Can AI Assistants Read America's Vacation Rental Websites?
Published October 20, 2026 · Corrected October 21, 2026 · Threshold
We checked 1,913 independent vacation rental managers' websites in 49 states in October 2026. Only 1 in 7 (15.0%) tells an AI assistant, in the format it looks for, that it offers places to stay. Four in five can be read but carry no such label, and about 1 in 25 hides from, blocks or turns away AI assistants altogether.
Key findings
Of the 1,863 sites that answered, 280 (15.0%) are labelled as a place to stay; the other 85% leave an AI assistant to guess what they offer, or show it nothing at all.
| What an AI assistant finds on the home page | Sites | Share of sites that answered |
|---|---|---|
| Readable, but not labelled as a place to stay | 1,508 | 80.9% |
| Readable and labelled as a place to stay | 280 | 15.0% |
| Blocked: the site tells AI assistants to stay away | 36 | 1.9% |
| Nearly empty: the text only appears once a browser runs the page's scripts | 17 | 0.9% |
| Turned away by the site's security service | 16 | 0.9% |
| Too large to read in full (more than 2 MB of code) | 6 | 0.3% |
A further 50 sites (2.6% of all 1,913) didn't answer at all.
- Blocking is mostly deliberate, and mostly aimed at ChatGPT. Of the 36 sites that block AI assistants in their robots.txt file, 35 block ChatGPT's crawlers, 32 Claude's, 27 Google Gemini's and 5 Perplexity's.
- Security services block assistants by accident. 14 of the 16 sites that turned the check away sit behind Cloudflare, whose bot settings can shut out AI assistants even when the site's own robots.txt welcomes them.
- Some sites have little for an assistant to read. The typical readable home page offers about 7,000 characters of text, but 97 offer fewer than 2,000, a few short paragraphs.
- Some pages are empty until a browser builds them. 17 sites put their text on the page only once a browser runs their scripts, which most AI crawlers never do, so an assistant sees a nearly blank page.
By state
Source: Threshold website check of 1,913 sites, October 2026.
The gap between states is wide even among big markets: Colorado, North Carolina and South Carolina label one site in five, while New York labels 2 of 97.
By region
The South leads and the Northeast trails: managers in Texas and Oklahoma are nearly three times as likely as those in the Northeast to label their sites.
| Region | States | Sites that answered | Labelled as a place to stay |
|---|---|---|---|
| South Central | TX, OK | 116 | 20% |
| Southeast | FL, GA, AL, MS, SC, NC, TN, KY, VA, WV, LA, AR | 674 | 18% |
| Mountain | CO, UT, AZ, NM, NV, ID, MT, WY | 357 | 15% |
| Pacific | CA, OR, WA, AK, HI | 294 | 16% |
| Midwest | OH, IN, IL, MI, WI, MN, IA, MO, ND, SD, NE | 181 | 11% |
| Northeast | ME, NH, VT, MA, RI, CT, NY, NJ, PA, MD, DE | 241 | 7% |
Why it matters
More travelers now ask an AI assistant such as ChatGPT, Claude or Gemini to find and compare places to stay, and the assistant can only recommend what it can read and understand.
When an assistant answers “find me a beach house for six in Gulf Shores,” it reads web pages for the answer. A page labelled with structured data says, in a form software reads without guessing, that this is a vacation rental, where it is and what it offers. An unlabelled page leaves the assistant to infer all of that from the wording, and a page that blocks it, or is empty without scripts, gives it nothing.
So a manager whose site an assistant can't read clearly is easier for it to pass over, in favour of sources it can read, often a listing on a booking site with that site's fees.
What managers can do
Most of these fixes are an afternoon's work for a web developer, and the first three cost nothing.
- Check what an assistant sees. A free check such as threshold.rentals/check reads a site the way an AI crawler does and reports which of the problems above it has.
- Unblock the assistants, if you want them. Remove the lines in robots.txt that name GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Google-Extended or PerplexityBot.
- Check your security service. If you use Cloudflare, look at its bot settings (Security, then Bots or AI Crawl Control): “Block AI bots” and Bot Fight Mode can turn assistants away even when robots.txt allows them.
- Label your pages. Add structured data: VacationRental on each rental's page and LodgingBusiness for the company. Google reads the same labels.
- Put the words in the page. If your rental platform builds pages in the browser, ask for server-side rendering, so the text and rentals are in the page itself.
- Say it in plain text. A few sentences on what you offer, where, and for how many guests give an assistant something to quote.
How we checked
We found independent vacation rental managers in every state, then read each one's home page the way an AI crawler does, all in the same week of October 2026.
- The sample. 1,913 local and regional managers with their own websites in 49 states, found through chamber of commerce and visitor bureau lists, local directories and web searches. National brands and their franchises, booking sites, hotels and single-owner listings were left out. We found no independent manager with its own site in Kansas.
- The test. For each home page we checked whether the site's robots.txt turns away the main AI assistants' crawlers, whether the page has at least 600 characters of readable text without running scripts, and whether it carries structured data labelling it as lodging (schema.org types such as VacationRental, LodgingBusiness, Accommodation, House or Hotel).
- The limits. The sample isn't a census: it leans toward managers that are easy to find online, and small states have few sites, so their figures are shown only where 20 or more sites answered. We checked home pages only; a site may label its individual rental pages even when its home page isn't labelled.
Correction, October 21, 2026: the first version counted 20 Washington websites as separate managers. They belong to one business, which runs a network of local sites. They are now counted once, which brings the total to 1,913 sites, 1,863 of which answered, and changes the national shares by less than one point. Washington is now 3 of 25 sites labelled. An earlier finding that one website platform hid much of Washington's market was wrong and has been removed.
You're welcome to quote these figures with credit to Threshold and a link to this page. Journalists can ask for the figures by state at hello@threshold.rentals.
The free website check runs the same test on your own site in under a minute.