What We Learned Asking ChatGPT to Recommend NYC Divorce Lawyers
AI recommendation behavior is measurable, but it does not look like a Google results page. The firms assistants name are the firms whose public pages give them enough specific language to reuse.
The question is no longer hypothetical
A prospective divorce client does not need to start with Google anymore. They can open ChatGPT, Claude, Gemini, or Perplexity and ask a version of the question they are afraid to ask out loud: who should I talk to if my marriage is ending, there are children involved, and I need someone serious in New York?
Viclaro Atlas exists to measure what happens next. We send structured buyer-style prompts to major AI assistants, record which firms they name, and publish the market map with methodology notes and confidence intervals. The point is not to crown the best law firm. The point is to show which firms AI assistants surface when buyers ask for help.
The first lesson: AI does not produce one stable market map
ChatGPT, Claude, Gemini, and Perplexity do not recommend the same firms at the same rate. One assistant may lean toward well-known institutional firms. Another may surface boutiques. Another may avoid naming specific firms until the user narrows the situation.
That variance is not a bug in the research. It is the market. Buyers use different assistants, and the assistants draw from different retrieval systems, training data, search integrations, and safety patterns. A serious audit has to show provider-level behavior instead of hiding it behind a single tidy score.
The second lesson: broad reputation is not enough
Some firms with strong public reputations still disappear on highly situational prompts. The issue is not necessarily quality. The issue is that their websites often speak in credentials and practice-area labels while buyers speak in crises, constraints, and decisions.
An assistant answering a prompt about hidden assets, custody pressure, physician income, or a same-week filing needs language it can confidently map to that situation. If the firm page only says "complex matrimonial matters," the assistant may choose a competitor whose page says the specific thing more clearly.
The third lesson: the useful output is the gap, not the rank
A rank is useful because it gets attention. The commercial value is underneath it: which prompts did the firm lose, who got named instead, and what language did that competitor publish that made the recommendation easier?
That is why Atlas and audits belong together. Atlas shows the public market map. The audit turns the map into a fix list for one firm: the answer block to add, the FAQ schema to publish, the page that needs sharper language, and the prompt set to re-run afterward.
Key takeaways
- AI recommendations are measurable, but they are provider-specific and prompt-specific.
- A firm can be respected offline and still poorly represented in buyer-situation prompts.
- The best sales asset is not a visibility score. It is a lost prompt with a named competitor and a fix.
Next step
Atlas shows the public map. A Viclaro audit turns that map into the prompts your firm is losing and the page edits most likely to change the next scan.
See the live NYC divorce and family law Atlas.