1. Article classification and sentiment analysis
Article bodies are analyzed with AI for entity matching, sentiment classification, and topic extraction. AI classifications may be wrong; corrections can be requested via /correction-policy.
2. Summaries on detail pages
The summary sentences on celebrity and brand pages are not written by AI. They restate aggregated figures (rank, score, article count) using fixed rules and are labelled 'data summary' on screen.
3. Backers100 Score computation
The scoring algorithm is identified by an algorithm version (e.g., v1.0.0); change history is available at /methodology/changelog.
4. Limits of AI analysis
AI may misclassify due to context misunderstanding, failure to detect satire, and similar issues. Results are for reference and cannot be used as legal evidence. Suspected misclassifications can be reported via /correction-policy.
5. AI training on user data
User cheer, donation, and search data are used only for Backers100's data analysis and are not provided as training data to external AI models.