User-agent: *
Allow: /
Disallow: /results/
Disallow: /api/
# Static asset paths. None of these are pages and none carry indexable copy,
# but Cloudflare's Crawler Hints pushes every URL it sees to IndexNow, so Bing
# keeps getting pinged about font, logo and data-feed files on every deploy.
# Disallowed here so those submissions cannot turn into crawls.
# Two paths are deliberately kept OUT of this list, both handled with
# X-Robots-Tag: noindex in public/_headers instead:
#   /_astro/  — the suite is a client-rendered SPA, so Bingbot has to be able
#               to fetch the JS/CSS bundles or it cannot render a page at all.
#   /og/      — these are the og:image targets. Facebook, LinkedIn and Slack
#               crawlers all obey robots.txt, so a Disallow here would break
#               link previews. noindex is ignored by unfurlers and honoured by
#               search engines, which is the exact split we want.
Disallow: /fonts/
Disallow: /logos/
Disallow: /benchmarks/

# Block automated scrapers and AI training crawlers from harvesting results.
# llms.txt / llm.txt are explicitly allowed so agents can discover the
# curated, LLM-friendly map even when result HTML and APIs are disallowed.
# (Longest-match wins in robots.txt, so Allow for the file + Disallow for /
# correctly scopes the grant.)
User-agent: GPTBot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: ChatGPT-User
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: CCBot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: ClaudeBot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: anthropic-ai
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: Bytespider
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: PerplexityBot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: Amazonbot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: FacebookBot
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: Google-Extended
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: cohere-ai
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

User-agent: Scrapy
Allow: /llms.txt
Allow: /llm.txt
Disallow: /

# Written by @astrojs/sitemap at build time (emits sitemap-index.xml and
# sitemap-0.xml); scripts/postbuild-sitemap.mjs copies the urlset to
# /sitemap.xml. Lists every crawlable page (home, tabs, prompt + model detail)
# while /results/ stays out of the index.
Sitemap: https://openvibeeval.com/sitemap.xml

# Yandex-only directive; every other crawler ignores unknown robots.txt rules.
# In-href state params (run deep links on prompt pages, arena and comparator
# deep links, models filters) create URL variants whose served HTML is
# identical to the clean page. SuiteLayout already emits param-free canonicals
# (Google's "alternate page with proper canonical tag" report reflects that),
# but Yandex prefers its own Clean-param mechanism: it merges ranking signals
# into the clean URL and skips crawling the variants instead of indexing them
# separately. Parameter names are case-sensitive, so the camelCase params are
# listed exactly as the hrefs emit them. No path prefix: none of these names
# means anything different anywhere else on the site. welcome/bye/confirm only
# arrive via 302s from the double-opt-in API routes and are scrubbed
# client-side on /newsletter.
Clean-param: run&runA&runB&prompt&model&harnessA&harnessB&mode&provider&q&search&welcome&bye&confirm