Where does Claude report when fetched content was retrieved?

Definition: Claude reports the retrieval time in the retrieved_at field of a web_fetch_tool_result response block. The field sits inside the web_fetch_result content object, beside the fetched URL and document content.

Key facts

• Anthropic describes web fetch as a server tool: the API retrieves the page during the request and inserts the result into the conversation. The web fetch documentation says the application does not run the tool or return a tool_result for the ordinary case.
• A successful result uses the web_fetch_tool_result block type and contains a web_fetch_result object with the fetched URL, a document block, and retrieved_at.
• Anthropic's Messages API schema describes retrieved_at as an optional string or null and specifies an ISO 8601 timestamp when the content was retrieved.
• The documentation's example value is 2025-08-25T10:30:00Z, which records a UTC time with a date and time component.
• Anthropic's tool-use overview places web_fetch among server tools that run on Anthropic's infrastructure and return results directly.

Where is the field in the response?

The field is nested in the result returned by the server tool. The web fetch documentation shows the server tool call first, followed by a result block whose content object has type web_fetch_result. Inside that object, url, the document content, and retrieved_at are peer fields.

{
  "type": "web_fetch_tool_result",
  "tool_use_id": "srvtoolu_01234567890abcdef",
  "content": {
    "type": "web_fetch_result",
    "url": "https://example.com/article",
    "content": {
      "type": "document"
    },
    "retrieved_at": "2025-08-25T10:30:00Z"
  }
}

The tool_use_id connects the result to the server tool invocation. The API schema includes a tool_use_id string for this result type.

What does retrieved_at describe?

It describes when the returned content was retrieved. It is a retrieval timestamp, so it should be stored separately from page metadata such as a publication date. The Messages API schema allows the value to be absent as null, which means an integration should handle a missing timestamp while parsing the result.

The result block carries the timestamp with the fetched document before Claude's response text and any optional passage citations.

How does caching affect the timestamp?

Anthropic documents automatic caching for web fetch results and warns that returned content may differ from the latest version at the URL. The fetch documentation says that use_cache: false requests a fresh fetch. The documented meaning of retrieved_at remains the time when the returned content was retrieved.

The same page says web fetch citations are optional and disabled by default. Enabling citations controls passage citations; it is separate from the retrieved_at field.

Related terms

• Claude web search
• Claude-SearchBot
• Claude-User
• Retrieval-augmented generation (RAG)

FAQ

Is retrieved_at in the request or the response?

It is in the response, inside the web_fetch_tool_result block. Anthropic's schema defines it on the successful web_fetch_result content object.

Is retrieved_at a publication date?

No. The API documentation defines it as the ISO 8601 time when the content was retrieved. The result schema provides the fetched URL separately from this field.

Does enabling citations create retrieved_at?

Citations and retrieval time are separate response details. The web fetch documentation describes citations as optional and disabled by default, while the result schema defines retrieved_at on the fetch result.

How can an integration request a fresh fetch?

The web fetch documentation says to set use_cache to false when fresh content is required. The returned retrieved_at value can then be recorded with the result.

Sources

1. Anthropic, "Web fetch tool", null
2. Anthropic, "Messages API", null
3. Anthropic, "Tool use with Claude", null

---

Check how a page reads to answer engines. Open the GEO score analyzer