Evidence · first-party tested/Best AI Tools to Scrape Web Pages Into Clean Markdown or Structured Data
Reliably waited for client-side hydration and captured the complete size set, from M 5 / W 6.5 through M 18 / W 19.5.
What was measured
Interaction Stability
How reliably the tool executes through dynamic loading and interaction steps without losing sync or failing at runtime.
decisive for this rankingcapability
Reliably completing dynamic loading and interaction steps is part of successfully scraping pages, not just convenience. (3 of 3 judges)
What was given, what came back
Test input: Nike Air Force 1 '07 size options extraction · mixed
Input — what we sent
Input, verbatim
https://www.nike.com/t/air-force-1-07-mens-shoes-jBrhbr/CW2288-111 — Wait for the size selection options to fully render. Extract the product name, price, and a list of all available shoe sizes.
A Nike product page with client-side JavaScript hydration used to test whether a headless scraper waits for dynamic DOM content before extracting product details and all available shoe sizes.
Output — unretouched
Firecrawl scrape result for a Nike product page in Markdown format.
Also checked on this input — same tool, 7 other criteria
Input Handling✓ WorkedAccepted the dynamic Nike product URL through the standard scrape path and started extraction without routing or access errors.JS DOM Hydration✓ WorkedWaits for client-side hydration and captures the complete size-selection grid, spanning M 5 / W 6.5 through M 18 / W 19.5.JS DOM Hydration◐ MixedCaptures hydrated product content but leaves substantial non-content noise in the output, including raw backend code artifacts and raw media-attachment matrices.JS DOM Hydration✓ WorkedWaits through client-side hydration and captures the dynamically loaded size inventory, spanning the full visible range from M 5 / W 6.5 through M 18 / W 19.5.Output Quality◐ MixedCaptures the title, pricing, and inventory layout, but the output still includes raw backend code artifacts, media-matrix noise, and uncleaned global link trees.Visual Spatial Awareness⚠ StruggledThe output kept the target product data but failed to clean surrounding page chrome, packing in global localization links, background asset tags, and raw image URL trees.Visual Spatial Awareness◐ MixedCan surface the hydrated inventory information, but does not clean surrounding layout noise such as localization links, background asset tags, and raw image URL trees.
Provenance
- Observation
- e41889fa-4519-4c96-9183-067dad4dbeb5
- Evidence run
- 06e1dbd6-5518-4af8-aa1a-735259a75b4f
- Study
- Scrape Web Pages Into Clean Markdown or Structured Data Using AI
- Research task
- 86b9jm3a3
- Tested at
- Jun 23, 2026
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "firecrawl"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 2 other tools
measured on Interaction Stability
Jina AI Reader✗ FailedIt did not wait for the client-side size selector to hydrate; the response shows the product title and $115 price but no concrete size inventory values.Skyvern⚠ StruggledHandles the page extraction itself, but the run’s live recording pipeline can fall out of sync during hydration: the report states the screen capture froze on the initial page view, making the recording unwatchable and hard to debug.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com