crawl4ai — Does "LLM-Ready Markdown From Any Webpage" Actually Work?

What This Repo Claims

Fast, LLM-friendly web crawler that converts any page to clean markdown. Supports fit_markdown — a filtered version of the page with noise removed — ideal for RAG pipelines.

50k+ stars. Actively maintained.

Original Test (PARTIAL verdict)

Tested on crawl4ai v0.8.6. Used result.fit_markdown to access filtered markdown output. Returned 0 characters on all pages tested.

Root cause (identified on re-test): result.fit_markdown was deprecated in v0.5 and fully removed in v0.8.6. The correct path is result.markdown.fit_markdown. The deprecation is documented in the official docs but not prominently surfaced in the README or quickstart examples at the time of original testing.

This was not a bug in crawl4ai. It was an API migration.

Re-Test

Environment:

• macOS, Python 3.13, crawl4ai 0.8.6
• Clean repoverifiertest user
• No prior crawl4ai state

import asyncio
from crawl4ai import AsyncWebCrawler, CrawlerRunConfig
from crawl4ai.content_filter_strategy import PruningContentFilter
from crawl4ai.markdown_generation_strategy import DefaultMarkdownGenerator

async def test():
    config = CrawlerRunConfig(
        markdown_generator=DefaultMarkdownGenerator(
            content_filter=PruningContentFilter()
        )
    )
    async with AsyncWebCrawler() as crawler:
        result = await crawler.arun('https://example.com', config=config)
        print('raw_markdown length:', len(result.markdown.raw_markdown or ''))
        print('fit_markdown length:', len(result.markdown.fit_markdown or ''))
        print('fit_markdown preview:', (result.markdown.fit_markdown or '')[:300])

asyncio.run(test())

Output:

raw_markdown length: 166
fit_markdown length: 166
fit_markdown preview: # Example Domain
This domain is for use in documentation examples without needing
permission. Avoid use in operations.
Learn more

What Works

• Install clean — no errors
• Crawler starts and fetches pages
• result.markdown.raw_markdown returns full page markdown
• result.markdown.fit_markdown returns filtered markdown with noise removed
• PruningContentFilter integrates correctly with DefaultMarkdownGenerator
• No API key required

API Migration Note

If you are upgrading from crawl4ai < v0.5, update your code:

# OLD — removed in v0.8.6
result.fit_markdown

# NEW — correct path
result.markdown.fit_markdown

Verdict: SOLID

The core claim holds. crawl4ai converts web pages to clean, LLM-ready markdown. fit_markdown works correctly using the current API. The original PARTIAL verdict was a testing error, not a crawl4ai defect.