Consensus

Consensus answers a simple, common need: you have a factual question and want to know what peer-reviewed research says, without wading through abstracts or trusting a chatbot's memory. Ask "does intermittent fasting improve insulin sensitivity?" and it returns a synthesized answer, the Consensus Meter showing how studies split, and snapshot cards for each paper.

What we like

The product knows exactly what it is. Every answer is built from retrieved papers, each claim links to its source, and the study snapshots — population, sample size, methodology — let you judge evidence quality without opening a single PDF. The Consensus Meter is the standout: seeing that 14 studies say yes, 3 say possibly, and 2 say no is a dramatically more honest answer format than a confident paragraph from a general-purpose chatbot. Quality filters for RCTs, systematic reviews, and human trials are exactly what health and nutrition questions need. At roughly $9/month for Premium, it is also one of the cheaper tools in this category.

Where it falls short

Consensus is built for questions with a yes/no shape, and it degrades on open-ended or exploratory queries — "what are emerging approaches to X" produces mediocre results. The Meter, useful as it is, can flatten genuinely contested literatures into false tidiness; vote-counting studies is not the same as weighing them. And serious researchers will outgrow it quickly: there is no extraction, no screening workflow, no citation management.

Verdict

Consensus is the best entry point into evidence-based AI search for students, clinicians checking claims, and writers who want to cite real studies. It is cheap, fast, and honest about uncertainty in a way most AI tools are not. Researchers running actual literature reviews should treat it as a complement to Elicit rather than a replacement — but for the question-in, evidence-out use case, it is the tool to beat.