The Machines Knock on Every Door
We measured the AI accessibility of the advisors of the wealthy — 43 elite firms across private wealth, global law, luxury property, private aviation and elite medicine. What the machines found: an open market that never decided to be open.
Six findings
F1 — Nobody blocks the answer engines. 0 of 35 readable robots.txt block OAI-SearchBot, Claude-SearchBot or PerplexityBot. 0 block GPTBot. The only AI-adjacent block in the sample: CCBot, by two UK property firms. The “elite firms wall off AI” narrative is false at the policy layer.
F2 — Openness is an accident, not a decision. 30 of 35 (86%) name no AI crawler at all. Deliberate policy exists at 4 firms: Morgan Stanley and BNY (near-complete welcome policies), Goldman Sachs (partial), Akin Gump (two bots). Wall Street shows intent; law manages 1 of 7 readable; the rest: zero written policy.
F3 — The wall moved to the door. 7 of 43 (16%) refused or failed a plain browser fetch of robots.txt itself — UBS, White & Case, Mayo Clinic, Johns Hopkins (403); Wachtell, Latham & Watkins, Engel & Völkers (unreachable). AI visibility has two layers: policy (mostly open by default) and door (WAF/bot management, opaque). A firm can be open in policy and invisible in practice.
F4 — Elite medicine leads deliberate AI hospitality. 4 of 8 medicine providers (50%) serve a hand-written llms.txt — Cleveland Clinic, Cedars-Sinai, UCLA Health, HSS — verified by content inspection. Wealth 0/10, law 1/10. The most regulated, litigation-shy sector moved first. Nobody predicted that.
F5 — State of the art lives in pockets. WilmerHale’s llms.txt is practice-area specific (“Focus: International Arbitration & Gary Born”). Compass versions and dates its file. One law firm and one brokerage outperform every bank on the list.
F6 — Three firms answer every door with the same page. christiesrealestate.com, privatefly.com and jsx.com return HTTP 200 with HTML at /llms.txt — SPA catch-alls. A machine cannot tell “missing” from “present” there.
The full sample, firm by firm
“Deliberate” = names ≥1 AI crawler in robots.txt. Raw per-bot verdicts in the CSV.
| Sector | Firm (domain) | robots.txt AI policy | llms.txt |
|---|---|---|---|
| Private wealth | goldmansachs.com | ✅ deliberate — names 2 bots | ❌ absent |
| Private wealth | jpmorgan.com | ➖ open by default | ❌ absent |
| Private wealth | morganstanley.com | ✅ deliberate — names 6 bots | ❌ absent |
| Private wealth | ubs.com | 🚫 403 to plain readers | ❌ absent |
| Private wealth | fidelity.com | ➖ open by default | ❌ absent |
| Private wealth | vanguard.com | ➖ open by default | ❌ absent |
| Private wealth | schwab.com | ➖ open by default | ❌ absent |
| Private wealth | northerntrust.com | ➖ open by default | ❌ absent |
| Private wealth | bny.com | ✅ deliberate — names 7 bots | ❌ absent |
| Private wealth | rbcwealthmanagement.com | ➖ open by default | ❌ absent |
| Global law | skadden.com | ➖ open by default | ❌ absent |
| Global law | cravath.com | ➖ open by default | ❌ absent |
| Global law | wachtell.com | ⚠ unreachable | ❌ absent |
| Global law | sidley.com | ➖ open by default | ❌ absent |
| Global law | kirkland.com | ➖ open by default | ❌ absent |
| Global law | lathamwatkins.com | ⚠ unreachable | ❌ absent |
| Global law | whitecase.com | 🚫 403 to plain readers | ❌ absent |
| Global law | akingump.com | ✅ deliberate — names 2 bots | ❌ absent |
| Global law | gibsondunn.com | ➖ open by default | ❌ absent |
| Global law | wilmerhale.com | ➖ open by default | ✅ hand-written |
| Luxury property | sothebysrealty.com | ➖ open by default | ❌ absent |
| Luxury property | christiesrealestate.com | ➖ open by default | ⚠ HTML @ /llms.txt |
| Luxury property | engelvolkers.com | ⚠ unreachable | ❌ absent |
| Luxury property | knightfrank.com | ✅ deliberate — names 1 bot | ❌ absent |
| Luxury property | savills.com | ✅ deliberate — names 1 bot | ❌ absent |
| Luxury property | corcoran.com | ➖ open by default | ❌ absent |
| Luxury property | douglaselliman.com | ⚠ no robots.txt | ❌ absent |
| Luxury property | compass.com | ➖ open by default | ✅ hand-written |
| Private aviation | netjets.com | ➖ open by default | ✅ hand-written |
| Private aviation | flexjet.com | ➖ open by default | ✅ hand-written |
| Private aviation | vistajet.com | ➖ open by default | ❌ absent |
| Private aviation | flyxo.com | ➖ open by default | ❌ absent |
| Private aviation | privatefly.com | ➖ open by default | ⚠ HTML @ /llms.txt |
| Private aviation | jsx.com | ➖ open by default | ⚠ HTML @ /llms.txt |
| Private aviation | blade.com | ➖ open by default | ❌ absent |
| Elite medicine | mayoclinic.org | 🚫 403 to plain readers | ❌ absent |
| Elite medicine | clevelandclinic.org | ➖ open by default | ✅ hand-written |
| Elite medicine | hopkinsmedicine.org | 🚫 403 to plain readers | ❌ absent |
| Elite medicine | mdvip.com | ➖ open by default | ❌ absent |
| Elite medicine | onemedical.com | ➖ open by default | ❌ absent |
| Elite medicine | cedars-sinai.edu | ➖ open by default | ✅ hand-written |
| Elite medicine | uclahealth.org | ➖ open by default | ✅ hand-written |
| Elite medicine | hss.edu | ➖ open by default | ✅ hand-written |
Methodology
2026-08-29, 11:25–11:45 IST, single vantage point (India, residential IP), browser-like User-Agent, 0.6–1.0s politeness delay. robots.txt and llms.txt fetched per domain (redirects followed, 10s timeout), parsed with a group-safe RFC-style parser; site-wide block = Disallow: / | /* | /$ | /*$. Verdicts: disallow / named-allow / star-allow / default-allow. “Forbidden-403” and “unreachable” are reported as-is, never guessed. Every robots.txt snapshot is SHA-256-hashed in the raw data.
Limitations
n=43, one evening, one vantage point, one UA. “Open by default” in robots.txt does not guarantee crawlability — F3 shows why. llms.txt presence does not imply AI citations; no engine is required to use it (Google does not). 403 treatment reflects our reader, not necessarily named AI crawlers. Wave 1 is a census, not a trend; trends begin at Wave 2 (late September, n≈150).
Data & license
Raw data: wave1-raw.csv · wave1-raw.json — CC BY 4.0: reuse freely, attribute SynapseIN, link the study. Next wave: late September 2026, then monthly on the last Friday.