Citation

Cite the dated page, not a screenshot.

CrawlDex claims should point to canonical URLs that carry dates, observation counts, confidence labels, evidence fields, and correction paths.

Canonical formatEvidence URLs firstNo screenshot citations

Preferred line

Use this line under charts, excerpts, benchmark rows, and report claims:

Source: CrawlDex, {n} observations, {month year} - {canonical URL}

Do not cite private runs, screenshots, local logs, raw traces, `.crawldex` files, or pages that require a login to verify.

Which URL

Use the most specific CrawlDex URL that supports the claim.

Specific public evidence is stronger than a homepage citation. Choose the first matching level in this order.

1. Specific Agent Trust Record

/r/<record_id>

Use this when the claim depends on a single issued record and its dated evidence fields.

2. Site-task evidence page

/sites/<site>/<task>

Use this when citing a public task verdict, blocker, freshness state, or site-level operational claim.

3. Report or benchmark page

/reports/... or /bench/...

Use this for aggregate claims, ranked lists, monthly index claims, or benchmark results.

4. Dataset page

/datasets

Use this when citing the public outcome graph, schema, collection method, or corpus scope.

BibTeX

Citation blocks

Dataset details

Project

@misc{crawldex,
  title = {CrawlDex: public reliability records for AI agents using websites},
  author = {{CrawlDex Project}},
  year = {2026},
  url = {https://crawldex.com/cite},
  note = {Public evidence-backed records, datasets, and reports for agentic web tasks}
}

Dataset

@dataset{crawldex-outcome-graph,
  title = {CrawlDex outcome graph},
  author = {{CrawlDex Project}},
  year = {2026},
  url = {https://crawldex.com/datasets},
  license = {CC BY 4.0},
  note = {Public site-task observations with blockers, freshness, confidence, and AES fields}
}

Bench

@misc{crawldex-bench-2026-07,
  title = {CrawlDex Bench 2026.07},
  author = {{CrawlDex Project}},
  year = {2026},
  url = {https://crawldex.com/bench/bench-2026.07},
  note = {Read-only browser-agent benchmark with server re-scored public submissions}
}

Site-task page

@misc{crawldex-site-task,
  title = {CrawlDex site-task evidence page},
  author = {{CrawlDex Project}},
  year = {2026},
  url = {https://crawldex.com/sites/<site>/<task>},
  note = {Use the page date, observation count, confidence label, and blocker fields from the cited URL}
}

Outcome graph dataset

Public site-task rows, blocker fields, freshness, confidence, and AES metadata.

CrawlDex Bench 2026.07

Read-only benchmark version and verified browser-agent stack results.

Public reports

Reviewed and gated report pages for aggregate reliability claims and ranked evidence.

Methodology

Evidence rules, confidence constants, freshness handling, and correction path.

Name usage

Citation is not endorsement.

Use the plain name CrawlDex when attributing records, datasets, or benchmark rows. Do not imply sponsorship, partnership, pay-for-score influence, removal rights, or endorsement by CrawlDex unless a separate written agreement says so. Scores, rankings, confidence, and dispute handling are evidence-derived only.