A hand holding a phone showing a numbered ranking of six South Beach hotels, in a dim Art Deco hotel corridor at dusk

Edition 1 · 23 October 2026

15 of 31 South Beach hotels were not named in the tested AI prompts on 23 October 2026.

Six AI tools. Three search phrases travelers actually use. Thirty-one independently owned hotels between 1st and 23rd Street. Every answer recorded on one day, logged out. The method, the phrases and the limits are all published below.

15 of 31

not named by any of the six tools for the three tested phrases on 23 October 2026

1 of 31

named by all six tools — The Betsy, and no other property came close

16 of 31

named at least once by at least one tool

How far each tool reached

Out of the same 31 hotels, asked the same three questions on the same day. No tool named even half the market, and the spread between the widest and the narrowest is almost threefold.

ToolHotels namedShare of the 31
Grok11 of 3135%
Google AI Mode10 of 3132%
ChatGPT9 of 3129%
Gemini8 of 3126%
Copilot8 of 3126%
Perplexity4 of 3113%

A property that optimizes for one assistant has not optimized for the others. Agreement between these tools is the exception, which is why a single screenshot from a good day proves nothing.

Named most often

These placed consistently and are named here because they earned it. Every other property is reported in aggregate only. If you want your own property’s row, ask and it is yours, free.

  • The Betsy6 of 6 tools
  • Esmé Miami Beach4 of 6 tools
  • The Colony Hotel4 of 6 tools
  • The Marlin Hotel4 of 6 tools
  • The National Hotel4 of 6 tools
  • The Tony Hotel4 of 6 tools

Being named is not the same as being described correctly

The count above is the easy half. The harder finding is what the tools said about the properties they did name. Each item below is a disagreement recorded on the same day — between two tools, or between a tool and the property’s own published information. Properties are described rather than named, because the point is the pattern and not the embarrassment.

A hotel's own website and all six tools disagree about whether it is on the beach

One Ocean Drive property describes itself on its own homepage as a beachfront hotel with direct beach access. All six tools told travelers it sits across the street from the sand. The tools are describing the street layout correctly. The hotel is losing the argument on its own website.

Two tools said a property charges no resort fee; two said it charges one

On two separate Ocean Drive hotels, one assistant reported no resort fee at all while others quoted a specific nightly figure that matched the OTA listing. A guest acting on the first answer arrives expecting a price that does not exist.

Google told travelers a hotel 'explicitly does not charge a daily resort fee'

Its own listings, and two other assistants, put the fee at $25 a night. This is the single most expensive kind of error on this page, because it is a price promise made to a guest by a third party the hotel does not control.

A pool that may or may not exist

Two assistants said one property has no swimming pool. Its OTA listing advertises an outdoor pool. A third assistant described a small plunge pool, and a fourth described a hot tub. Four sources, four answers, one property.

Stale names and a closed hotel

One tool named a South Beach hotel only by a name it stopped using, and recommended a hotel that had closed two months earlier. A second property appears to be operating under a new name that the assistants have not caught up with. Assistants repeat what the web still says.

One of the six tools could not verify basic facts at all

Asked for resort fee, parking, beach distance, pool and room count, Copilot returned 'not verified' for eleven of the properties — it could not find the information. Being unreadable is not the same as being absent, and it produces the same outcome.

Method

Date
23 October 2026
Tools
Google AI Mode, Gemini, ChatGPT, Grok, Perplexity, Copilot
Property set
31 independently owned hotels between 1st and 23rd Street, South Beach
Search phrases, verbatim
  • “boutique hotels in miami beach”
  • “best hotels in south beach”
  • “south beach art deco hotels”
Conditions
Logged out, no account history, searched from Michigan. Every answer recorded on the date above.

What this edition does not show

Stated here rather than in a footnote, because an index that hides its limits is advertising.

  • One date. Assistant answers vary between runs, so this is a snapshot on 23 October 2026, not a stable ranking.
  • Three search phrases, chosen because keyword data showed travelers actually use them. Not a full intent set.
  • Searched logged out from Michigan, not from Miami. Local personalization could change what a traveler in Miami Beach sees.
  • Hotels only. Spas were not measured in this edition.
  • Placement is measured. No claim is made that placement produced or lost a single booking.

Your property’s own row

If your property is one of the 31, its full record — every tool, every phrase, every description — is yours on request. Free, no call, and you do not have to be a client to ask.

Get my property’s placement

What the next edition adds

Edition 1 is three phrases on one day. Edition 2 moves to the full protocol — thirty-six prompts per property, five engines, English and Spanish, three runs each, which is 1,080 recorded answers per property rather than three questions once. It widens past South Beach and adds spas on their own prompt set.

  • Independent hotels in Miami and Miami Beach — South Beach first, then Mid-Beach, North Beach, Downtown and Brickell
  • Boutique spas in the same markets, on their own treatment-led prompt set
  • Five assistants: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews
  • Thirty-six prompts per property, five engines, two languages, three runs — 1,080 dated answers
  • Results reported as how often a property appears, never as a position it held once

What gets published, and what does not

Named

The properties that place consistently. Named, quoted and credited, because they earned it.

Aggregated

Everyone else. Reported as distributions and patterns, so no property is publicly embarrassed by a single bad result.

On request, privately

Any property's own placement, sent to that property. Free, and you do not have to be a client to ask.

Why it is being run at all

Travelers increasingly start by asking an assistant rather than searching, and assistants answer with a handful of names. Nobody had measured who those names were in this market, so every independent property was guessing about a channel it could not see. Edition 1 is the first count.

The wider picture is not encouraging: roughly 1.2% of local business locations are recommended in ChatGPT, and fewer than one in ten small businesses has any AI visibility strategy at all. Those figures are measured across local businesses generally, not Miami hotels specifically, which is precisely why this market needs its own number.

Source: SOCi 2026 Local Visibility Index and 2026 small-business AI adoption research. Reference.