About Filtir
AI-assistants like ChatGPT offer tremendous benefits for content authors.
But there's a catch.
They "hallucinate" factual claims that are demonstrably false.
Filtir is a fact-checking ChatGPT Plugin (also available as a web app).
Filtir works by assessing whether there is public, auditable evidence to support the claims identified in written body of text.
When a claim lacks support, we let you know.
Interactive Sample results
Below you can find a sample fact-checking results (click on sentence to view sources for the facts):
Input
Sample output
Born in 1906 in Linden, Germany
(now a district of Hanover)
,
Hannah Arendt was raised in a secular, intellectual Jewish family
.
Her father, a
respected engineer
,
passed away when Arendt was just seven years old
.
Her mother, Martha Arendt
,
remarried in 1914
,
and the family moved to Königsberg
(now Kaliningrad, Russia)
,
where Arendt would come of age
.
Supporting evidence score: 9.5/11
Factchecking summary:
We found 11 relevant facts in the provided text. Out of the 11 facts, we could verify 9 and partially verify 1.Meet Our Team
Our team has many years of experience building advanced machine learning systems in industry and academia.
If you want to fact-check your AI generations, we are here to serve you.
Samuel Albanie
Vlad Bogolin
Anthony Buonomo
Ioana Croitoru
Frequently Asked Questions
How do you determine whether a factual claim is true?
We do not arbitrate truth. Instead, we provide evidence in support of factual claims made by a piece of text. Where we are unable to identify appropriate evidence, we flag the claim as "unsupported". This does not mean that it is not true, it simply means that we cannot readily find public evidence in support of the claim.
Which sources of evidence do you consider acceptable? How do you know if a source is "trustworthy"?
Our objective is to provide supportive evidence where we can find it. At present, we do not discard evidence based on reputation metrics - instead, we trust the user to make an assessment. However, we will be working closely with our users and may revise this strategy in future if we learn that a different approach better serves user needs. If you have particular requests relating to this topic, do not hesitate to reach out to us at support@filtir.com
We do not arbitrate truth. Instead, we provide evidence in support of factual claims made by a piece of text. Where we are unable to identify appropriate evidence, we flag the claim as "unsupported". This does not mean that it is not true, it simply means that we cannot readily find public evidence in support of the claim.
Which sources of evidence do you consider acceptable? How do you know if a source is "trustworthy"?
Our objective is to provide supportive evidence where we can find it. At present, we do not discard evidence based on reputation metrics - instead, we trust the user to make an assessment. However, we will be working closely with our users and may revise this strategy in future if we learn that a different approach better serves user needs. If you have particular requests relating to this topic, do not hesitate to reach out to us at support@filtir.com
Does filtir sell my data?
No. We do not sell your data or share your content with third parties for marketing purposes.
See our privacy policy and api data policy for more details about how your data is used.