An AI fact checker that reads the original papers
By Steven Van ·
Bearly checks research claims in an attached draft against the papers themselves with Scholar, available on every plan across web, desktop and mobile.
Attaching a draft to a Bearly chat and asking for a fact check now returns a claim-by-claim table. Bearly's Scholar documentation describes the paper search behind it. Bearly pulls out each factual claim, checks research claims against the papers themselves with Scholar, and checks everything else against the web. The table lists the claim, a verdict, what the source actually says, and a link to it. The user then decides what to change.
A test with planted errors
The Bearly team wrote a five-paragraph post on the four-day week with a dozen checkable claims, two of them deliberately wrong. They attached it and asked GPT 6 Sol to give a verdict (supported, wrong, or unclear) for every claim, checking research claims against the papers rather than news coverage. Bearly ran four paper searches, read passages from one paper, and searched the web for the rest. About five minutes later, the reply opened with a verdict on the whole draft, followed by a 12-row table.
- Remote work: the draft said more than half of American workers are fully remote. Bearly marked it wrong. In August 2026, 10.6% of people at work teleworked all their hours and 21.6% teleworked some or all of them, according to the Bureau of Labor Statistics.
- Open offices: the draft said a Harvard study found open offices increased face-to-face interaction by 70%. The study followed two companies that moved to open-plan headquarters and found face-to-face interaction fell by about 70%, while email and messaging rose.
- An unplanted catch: the draft said Microsoft Japan's productivity rose 40% during its 2019 four-day week. The figure is real, since sales per employee in August 2019 were 39.9% higher than a year earlier. Microsoft later amended its announcement to say the trial wasn't the only cause, so Bearly marked the claim unclear and suggested a causal caveat.
Supported claims came back more precise than they were written. Examples were the 13% performance gain among call-center volunteers in a 2015 study of a Chinese travel company, and the UK's 2022 pilot, where 56 of 61 companies were continuing the four-day week when the trial ended and 18 had made it permanent.
How Scholar works
Scholar adds a paper index beside Web Search. When a question should rest on research literature, Bearly searches papers, reads the passages that carry a claim, and cites them with superscript numbers. A Papers card under the reply lists what it found, and marks papers it read along with the passage it relied on.
Coverage comes from arXiv, bioRxiv, medRxiv, and PubMed. Journal articles outside those sources can appear through academic web search, without passages. Full-text passages are available for many arXiv papers, while papers indexed only through PubMed give Bearly the abstract, so exact figures may need the paper opened. Bearly says so when a finding rests only on a preprint.
Availability and controls
Scholar works in Bearly on the web, desktop, iOS, and Android, on every plan, and is on by default. A team policy can turn it off. Under Search in the sliders control beside the chat input, it can be set to Auto or Off for the current chat, and typing /scholar uses it for a single message while it is off.
