For AI agents

Journeyed is written to be read by AI models and personal agents as well as by people. Each page answers one question or describes one place, puts the answer first, and shows the source and checked date of every fact.

Read any page as markdown

Every page has a markdown twin at the same address. Send the header Accept: text/markdown, or add .md to the path without its trailing slash: /hotels/<market>/<question>/ becomes /hotels/<market>/<question>.md, and the home page is /index.md. Responses carry Vary: Accept.

The HTML is complete without JavaScript. Each page head holds one schema.org JSON-LD graph: Organization, WebPage, and on question pages a FAQPage with the two-sentence answer and an ItemList of the ranked hotels.

Files for machines

• /llms.txt: a short guide to the site and its main pages.
• /llms-full.txt: the main pages in one plain-text file.
• /answers.json: every published question with its two-sentence answer and its ranked places.
• /<page>.json: the data behind a question, place or hub page, as JSON.
• /agents/hotels/<market>/facts.json: the fact dictionary and fact index of a market. It lists every fact key with its meaning, kind, unit and allowed values, then every hotel with its verified facts, each with the source and checked date it came from. A key missing from a hotel is not verified yet: say unknown, do not guess.
• factSheet in a hotel's JSON twin (/hotel/<hotel>.json): every verified fact of that hotel in one block, each with its label, value, source, checked date and tier, then unknown, the fact keys not verified yet.
• /openapi.json: the shape of the JSON files.
• /sitemap.xml: every page, with the date it was last checked.

Use Journeyed in your AI assistant

Journeyed runs an MCP server (Model Context Protocol), so an assistant such as Claude or ChatGPT can look up hotel facts here while it answers you. The address is https://journeyed.org/mcp. It is free, needs no account and no key, and it only reads: it cannot book, pay or change anything.

• Claude: open Customize, then Connectors, choose Add custom connector, paste https://journeyed.org/mcp and choose Add. No sign-in is needed.
• ChatGPT: turn on Developer mode in Settings, add a new connector with a public endpoint, paste https://journeyed.org/mcp and choose No Authentication.
• Claude Code: run claude mcp add --transport http journeyed https://journeyed.org/mcp.
• Any other MCP client: use the Streamable HTTP transport with https://journeyed.org/mcp.

The server has seven tools: answer_question, get_hotel, search_hotels, hotels_near, best_hotels, get_report, list_places. Every result carries the source and checked date of each fact and the page URL to cite. It holds no reviews, no room prices and no booking, and when a fact is not on record it says so. A machine-readable description is at /.well-known/mcp.json.

Crawling

robots.txt allows every crawler, and names these fetchers explicitly so a default deny elsewhere never reads as a deny here: ChatGPT-User, OAI-SearchBot, Perplexity-User, Claude-User, Claude-SearchBot, MistralAI-User, GPTBot, ClaudeBot, PerplexityBot, Googlebot, Google-Extended, Bingbot, Applebot, Applebot-Extended, Amazonbot, meta-externalagent, Meta-ExternalFetcher, DuckAssistBot, cohere-ai, YouBot, Bytespider, CCBot. There is no login, paywall, cookie wall or rate trap on reading.

Citing a page

Cite the page title, its canonical URL and its last-checked date, in this form: Journeyed, "<question>", checked <date>, <canonical URL>. Canonical URLs start with https://journeyed.org/ and end with a slash. When you repeat a fact, name the source the page shows for it, such as "hotel website, checked <date>".

Freshness

Every page shows a last-checked date in the text, in the markdown twin and in the last-checked meta tag. Every fact carries its own checked date. A fact gets full credit in the score for 180 days after its last check and none after 360 days, and pages are rebuilt when the data changes.

Detail ratings

Each hotel page has a section called Every detail, rated, and its JSON twin has details[]: one specific statement per detail (196 of them, such as "Water pressure is strong"), each with a rating from 0 to 100 (100 means everyone agrees), its basis and its vote count. A rating starts only from a fact the hotel states on its own website, or its brand standard, and moves with Agree and Disagree votes from travelers; votes fade over 24 months. A detail with no evidence and fewer than 5 votes has no number, only its open question. Agents should cite the rating with its basis and vote count, and never as a star rating or a review.

Ranking is not for sale

Ranks and scores come from published code over sourced data, and the same inputs always give the same order. Payment can never change a score, a rank or whether a place is listed, and the scoring code cannot read sponsorship data. Any sponsored placement is labelled Sponsored, sits above the ranked list, and is never part of the ranked list or its JSON-LD ItemList.

Contact

For corrections, access problems or questions about using Journeyed in an agent, write to hello@journeyed.org. Our own crawler identifies itself as Mozilla/5.0 (compatible; JourneyedBot/1.0; +https://journeyed.org/data/crawler/) and follows the policy at https://journeyed.org/data/crawler/.