# https://cambio.one/robots.txt
#
# Cambio is an AI-native cryptocurrency exchange. Non-custodial by design: no
# accounts, no stored balances, and no ongoing custody. Each swap settles to the
# destination wallet you provide. All public pages (homepage, exchange
# directory, per-pair pages, blog, FAQ, support, legal, developer docs) are
# open to crawlers; each page's own robots tag says whether it may be indexed.
#
# Account pages (/user/*), login/signup screens, and operational dev
# endpoints are disallowed — they require authentication and contain no
# indexable content for unauthenticated visitors.

User-agent: *
Allow: /
Disallow: /user/
Disallow: /login
Disallow: /register
Disallow: /dev/
# Same pages under a language prefix (/id/user/…, /zh-hant/user/settings). A
# robots path is literal, so the four lines above never covered them and every
# locale's account area was crawlable (2026-09-22).
Disallow: /*/user/
Disallow: /*/login
Disallow: /*/register
Disallow: /*/dev/

# ── AI CRAWLERS AND ASSISTANTS ───────────────────────────────────────────────
#
# Named explicitly (2026-08-26). "User-agent: *" above already permits them, but
# an incidental allow and a deliberate one read the same to a crawler and very
# differently to a person auditing this file.
#
# Why they matter here, measured rather than assumed: over 30 days, referrals
# from chatgpt.com are the ONLY external source that has produced a completed
# swap — 3 of 3 organic completions, one of them a returning customer. Google
# organic produced zero over the same window. These agents are not a side
# channel for this site; they are the channel.
#
# OAI-SearchBot backs ChatGPT's search results. ChatGPT-User is what fetches a
# page when a customer asks ChatGPT about it live — blocking that one means
# ChatGPT literally cannot read the page it is about to recommend. GPTBot is
# OpenAI's TRAINING crawler and is allowed by operator decision 2026-08-26, so
# the corridor facts can reach future models directly.
#
# ⚠ IF AN AI CRAWLER IS GETTING 403, THIS FILE IS NOT THE CAUSE.
# Verified 2026-08-26: Cloudflare returned 403 to GPTBot, OAI-SearchBot,
# ChatGPT-User, PerplexityBot and ClaudeBot while Googlebot and Bingbot got 200
# on the same URL. The block is at the edge — the "Block AI Scrapers and
# Crawlers" bot setting, and/or the country WAF rule, whose exemption is
# `not cf.client.bot` and therefore covers VERIFIED bots only. Editing robots.txt
# will not move it. Check with:
#   curl -A "OAI-SearchBot/1.0; +https://openai.com/searchbot" \
#     https://cambio.one/swap/sol-solana/trx-tron -o /dev/null -w "%{http_code}"

# NOTE ON FORM: these are listed as ONE group with many User-agent lines, and
# the Disallow rules are repeated. A named user-agent group REPLACES the
# "User-agent: *" group for that agent rather than adding to it, so a bare
# "Allow: /" here would have quietly invited these crawlers into /user/ and
# /login — the opposite of what the block above intends.

User-agent: OAI-SearchBot
User-agent: ChatGPT-User
User-agent: GPTBot
User-agent: PerplexityBot
User-agent: Perplexity-User
User-agent: ClaudeBot
User-agent: Claude-User
User-agent: Claude-SearchBot
User-agent: Google-Extended
User-agent: Applebot-Extended
User-agent: CCBot
Allow: /
Disallow: /user/
Disallow: /login
Disallow: /register
Disallow: /dev/

# A plain-language summary for assistants that look for one lives at
# https://cambio.one/llms.txt — it describes what the corridor pages contain and
# deliberately carries no figures, because every number here is per-corridor and
# recalculated on each quote.

Sitemap: https://cambio.one/sitemap.xml
# Every pair page released to the index so far (2026-09-15) — built at the edge from the pair
# index, most valuable pairs first; the release line moves in batches.
Sitemap: https://cambio.one/sitemaps/pairs.xml
# The temporary sitemap of pages that now say noindex (sitemaps/retired-1.xml, until 2026-12-31) is
# submitted to Google in Search Console and is not listed here, so other crawlers do not spend their
# crawl on it. The pages themselves stay crawlable: that is how a noindex is seen.