beyond being widely and consistently described on the public web over time.

Answers with live retrieval. When search is on, which is now common for "find me a lawyer" style questions, the engine fetches pages, reads them, and writes an answer with links. This is where firms can act, because the engine is reading the web as it is today.

Most of this guide is about the second case.

The inputs engines appear to weigh

No engine publishes its full ranking method, so treat the list below as observed patterns, not rules.

1. Entity clarity

The engine needs to be sure who you are. That means one firm name, one address and one phone, written the same way on your site, your Google Business Profile, your state bar listing and the major directories. Small differences, like "Suite 200" on one and "Ste. 2" on another, create doubt.

2. Pages that answer the question

If a client asks "how much does an uncontested divorce cost in Texas", the engine looks for a page that answers that directly. Pages that put the answer in the first paragraph, then explain, are easier to quote than pages that open with a long firm history.

3. Third-party mentions

Directories, review sites, bar association pages, local news and "top lawyers" lists are frequently cited sources. When an engine names a firm, it often links one of these rather than the firm's own site. Knowing which third-party sources win in your market tells you where to be present.

4. Reviews and recency

Review count, rating and how recent the reviews are show up in many cited sources. A profile with recent reviews looks more current than one with a burst of reviews from years ago.

5. Crawl access and rendering

If your site blocks AI crawlers in robots.txt, or renders key content only after heavy scripts load, some engines may not read it. Check that the bots you want are allowed and that your main text is in the page HTML.

6. Structured data

Schema markup such as LegalService, LocalBusiness and FAQPage helps machines parse your pages. It does not promise anything, but it removes ambiguity when it matches what is visible on the page.

Why results differ by engine

Each engine uses its own index and sources. Google AI Overviews lean on Google's own index. Perplexity shows numbered sources and often pulls from a wide set of sites. ChatGPT and Claude run their own web search. That is why a firm can do well on one and be absent on another, and why testing only one engine misleads.

What a practical plan looks like

1. Baseline. Test 20 to 30 client questions on ChatGPT, Claude, Perplexity and Google AI Overviews, twice each. Record named, cited or missing, and who appeared instead.
2. Fix facts first. Correct any wrong address, phone, practice area or attorney name the engines repeat.
3. Align your profiles. Make your name, address and phone identical across your site and the directories that were cited in your baseline.
4. Publish answer-first pages. For each question group where you were missing, write one page that answers it in the first two sentences and adds an FAQ.
5. Add matching schema. Mark up the firm, attorneys and FAQs, and validate it.
6. Check crawl access. Review robots.txt and confirm main content loads without scripts.
7. Re-measure monthly. The trend is what tells you whether the work is landing. FirmCited Monitor runs this scan every month.

A worked example (illustrative)

Picture a two-attorney family law firm in a mid-size US city. It tests 25 questions on four engines, twice each: 200 answers. It is named or cited in 14, a 7% rate. Brand questions mostly work. Recommendation questions mostly do not, and the winners are linked through two directories and one local "top lawyers" page. The firm's own fee page is cited once, on Perplexity only. The fix list writes itself: align the two directory profiles, ask recent clients for reviews on the cited profile, rewrite the fee page answer-first, and re-measure next month. This is an illustration, not a client result.

What to be skeptical of

• Promised placements or rankings in AI answers.
• A single "AI score" with no list of the questions tested.
• Results from one engine presented as the whole picture.
• Reports that do not show which sources were cited instead of you.

Where FirmCited fits

FirmCited is software for US law firms that want to measure this instead of guessing. The FirmCited audit is USD 490 one time: 25 prompts, up to 4 engines, a 10-page fix list, under 48 hours. It records each answer verbatim, counts every firm the engines cite, and shows competitor share of voice next to that fix list. The engines are ChatGPT, Claude, Perplexity and Google AI Overviews. FirmCited Monitor is USD 299 per month, cancel anytime, and repeats the measurement so you can see the trend.

Related reading

• Is my law firm cited by ChatGPT?
• The recurring compliance deadlines small law firms miss
• One-time vs ongoing OFAC screening

Related tools

• FirmCited audit. The FirmCited audit is USD 490 one time: 25 prompts, up to 4 engines, a 10-page fix list, under 48 hours.
• FirmCited Monitor. FirmCited Monitor is USD 299 per month, cancel anytime.

FAQ

How does ChatGPT decide which law firm to recommend?

When search is on, it retrieves pages, reads them and writes an answer. Firms that are described consistently across trusted sources and that answer the question clearly on their own pages appear more often. The exact method is not published. The same pattern shows up on Claude, Perplexity and Google AI Overviews, and the four engines often disagree.

Do legal directories matter for AI search?

Often, yes. Directories and review profiles are frequently linked as sources in AI answers about lawyers. Which ones matter varies by market, so check which domains appear in your own tests.

Does schema markup help with AI answers?

It can help engines understand your pages, but it is not a promise. It should match what is visible on the page.

Should I block AI crawlers?

That is a business decision. If you want to be cited in AI answers, blocking the crawlers those engines use may reduce your chances.

Is AI search replacing Google for law firms?

Not today for most firms. Google organic and map results still drive many calls. AI answers are an added layer worth measuring.

How long until changes show up?

It varies by engine and by how often your pages and the cited sources are re-crawled. Monthly measurement is the practical way to find out.

What is the difference between GEO and AEO?

Both terms describe making content easier for AI and answer engines to use. Generative engine optimization usually refers to AI-written answers. Answer engine optimization is an older term that also covers featured snippets and voice answers.

What does the FirmCited audit include?

The FirmCited audit is USD 490 one time: 25 prompts, up to 4 engines, a 10-page fix list, under 48 hours. The engines are ChatGPT, Claude, Perplexity and Google AI Overviews. FirmCited Monitor is USD 299 per month, cancel anytime.

Disclaimer

General information and software only - not legal advice, no attorney-client relationship. Consult a licensed attorney.

Reports are generated with AI-assisted software and may contain errors. Review before relying on them.