Bias Analyzer

Analyze the political bias of some text. Bias is given the following ratings:

Direction of Bias:


Degree of Bias


Read more about how the machine learning model was built.

Analysis may take 5-15 seconds and is more accurate on longer content (100+ words).

Press? Contact us

Email for updates to our A.I. future related projects. We don't spam.


(3/17/2020): This tool now runs on our most powerful model, based on Roberta-Large

(2/14/2020): We now offer a Reddit bot to analyze links you find on popular political subreddits.

(12/22/2019): We have updated the tool to run on our second most accurate model, the full BERT regressor.

Thoughts, comments, or feedback? Leave them below


  • comment-avatar
    • comment-avatar

      That is a good question. If this tool is to help show bias I would like to know how that paragraph is biased right as well. I was about to share this tool for others to use, but I am unsure about sharing something that does not seem to work properly.

    • comment-avatar

      Allow me to clarify. As we documented in our article on building the AI, it has an average error deviation of +/- 0.07. That means there’s gonna be some margins of error. It’s not perfect, but generally very accurate.

    • comment-avatar
      Cody Stough February 6, 2022

      Another explanation could be that your text isn’t inherently politically biased, especially if it doesn’t discuss politics. Looking at other ML model, for example, ones that classify images as either cats or dogs, if you feed an image that is neither of those, the model still has to assign either the label of cat or dog to it, even though its not either of those. This model seems to work the same way.

  • comment-avatar

    How are articles written about the death of the photographer Gerth, at a protest during his involvement in BLM not biased? When it is the articles written by his parents, one of which being the video director of the publishers (USAtoday, Gannett and Courier Journal) and calling the shots for the publication.

    I mean most articles, even the ones that the household works for, do not even mention that the household (Father, mother, son) were directly involved in the stories from the beginning. The whole house is still trying to omit they are extremely biased in a narrative they have helped create while pretending to be reporting without bias.

    Your software claims the article is biased right, which is wrong on so many levels. The entire household has been reporting on events and stories that they refuse to admit that they are directly involved in. Furthermore, they are merely using the death of their own son to fuel the narrative their created and controlled from the beginning from the inside. Now so is your software.

    This is immoral and shameful!

  • comment-avatar

    Can this system work with spanish articles?

  • comment-avatar

    How is the bias of chapter 3 of Das Kapitas “minimal right”?