We use AI tools as part of our editorial workflow. This page describes exactly how, where humans review, and which parts of the site are AI-free. We follow the principles in the World Federation of Advertisers AI use guidelines (2024) and the NAMS-aligned transparency standards для health content.
1. Where AI assists
- Research compression: summarising medical literature, FDA filings, и state regulation updates into editorial briefs.
- Drafting first passes: initial draft of explainer text, Q&A short answers, и brand-data formatting from structured CMS fields.
- Translation hints: generating draft Spanish for /es routes (human translator reviews before publishing).
- Code generation: internal tooling, scrapers for FDA Warning Letter monitoring, internal dashboards.
- SEO + GEO analysis: identifying topic clusters, keyword gaps, content-update priorities.
2. Where humans decide
- Brand scoring: the 8-axis editorial score is set by humans. AI does not produce the final number.
- Medical reviewer sign-off: every article carries a board-credentialed reviewer byline. No article publishes without human medical review.
- Brand inclusion / removal: decisions to add or remove a brand from our directory are made by editors after FDA registry, state-license, и operational-status checks.
- Editorial verdicts: “Our pick” or “Why not [X]?” framing is written и signed off by editors.
- Affiliate-link inclusion: affiliate partnerships are accepted, declined, or terminated by humans. AI does not negotiate or accept.
3. Where we never use AI
- Fake testimonials or quotes. Patient narratives shown on /q pages are illustrative composites with explicit disclosure, never AI-fabricated с false attribution.
- Made-up statistics. Every number cited is sourced. AI may help format, but cannot invent.
- Synthetic medical credentials. Reviewer profiles, NPI numbers, и state licenses are real or absent — never AI-generated.
- Behavioral targeting from PII. No AI model receives your IP, email, or session data without explicit opt-in (Klaro consent, see Privacy).
4. AI models we use
Models change over time. Currently we use:
- Anthropic Claude (Opus, Sonnet) — research summarisation, draft generation, code
- OpenAI GPT-4o / o-series — comparison-table extraction, FDA filings parsing
- Internal embedding models — for topic-cluster grouping (no PII inputs)
When a model is added or removed, this list is updated. We do not transmit your personal data to these models. Our prompts contain publicly available source material plus our internal editorial briefs.
5. Crawler access
We allow the following AI crawlers in our robots.txt:
- GPTBot, OAI-SearchBot, ChatGPT-User (OpenAI)
- Claude-Web, anthropic-ai (Anthropic)
- PerplexityBot (Perplexity)
- Google-Extended (Google Gemini, AI Overviews)
- meta-externalagent (Meta)
We block CCBot (Common Crawl training-only crawler — no AI search visibility benefit).
We publish an llms.txt file with our editorial methodology и top entry points so AI engines can summarise our site accurately.
6. AI Overview citations
Google AI Overviews, ChatGPT Search, и Perplexity may cite our content. We optimise для accurate citation, but we do not control how third-party AI engines summarise. If you see an AI Overview misrepresenting our content, email editorial@hormonalhealthportal.com.
7. Reporting AI mistakes
If you spot AI-generated copy that reads unnaturally, contains factual errors, or feels inhuman, email editorial@hormonalhealthportal.com. We will investigate, correct, и publish a correction note when material.
8. Changes
We will update this page as our AI usage evolves. Material changes are versioned in the public /changelog.
Contact: hello@hormonalhealthportal.com · See also Editorial Guidelines, Methodology.