Datasets:

xlth
/

CRED-1

license: cc-by-4.0
language:
  - en
tags:
  - arxiv:2604.20856
  - credibility
  - misinformation
  - disinformation
  - fact-checking
  - news-credibility
  - domain-credibility
  - media-bias
  - information-integrity
  - ai-safety
  - content-moderation
task_categories:
  - tabular-classification
  - text-classification
size_categories:
  - 1K<n<10K
pretty_name: 'CRED-1: Open Multi-Signal Domain Credibility Dataset'
configs:
  - config_name: default
    data_files:
      - split: train
        path: cred1_current.csv

CRED-1: Open Multi-Signal Domain Credibility Dataset

CRED-1 is an open, reproducible domain-level credibility dataset combining multiple openly-licensed source lists with computed enrichment signals. It provides credibility scores for 2,672 domains known to publish mis/disinformation, conspiracy theories, or other unreliable content.

Key Features

2,672 domains with credibility scores (0.0 to 1.0)
Fully reproducible Python pipeline rebuilds the dataset from scratch (see GitHub repo)
Multi-signal scoring combining source labels, domain age, web popularity, fact-check frequency, and threat intelligence
Privacy-preserving designed for on-device client-side deployment (no server calls needed)
Two openly-licensed sources, no proprietary data dependencies

Quick Start

Load with `datasets`

from datasets import load_dataset

ds = load_dataset("xlth/CRED-1")
print(ds["train"][0])
# {'domain': '100percentfedup.com', 'category': 'mixed', 'credibility_score': 0.173, ...}

Lookup a domain

from datasets import load_dataset

ds = load_dataset("xlth/CRED-1", split="train")
lookup = {row["domain"]: row for row in ds}

domain = "infowars.com"
if domain in lookup:
    print(f"{domain}: credibility = {lookup[domain]['credibility_score']}")
else:
    print(f"{domain}: not in dataset (treat as neutral)")

Dataset Schema

Field	Type	Description
`domain`	string	Domain name (lowercase, no scheme)
`category`	string	Category from source lists (e.g. `fake`, `conspiracy`, `unreliable`, `mixed`, `clickbait`)
`credibility_score`	float	Aggregated credibility score (0.0 = least credible, 1.0 = most credible)
`sources`	int	Number of source lists the domain appears in
`iffy_factual`	string	Factual reporting rating from Iffy.news (VL, L, M, H, VH)
`iffy_bias`	string	Bias rating from Iffy.news (e.g. FN = fake news, CP = conspiracy)
`iffy_score`	float	Normalized Iffy score
`tranco_rank`	int	Tranco web popularity rank (1 = most popular)
`domain_age_years`	float	Age of the domain in years
`domain_registered`	string	ISO 8601 registration date
`factcheck_claims`	int	Count of fact-check claims targeting this domain
`safe_browsing_flagged`	bool	Whether Google Safe Browsing flagged this domain
`score_*`	float	Individual signal contributions to the aggregate score

Full schema in CODEBOOK.md.

Files

cred1_current.csv (~250 KB): Full dataset, recommended for load_dataset and the dataset viewer
cred1_current.json (~1 MB): Same data as nested JSON with domain as key
cred1_compact.json (~170 KB): Minimal {domain: score} mapping for lightweight on-device lookups
CODEBOOK.md: Field definitions, scoring methodology, source provenance
figures/: Project banner

Intended Use

CRED-1 is designed for:

Browser extensions and on-device clients that need fast, offline domain credibility lookups
Research on misinformation detection, news verification, and platform governance
Pre-bunking pipelines that flag suspicious sources before users engage
Educational and digital-literacy tooling

Limitations

Coverage is concentrated on English-language misinformation sources
The dataset captures domains, not individual articles; legitimate journalism on otherwise unreliable domains is not distinguished
Source lists carry their own biases; consult CODEBOOK.md for provenance

Citation

Canonical citation (Zenodo DOI for this dataset):

@dataset{loth_cred1_2026,
  author       = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},
  title        = {{CRED-1}: An Open Multi-Signal Domain Credibility Dataset},
  year         = 2026,
  publisher    = {Zenodo},
  version      = {v1.0},
  doi          = {10.5281/zenodo.18769460},
  url          = {https://doi.org/10.5281/zenodo.18769460}
}

Accompanying preprint:

@article{loth_cred1_preprint_2026,
  author  = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},
  title   = {{CRED-1}: An Open Multi-Signal Domain Credibility Dataset for Automated Pre-Bunking of Online Misinformation},
  year    = 2026,
  journal = {SSRN Preprint},
  doi     = {10.2139/ssrn.6448466},
  url     = {https://ssrn.com/abstract=6448466}
}

License

CC BY 4.0. Free to use with attribution. See LICENSE.