# Content Signals (contentsignals.org / draft-romm-aipref-contentsignals):
# declared usage preferences for this site's content. Deliberately permissive on all three.
#   ai-train=yes  -> may be used to train AI/ML models. Note this brings us no traffic and no
#                    citations (training data is baked into model weights, not linked back);
#                    it is allowed as an open-by-default stance, not as an acquisition channel.
#   search=yes    -> may be indexed for search results (never disable: this is our SEO)
#   ai-input=yes  -> may be used as input/grounding for AI-generated answers. This is the one
#                    that actually drives visits: real-time assistants read the live page and
#                    cite us with a link, so prices/stock are always current.
# Consistent with the ChatGPT-User / Claude-User / Perplexity-User groups below, which are
# the real-time browsing agents (not the training crawlers).
# Applies to the User-agent group it is declared in (here: all crawlers).
User-agent: *
Content-Signal: ai-train=yes, search=yes, ai-input=yes
Allow: /
Disallow: /*?*srsltid=
Disallow: /api/*

# Search results pages: low-value, effectively infinite URLs. Route slugs are translated per
# locale, so a bare "/search*" only ever matched the en-* domains — every other region served
# /pesquisa, /buscar, /recherche or /onderzoek and was crawled freely. One rule per slug so the
# same file works on every regional deployment.
#
# These rules are root-anchored, which only covers each region's DEFAULT locale: a non-default
# locale is served under a 2-letter path prefix (routes/CLAUDE.md, "Locale prefixing"), so
# /ca/cerca was crawled freely while root /cerca — which exists on no deployment — was blocked.
# Per config/locales.php the only non-default locales active in production are ca-es (iservices.es),
# nl-be (iservices.be) and en-nl (iservices.nl); every other region serves its default locale
# unprefixed and its en-* locale is backoffice_only, injected on local/demo boxes only. Hence
# exactly three prefixed rules. Do NOT collapse them into a leading wildcard — see /recon below.
Disallow: /pesquisa*        # pt-pt
Disallow: /buscar*          # es-es, es-ic
Disallow: /cerca*           # ca-es  (root form only reachable on local/demo)
Disallow: /recherche*       # fr-fr, fr-be
Disallow: /onderzoek*       # nl-nl, nl-be
Disallow: /search*          # en-*   (root form only reachable on local/demo)
Disallow: /ca/cerca*        # ca-es on iservices.es
Disallow: /nl/onderzoek*    # nl-be on iservices.be
Disallow: /en/search*       # en-nl on iservices.nl

# NOT blocked: the comparator (/comparar?products=a,b). Its URLs are combinatorial in theory,
# but nothing in the rendered HTML links to them — combinations are built client-side from
# localStorage and navigated via window.location, so crawlers have no path to discover them.
# Revisit only if Search Console starts reporting indexed ?products= URLs.

User-agent: ChatGPT-User
Allow: /

User-agent: Claude-User
Allow: /

User-agent: Perplexity-User
Allow: /

Sitemap: https://iservices.es/sitemap.xml