October 1, 2026, (Inside AI) — Cloudflare announced today that its AI Search product is now generally available, marking a significant milestone for the managed retrieval service that combines Workers AI, Vectorize, R2, and Browser Run into a single pipeline. The company also revealed that billing will commence on November 1, 2026, following a preview period that began over a year ago.
The general availability release introduces native multimodal embedding and retrieval, allowing developers to search images directly by their visual content rather than relying solely on text captions. This shift addresses a long-standing limitation in AI-powered search systems, where visual details such as product texture, chart relationships, and document layout often get lost when images are converted to text.
Cloudflare's AI Search now embeds image pixels directly using the Qwen3-VL-Embedding model, while also retaining captions for textual understanding. The system leverages Matryoshka Representation Learning (MRL) to keep embeddings efficient and storage manageable. According to the company, this dual approach enables searches that are difficult to express with words alone, such as finding visually similar products or matching screenshots.
"An image is more than the sentence used to describe it," Cloudflare stated in its announcement. "Product texture, screenshot state, chart relationships, document layout, and fine visual detail can all disappear when pixels are compressed into a caption."
Read: AWS Launches Amazon CloudWatch Omni for AI Agent Observability
The update also expands file support to 10 MiB for text files and PDFs, up from 4 MiB, and adds optical character recognition (OCR) for scanned documents. OCR is available to all accounts and will be billed as image processing ingestion tokens under the new pricing model.
Cloudflare's pricing structure for AI Search is designed for predictability. Customers pay for three components: content ingestion, data storage, and queries. The processing steps in between, including parsing, chunking, embedding, keyword indexing, and reranking, are included at no extra cost. The free tier now offers 1,000 semantic queries and 1,000 full-text queries per month, a change from the previously announced shared pool of 2,000 queries.
"We'll send a reminder email before billing is enabled," the company noted, emphasizing a smooth transition for existing users.
The general availability follows Cloudflare's August 2026 Agents Week, where preview pricing was first introduced. Since its initial launch, developers have used AI Search for internal documentation and website search, including Cloudflare's own blog and developer docs.
Looking ahead, Cloudflare plans to extend the ingestion pipeline to support full video and audio processing, refactor its keyword search engine for better scalability, and simplify index creation for websites already on its platform. These developments aim to make it easier for AI agents to discover and consume content.
Read: AWS Adds AI Agent Skills to MCP Server for SES and End User Messaging
AI Search is available today, and developers can start using the new multimodal embeddings, larger file support, and OCR features immediately.
