§ How we ran the test
Method, so you can replicate it: we wrote 10 prompts a founder would genuinely type — not marketing keywords, real hiring questions:
- Best startup lawyer in NYC
- Can you recommend a startup attorney in New York for a seed-stage company?
- Which law firm should I use for my Series A in NYC?
- Affordable startup lawyer in New York for incorporation
- Best law firm for SAFE notes in NYC
- Who are the top venture capital lawyers in New York?
- I'm a founder raising $2M — which NYC lawyer should I talk to?
- Best boutique startup law firm in NYC (not BigLaw)
- Startup lawyer NYC with flat-fee pricing
- Which attorneys do YC founders in NYC use?
We ran these through Claude — one of the four engines we track, and a useful first leg because its answers come primarily from model knowledge (the "reputation memory" built from years of press, deal coverage, and citations) rather than live retrieval. We recorded every firm named. ChatGPT, Gemini, and Perplexity legs are running next — each engine retrieves differently, and we'll publish each leg the same way.
§ What the AI actually answered
| Firm | Named in (of 10 prompts) |
|---|---|
| Gunderson Dettmer | ~9 |
| Cooley | ~9 |
| Fenwick & West | ~7 |
| Wilson Sonsini | ~6 |
| Goodwin | ~5 |
| Lowenstein Sandler | ~5 |
| Orrick | ~3 |
| Boutiques ranking page 1 of Google for "startup lawyer NYC" | 0 |
Three more findings that matter as much as the table:
The Google–AI overlap was roughly 20%. The firms dominating Google's results for NYC startup-law searches — mostly boutiques with strong SEO — and the firms the AI recommends barely intersect. This mirrors what large-scale studies show across industries: Ahrefs found only 12% of AI-cited URLs rank in Google's top 10 for the same query.
Three prompts returned no firm at all. "Affordable startup lawyer," "flat-fee pricing," and "boutique, not BigLaw" produced generic advice and marketplace suggestions — but no named firm. In the most expensive legal market in America, the price-sensitive founder questions have no incumbent answer.
The engine has a personality. This leg of the test shows what we call the Reputation Snob pattern: model-memory answers favor whoever history already talks about — decades of Chambers rankings, deal press, and legal media compound into recommendation dominance for BigLaw. Retrieval-based engines behave differently (they fetch live web results, often through Bing's index, where 87% of ChatGPT search citations have matched Bing's top organic results) — which is exactly why the next legs of this audit may tell a different story, and why we're publishing each one.
§ Why do boutiques that win Google lose the AI answer?
A boutique can win Google with excellent SEO while remaining, from the AI's perspective, a firm with no evidence trail. The practical translation: your #1 Google ranking doesn't transfer to AI answers — and for boutiques, the reputation game (BigLaw's moat) isn't the one to play anyway.
§ The real opportunity is the empty slots
Here's what should interest every boutique managing partner in the data: nobody owns "affordable," "flat-fee," or "boutique" in the AI's answers. These aren't consolation-prize queries — they're the exact questions asked by the seed-stage founders boutiques most want, at the moment of hiring intent.
Unlike "best startup lawyer NYC" — where displacing Gunderson from a model's memory is a years-long reputation project — an empty slot has no incumbent.
The first firm that makes itself readable, trustable, and citable for those questions — indexed where engines retrieve, answering the exact question in extractable form, corroborated by third-party sources — takes the slot. Recent-content bias works in the challenger's favor too: fresh, well-structured pages are disproportionately cited.
That's not a theory we're asking anyone to take on faith. We're currently running a public test of it: five NYC boutique firms, 60 days, weekly published results — win or fail. The before-measurements use exactly the method above.
§ How to check your own firm (free, 20 minutes)
- Write the 10 questions your real clients would ask an AI — practice area + city + the price-sensitive variants.
- Ask them in ChatGPT, Gemini, Perplexity, and Claude. Record every firm named. Screenshot everything — that's your baseline.
- Check
site:yourfirm.comon Bing (the index behind much of ChatGPT's retrieval). Sparse results = largely invisible to the most-used AI assistant. - Note which questions returned no firm. Those are your open slots.
§ Frequently asked questions
Does AI recommend law firms by name?
Why do boutique firms that rank on Google lose the AI answer?
How can a law firm check its own AI visibility for free?
What is the biggest AI-visibility opportunity for boutique firms?
Find your firm's empty slots
FirePencil is the autonomous AEO agent built for exactly this gap: it baselines what AI engines say about your firm across the questions that matter, then executes the fixes — structure, content, entity and authority signals — on weekly sprints, every change owner-approved, bar-advertising rules respected. We're in private beta with law firms now.
FirePencil.AI is an autonomous AEO agent built for law firms. This article reports original research (the Claude leg of a four-engine NYC startup-law audit, July 2026); firm-name counts are approximate and marked with "~". It is general information, not legal advice, and is not affiliated with or endorsed by any firm named. Third-party names (ChatGPT, Gemini, Perplexity, Claude, Google, Bing) are trademarks of their respective owners; use is descriptive.