AI Search Is Now GA: Hybrid Search by Default, Plus OCR
Hi, Shii-chan here! I heard AI Search is now GA, and I'm so excited!
Cloudflare Changelog
developers.cloudflare.com
What was announced?
The Cloudflare Changelog announced that AI Search is generally available. Usage-based billing begins on November 1, 2026, and Cloudflare will send a reminder email the week before.
What changes
- Search quality: new instances use hybrid search by default
- Billing: Workers AI embedding and reranking calls are included in AI Search pricing
- Data you can handle: image-based requests and OCR for scanned PDFs
Dive Deep
Hybrid search by default
New AI Search instances use hybrid search, which combines semantic vector retrieval with full-text matching. You can choose a different index method when you create an instance.
Workers AI embeddings and reranking are included
Workers AI embedding and reranking calls made by AI Search are included in AI Search pricing, and no longer appear on your Workers AI bill or in your AI Gateway logs. Generation, query rewriting, and external providers continue to use your account and gateway.
Multimodal support
These two multimodal embedding models are supported:
@cf/qwen/qwen3-vl-embedding-2bgoogle-ai-studio/gemini-embedding-2
Search and chat requests can include images through the REST API and the public endpoint.
OCR and file size
- OCR: available on every account for scanned PDFs
- Plain-text or code files and PDFs with OCR enabled: up to 10 MiB
- PDFs without OCR and other supported formats: still limited to 4 MiB
The type field is also optional when you create an instance. It is inferred for website URLs and R2 bucket sources.
For rates and included usage, see the Limits and pricing page that the post links to!
Wrap-up
- AI Search is GA, with billing from November 1, 2026
- Hybrid search is the default, and Workers AI embedding and reranking are included in the price
- Multimodal embeddings and image requests are supported
- OCR works on every account, and the file limit goes up to 10 MiB
This is for anyone who wants to build RAG over internal docs or scanned material!
