Consensus Everywhere: Wherever you work, research is within reach

Ask ChatGPT whether intermittent fasting affects LDL and you’ll get a clear, organized answer in seconds. It may sound authoritative. It may even be right.

But where’s the paper?

Finding out often means leaving the conversation and running the search again somewhere else. Many people skip that step, not because they don’t care about the evidence, but because they’re busy.

We built Consensus as a destination: a search engine over 220M+ peer-reviewed papers and preprints, with models that rank and read them. That made sense when research started in a browser tab. Increasingly, it starts in an AI assistant. So we’ve brought Consensus there, too.

Research should happen where the question does

A question rarely arrives at a convenient moment. It turns up three paragraphs into a draft or in the middle of an unrelated conversation. Going elsewhere to answer it breaks your flow, and busy people eventually stop making the trip.

You can now add Consensus to ChatGPT or Claude directly from either app’s directory. Ask what the research says, and Consensus returns papers you can open, filter by study type or journal quality, and cite without leaving the conversation. It’s free to start. You just need a Consensus account.

Your evidence standard shouldn’t depend on your AI

Researchers don’t always get to choose their AI tools. If a university provides Copilot, that’s often where faculty and staff work, whether or not it’s their first choice. Until now, checking its answers against the literature meant leaving that approved workflow and starting over somewhere else.

And now, Consensus for Copilot is now available from the Microsoft Marketplace. It runs in Copilot Chat, Cowork, Word, Excel, and PowerPoint, so the papers land in the manuscript you’re drafting instead of a tab you mean to get back to.

A model's memory is not the literature

We tested whether Consensus improves on what an assistant returns on its own. Four PhD-level researchers evaluated the papers returned for 100 anonymized user queries on a 0–4 relevance scale, without knowing which system produced them.

Across the top 20 results, Consensus achieved 4.5% higher precision than GPT-5.5 and 13.2% higher precision than Opus 4.8. It returned the strongest overall result set for 59 of the 100 queries.

The assistant will keep changing

The assistant you use today may not be the one you use next year. You may switch from ChatGPT to Claude, or your institution may choose Copilot for you. Either way, the foundation of your research should remain the same: relevant papers you can inspect for yourself.

Reading the papers is still your job. Finding them shouldn't need a different tab.

Connect Consensus to the assistant you already use →

At an institution? Send the setup guide to whoever manages your AI tools, and they can switch it on for everyone.