Context
Cloudflare announced general availability of its managed AI Search pipeline, adding native image embeddings, OCR for scanned PDFs, larger files and a documented hybrid retrieval path.
What changed
AI Search reached general availability on October 1, 2026, with multimodal embeddings, image-caption fallback, scanned-PDF OCR and support for files up to 10 MiB. Cloudflare documented optional query rewriting, parallel vector and keyword retrieval, fusion and reranking.
Why it matters
The release provides a concrete implementation of how parsing, multimodal indexing, retrieval and reranking govern which evidence reaches downstream answer generation in developer-built systems.
Visibility OS interpretation
Visibility OS interpretation: use this as evidence of a retrieval architecture and separate ingestion, ranking and answer generation; do not extrapolate to ChatGPT Search, Gemini or Google ranking.
What remains uncertain
What are real-world retrieval-quality tradeoffs across embedding models and modalities? Do downstream applications cite retrieved sources faithfully? No consumer search citation effect is established.