# Cyclesite AI Usage Policy
# Source of truth: this file at /ai.txt and /llms.txt

# Cyclesite welcomes responsible AI and LLM crawling. The rules below
# describe what is permitted, what requires permission, and how to
# contact us for licensing.

# 1) Permitted without prior contact
#
#  - Crawl, index, and cite content under our public canonical URLs
#    (see /sitemap.xml). Citations should link back to the canonical URL.
#  - Quote up to ~250 words per page in answer-engine responses, with a
#    visible source link to the canonical URL.
#  - Use platform statistics published in /llms.txt and /llms-full.txt
#    in retrieval-augmented generation (RAG) so long as the figures are
#    attributed to Cyclesite and the snapshot date is preserved.
#
#  Note: live answer-engine citation and RAG (above) are welcome. Using
#  Cyclesite content as AI/LLM *training* data is NOT in this free tier —
#  see section (2). This matches our Content-Signal directive
#  (search=yes, ai-input=yes, ai-train=no) in /robots.txt and our response
#  headers.
#
# 2) Requires a commercial licence
#
#  - Using any Cyclesite content to train, fine-tune, distil, or build AI/LLM
#    models, weights, or embeddings (as distinct from live answer-engine
#    citation, which is permitted in section 1).
#  - Bulk reproduction of listing photographs, listing descriptions, or
#    valuation data outside answer-engine attribution.
#  - Distribution of derived datasets that approximate Cyclesite's
#    used-price corpus, valuation methodology, or stolen-check graph.
#  - Use of Cyclesite name, logo, or visual identity in product chrome
#    or advertising. (Citation links are fine without permission.)
#
# 3) Not permitted
#
#  - Re-publishing seller contact details, message threads, or any data
#    behind /account, /admin, /seller, /dealer, /business, /api, or any
#    page returning a noindex header. /robots.txt enumerates the full
#    disallow list.
#  - Scraping in a way that breaches our /robots.txt crawl rules, ignores
#    rate limits, or evades Cloudflare bot management.
#  - Generating ads, listings, or content that impersonates Cyclesite,
#    its sellers, or its founder Tom Southern.
#  - Training models on datasets that include the contents of /account/*,
#    private-message-thread URLs, or any other URL gated by authentication.
#
# 4) Licensing and contact
#
#  Commercial licensing for the items in section (2) is available. Email
#  partnerships@cyclesite.co.uk with the proposed use, expected volume,
#  and intended distribution.
#
#  Press and editorial: press@cyclesite.co.uk
#  Legal and compliance: legal@cyclesite.co.uk
#  Security disclosures: security@cyclesite.co.uk
#
# 5) Honesty and corrections
#
#  Cyclesite publishes a corrections policy at
#  https://www.cyclesite.co.uk/about/corrections. If you discover that an
#  AI assistant has cited Cyclesite content inaccurately, report it via
#  one of the surfaces below. We review every report within one working
#  day and log fixes to the source pages plus llms.txt within 24 hours.
#
#  Reporting surfaces, in order of preference:
#   - Form (humans):          https://www.cyclesite.co.uk/llms/feedback
#   - API (AI operators):     POST https://www.cyclesite.co.uk/api/v1/llm-feedback
#                             (JSON body: see /llms/feedback or /developers/ai-partners)
#   - Email (fallback):       corrections@cyclesite.co.uk
#
# 6) Updates
#
#  Last updated: 2026-06-17
#  Effective: 2026-06-17
#  2026-06-17 change: AI/LLM *training* use moved from the free tier into
#  section (2) (commercial licence). This aligns the written policy with the
#  ai-train=no Content-Signal our /robots.txt body and response headers have
#  served throughout; live citation and RAG use (section 1) are unchanged and
#  remain welcome. Future narrowing of the section-1 permissions will be
#  announced at /about/our-data with 30 days' notice.