crawl4ai version
0.9.4 (develop @ 86e6464); also reproduced on 0.9.0
Expected Behavior
cleaned_html keeps the page's body content. <noscript> fallback blocks are removed, as they are today.
Current Behavior
On pages that contain nested <noscript>, cleaned_html loses part or all of the body, even though result.html contains the full page.
Lazy-load plugins commonly wrap an existing noscript (typically Google Tag Manager's), so the served page contains:
<noscript><iframe ...></iframe><noscript><iframe src="https://www.googletagmanager.com/ns.html?id=..."></iframe></noscript></noscript>
The browser has scripting enabled, so it treats noscript contents as raw text and ends the element at the first </noscript>. The stray second closing tag is dropped when page.content() serializes the DOM, so the capture has more opening tags than closing ones. lxml parses with scripting disabled, so it nests the rest of the document inside the still-open <noscript>, and LXMLWebScrapingStrategy._scrap then deletes that content when it removes noscript elements.
One common source is LiteSpeed Cache's iframe lazy-load (~7M active WordPress installs), which rewrites GTM's noscript iframe this way: WordPress.org support thread.
Visible text in cleaned_html for affected pages, default configs (2026-09-24):
In a random sample of 2,972 Tranco top-1M homepages (ranks 20k–1M), 6 served nested noscript, and 4 of those lost content this way.
Is this reproducible?
Yes
Inputs Causing the Bug
- URL(s): https://www.comunio.es/ , https://halalvlees.nl/
- Settings used: defaults (BrowserConfig(), CrawlerRunConfig())
- Input data: any HTML with more <noscript> opening tags than closing tags
Steps to Reproduce
1. Crawl https://www.comunio.es/ with default settings
2. Compare result.cleaned_html with result.html
3. cleaned_html has an empty body; result.html has the full page
Code snippets
# Offline repro: the shape page.content() produces for such a page
from crawl4ai.content_scraping_strategy import LXMLWebScrapingStrategy
html = """<html><head><title>Example shop</title></head><body>
<noscript><iframe src="about:blank"></iframe><noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-TEST"></iframe></noscript>
<div><h1>Real heading</h1><p>Real body copy.</p></div>
</body></html>"""
print(LXMLWebScrapingStrategy().scrap("https://example.com/", html).cleaned_html)
# Actual: <html><head><title>Example shop</title></head></html>
# Expected: the <h1> and <p> are kept# Live repro
import asyncio
from crawl4ai import AsyncWebCrawler
async def main():
async with AsyncWebCrawler() as crawler:
result = await crawler.arun("https://www.comunio.es/")
print(len(result.html), repr(result.cleaned_html[-200:]))
asyncio.run(main())OS
macOS 26.6 (arm64)
Python version
3.12.11
Browser
Chromium (Playwright headless shell)
Browser version
153.0.8010.12
Error logs & Screenshots (if applicable)
No error is raised; result.success is True and the status code is 200.
A fix is proposed in #2284.
crawl4ai version
0.9.4 (
develop@ 86e6464); also reproduced on 0.9.0Expected Behavior
cleaned_htmlkeeps the page's body content.<noscript>fallback blocks are removed, as they are today.Current Behavior
On pages that contain nested
<noscript>,cleaned_htmlloses part or all of the body, even thoughresult.htmlcontains the full page.Lazy-load plugins commonly wrap an existing noscript (typically Google Tag Manager's), so the served page contains:
The browser has scripting enabled, so it treats noscript contents as raw text and ends the element at the first
</noscript>. The stray second closing tag is dropped whenpage.content()serializes the DOM, so the capture has more opening tags than closing ones.lxmlparses with scripting disabled, so it nests the rest of the document inside the still-open<noscript>, andLXMLWebScrapingStrategy._scrapthen deletes that content when it removesnoscriptelements.One common source is LiteSpeed Cache's iframe lazy-load (~7M active WordPress installs), which rewrites GTM's noscript iframe this way: WordPress.org support thread.
Visible text in
cleaned_htmlfor affected pages, default configs (2026-09-24):cleaned_htmlresult.htmlIn a random sample of 2,972 Tranco top-1M homepages (ranks 20k–1M), 6 served nested noscript, and 4 of those lost content this way.
Is this reproducible?
Yes
Inputs Causing the Bug
Steps to Reproduce
1. Crawl https://www.comunio.es/ with default settings 2. Compare result.cleaned_html with result.html 3. cleaned_html has an empty body; result.html has the full pageCode snippets
OS
macOS 26.6 (arm64)
Python version
3.12.11
Browser
Chromium (Playwright headless shell)
Browser version
153.0.8010.12
Error logs & Screenshots (if applicable)
No error is raised;
result.successisTrueand the status code is 200.A fix is proposed in #2284.