[BUG] WebFetch summarizer model drops critical details from static HTML pages

Resolved 💬 3 comments Opened Apr 19, 2026 by catsoc Closed Jun 16, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

WebFetch successfully fetches static HTML content but the intermediate summarizer model loses critical details from the page body, returning only headline-level information. The main model then makes incorrect conclusions based on incomplete data.

Concrete example: Fetched a WordPress blog post about restaurant reservation rules. The page body clearly lists three different stores with different advance booking windows (35 / 60 / 61 days) and a time-slot-by-time-slot release mechanism. WebFetch was called twice with different prompts — both times the summarizer only returned the headline description "book 2 months ahead", dropping all per-store breakdowns and the critical time-slot release rules.

URL: https://plusweiplus.com/nikuichi-reserve/

What Should Happen?

WebFetch should return the key details from the page body, not just headline/summary-level text. At minimum, it should flag when the summarizer has dropped significant portions of the source content so the main model knows the result may be incomplete, rather than presenting partial info as complete.

Error Messages/Logs

No error. The tool returned successfully — the problem is that the returned content was silently incomplete.

Steps to Reproduce

  1. Use WebFetch to fetch: https://plusweiplus.com/nikuichi-reserve/
  2. Prompt: "Extract reservation booking rules — how many days in advance, when do slots open?"
  3. Observe that the response only mentions "2 months in advance" from the headline
  4. The page body actually contains per-store breakdowns (35/60/61 days) and time-slot release rules — all missing from the response

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.63

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

VS Code integrated terminal

Additional Information

The target page is a standard WordPress blog with static server-rendered HTML — no JavaScript-dependent content loading. The issue is not about fetching (the fetch succeeds) but about the summarizer model aggressively truncating body content and only preserving headline-level information.

This caused the main model to write incorrect reservation dates into the user's travel itinerary, which the user had to manually correct by pasting the page content directly.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗