Can you please tell me where can I find good pre-labeled datasets with real and fake news, but not tweets and facebook posts?

I found 4-5 on Kaggle, but I read that those are not real news (those are not real articles). I have also looked UCI ML Repository but those datasets are similar.

Basically I need dataset that has 3 columns, "title/headline", "text / paragraph text" and "label (true of fake)". Label does not need to be boolean, it can be more than 2 values, but dataset needs to be labeled.

Also, I do not want twitter and facebook posts, im interested in news articles.

Can you tell me other websites where I can find datasets like this? Can you share some links where I can download such a thing?

  • Is the truth of newsmedia really a Boolean, and if so, who is defining it?
    – philshem
    Apr 3, 2020 at 19:46
  • Funny @philshem ... there is truth in that!
    – Marcus D
    Apr 4, 2020 at 15:35
  • @philshem I agree, but now, I just need a datasets. There is no need to be just boolean (FAKE or TRUE), it can be more than 2 labels, but I need dataset that has headline, paragraph text and labels
    – taga
    Apr 4, 2020 at 16:47
  • duplicate?? opendata.stackexchange.com/q/10584/1511
    – philshem
    Apr 6, 2020 at 11:03
  • Not duplicate, that answer was more than 3 years, its outdated
    – taga
    Apr 6, 2020 at 12:56

1 Answer 1


From the Google Site.

Fact Check Tools API Getting Started The Google FactCheck ClaimReview Read/Write API Typical use cases

Adding ClaimReview markup, editing ClaimReview Markup, deleting ClaimReview Markup for a site’s fact checking articles (URLs). To learn more about ClaimReview, visit the Fact Check developer docs.


Checking "Facts" are dependent upon the developers and/or readers perspective. Each may be completely different. I wish you luck but I wouldn't touch this with a 5000 foot pole.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.