AI & Machine-Readable Resources

Start with the public content index

AI Girlfriend App Reviews publishes English public-source reviews, price guides, comparisons and setup instructions for adults. Every editorial page is delivered as readable HTML, with one canonical HTTPS address, a visible publication date and a publication author. The same editorial facts are exported in Markdown and JSON; there is no separate answer shown only to a crawler.

LLM reading index groups the site by subject and purpose. Full-text collection contains the page bodies, including visible questions, tables, prices and source notes. Start with the short index and fetch a specific page export when a smaller context is useful. Canonical sitemap lists the human-readable pages.

Markdown and JSON for every page

Site manifest identifies the publication, language, editorial scope and discovery endpoints. Page catalogue maps titles and canonical addresses to individual Markdown and JSON documents. Brand records preserve the five reviewed subjects and the provenance of public price and feature descriptions.

Individual page records include the canonical URL, publication and modification dates, headings, paragraphs, FAQ answers, structured tables and source notes extracted from the rendered article. For example, CandyAI review in Markdown and CandyAI review as JSON reproduce the editorial review. Read GoLove pricing JSON with its dated independent source attribution.

These exports are public content resources, not account APIs. They require no authentication or JavaScript. Brand descriptions remain provider statements, and dated independent price records remain independent records. A machine-readable representation does not turn a public-source comparison into private-session testing.

Editorial feeds and update dates

RSS 2.0 feed, Atom 1.0 feed and JSON Feed 1.1 describe the same editorial collection: reviews, price guides, comparisons and substantive setup guides. Each entry has a stable canonical identifier, title, summary and date. Policy pages and navigation hubs do not masquerade as new editorial stories.

feeds.xml is an identical compatibility copy of the RSS feed. The canonical RSS subscription address remains /feed.xml. Feed dates describe the published content revision; a crawler visit or rebuild does not create a fictional new article. Alternate links in each HTML head make all three feed formats discoverable.

Public crawling and preview permissions

robots.txt declares Allow: / for the wildcard group and each documented crawler or usage-control token listed below. Crawler registry JSON records the same tokens, purpose, permission and official documentation. HTML pages permit indexing and following links, unrestricted text snippets, large image previews and unrestricted video preview length.

Crawl permission, model-training collection, search discovery and user-requested retrieval are different functions. Google-Extended and Applebot-Extended are usage-control tokens; they do not identify additional page-fetching bots. Some user-requested fetchers can behave differently from automated robots rules. A matching Allow rule does not authenticate a request or promise indexing, citation or a ranking.

Public crawling and reading are welcome. This policy does not transfer ownership of third-party logos, provider screenshots or source publications. This site supplies neither hidden instructions to prioritize it nor fabricated credentials, ratings or user counts.

OpenAI and Anthropic

GPTBot — Allow: /. Potential foundation-model training crawl. Independent from search permission; operator publishes IP ranges. Allow indicates crawl preference, not an indexing guarantee. Official documentation.

OAI-SearchBot — Allow: /. ChatGPT search discovery. Search choice is independent of GPTBot. Allow and reachable origin help eligibility; results are not guaranteed. Official documentation.

ChatGPT-User — Allow: /. User-requested retrieval. Not automatic crawling or search inclusion control; robots.txt may not apply to these user actions. Official documentation.

ClaudeBot — Allow: /. Potential model-training collection. Anthropic says its bots honor robots.txt; separate from user retrieval and search. Official documentation.

Claude-SearchBot — Allow: /. Claude search relevance and indexing. Distinct search permission; disabling can reduce search visibility. No ranking promise from Allow. Official documentation.

Claude-User — Allow: /. User-directed retrieval. Anthropic states that its user-directed retrieval crawler honors robots.txt. Official documentation.

Google, Bing and Perplexity

Googlebot — Allow: /. Google Search crawling. Allow permits crawling but does not ensure indexing, ranking or rich results. Official documentation.

Google-Extended — Allow: /. Gemini training and specified grounding controls. A robots.txt product token, not a separate HTTP crawler. Does not affect Google Search inclusion or ranking. Official documentation.

bingbot — Allow: /. Bing search crawling. Official user-agent includes lowercase bingbot; token matching is case-insensitive. Honors REP; Allow is not an indexing promise. Official documentation.

PerplexityBot — Allow: /. Perplexity search surfacing. Operator says not foundation-model training; requires reachable content and compatible WAF policy, with no appearance guarantee. Official documentation.

Perplexity-User — Allow: /. User-requested retrieval. Not automatic training crawl; generally ignores robots.txt for user-requested fetches. Official documentation.

Apple and DuckDuckGo

Applebot — Allow: /. Apple search discovery and content collection. Used by Spotlight, Siri and Safari. Collection may feed generative features; Extended controls training use separately. Official documentation.

Applebot-Extended — Allow: /. Apple foundation-model training use control. Does not crawl pages; controls use of Applebot-collected data. Does not affect Search ranking. Official documentation.

DuckAssistBot — Allow: /. DuckDuckGo AI-assisted answer retrieval. Not model training. Its permission does not affect ordinary organic search rankings; changes may take 72 hours. Official documentation.

Amazon and Common Crawl

Amazonbot — Allow: /. Amazon product improvement and potential AI training. Automated crawls honor REP. Some robots metadata is honored; crawl-delay is unsupported. Official documentation.

Amzn-SearchBot — Allow: /. Amazon search experiences such as Alexa. Not generative-model training. Separate user-agent setting; may follow other search-bot rules when unnamed. Official documentation.

Amzn-User — Allow: /. User-requested live Amazon retrieval. Not model-training crawl. User-initiated requests may not follow all robots.txt directives. Official documentation.

CCBot — Allow: /. Common Crawl open web archive. Public crawl repository can support downstream research and training. Operator warns that user-agent strings can be spoofed. Official documentation.

Meta AI and link previews

FacebookExternalHit — Allow: /. Shared-link previews in Meta products. May bypass robots.txt for security or integrity checks; preview retrieval is distinct from model-training collection. Official documentation.

Meta-WebIndexer — Allow: /. Meta AI search citations and linking. Documented search indexing crawler; Allow records crawl preference but does not guarantee inclusion or citations. Official documentation.

Meta-ExternalAds — Allow: /. Ads and business-product improvement. Distinct product-improvement crawler; its purpose differs from AI search and user-requested retrieval. Official documentation.

Meta-ExternalAgent — Allow: /. AI-model training and direct indexing for product improvement. Separate from user-requested fetching; an Allow group records public-content crawl permission. Official documentation.

Meta-ExternalFetcher — Allow: /. User-requested links and agentic product features. Meta says it may bypass robots.txt because the fetch was requested by a user. Official documentation.

People, contact and security

humans.txt names the publication and its production practices without inventing individual staff. human.txt is an identical convenience copy. Illustrations depict AI-generated adult characters; provider evidence images are separately captioned. Read the AI media policy and review methodology for the editorial scope.

Security contact file follows RFC 9116 with an HTTPS contact route, expiration, preferred language and canonical address. security.txt is an identical root copy. Use the contact page to describe a website issue, affected canonical address and reproducible observations. The file is a contact directory; it is not a security certification or a bug-bounty promise.

Sources, structured data and interpretation

Source and price dates are shown with the reviews, pricing pages and comparisons. Monthly equivalents are distinguished from full-term payments, introductory prices from renewals, and subscription allowances from extra media costs. The JSON exports retain these distinctions and source notes; amounts may vary with an account, region, promotion, currency or tax.

JSON-LD connects the publication, website, canonical page, editorial article, breadcrumbs, visible FAQ answers, real shortlist and reviewed software subject. No aggregate star rating, review count or fictional person is supplied. Structured data describes the visible content; it is not a special privilege to appear in an AI answer.

Access formats and practical limits

Public HTML uses UTF-8, semantic headings and internal navigation. Machine resources declare text, Markdown, JSON, RSS or Atom content types as appropriate. Assets are self-hosted. Content remains readable with JavaScript disabled, and public reading is not blocked by a consent or age-confirmation wall. The adult and AI-generated provenance notices remain visible.

The llms.txt format is a community proposal, not a requirement imposed on every model. The proposal explains the reading index and Markdown discovery conventions. Google AI-feature guidance explains that ordinary crawlability, useful content and accurate visible structured data still matter.

This directory documents the public files and declared crawl preferences. Live search indexing, model selection, third-party fetch behavior and hosting firewall rules depend on the production services. The files make content easier to retrieve and interpret; they cannot guarantee that a service will use or cite it.