If you’ve ever asked ChatGPT a science question and quietly wondered whether it just made the answer up, Consensus is the tool built to fix exactly that. Rather than drawing on the open web or a model’s training data, Consensus (consensus.app) searches a corpus of over 200 million peer-reviewed academic papers, then uses large language models to synthesise what the literature actually says — with a citation behind every claim.

That focus has made it the standout name in academic AI search. Backed by a $30 million Series A, partnered with 170-plus university libraries and used by millions of researchers, it turns the slow grind of a literature review into something you can do in minutes — without sacrificing the one thing that matters most in research: trust.

What Consensus actually does

The core workflow is simple: ask a research question in plain English, and Consensus retrieves the most relevant papers, then writes a cohesive, citation-backed summary of the findings. Its corpus is built primarily from OpenAlex and Semantic Scholar — around 220 million paper records — plus exclusive full-text partnerships with publishers like Taylor & Francis, Sage and the American Chemical Society. The features that make it sing:

• The Consensus Meter — its signature feature. For yes/no questions, it visualises how much the studies agree (yes / no / possibly), so you can see at a glance whether there’s genuine scientific consensus.
• AI synthesis — a clear summary of the current state of research with inline citations, not a copy-paste of abstracts.
• Study Snapshots — quick breakdowns of individual papers (population, methods, outcomes).
• Chat with a paper — ask questions of a single study directly.
• Advanced filters — by date, study type, sample size and journal quality.
• Search depth — Quick (around 10 papers), Pro (around 20) or Deep (50–1,000) for exhaustive reviews.

How Consensus scores

Overall score: 8.3 / 10, the average of the eight categories above.

What Consensus does well

Evidence you can actually trust

This is the whole point, and Consensus nails it. Because it searches only peer-reviewed papers and cites every claim, you get the credibility a general-purpose chatbot simply can’t match. For medical professionals, policymakers, journalists fact-checking a claim, or anyone burned by an AI hallucination, that grounding in real literature is worth the subscription on its own.

The Consensus Meter is genuinely clever

Most tools give you an answer; Consensus tells you how settled that answer is. Ask “does intermittent fasting aid weight loss?” and the Meter shows the balance of yes / no / possibly across the studies it analysed. It’s a brilliantly simple way to surface where the science is solid and where it’s still contested — and nothing else in this space does it as elegantly.

It makes serious research accessible

You don’t need a library-science degree to use it. The interface is clean and beginner-friendly, yet the depth is there when you need it — Study Snapshots, journal-quality filters and Deep mode for exhaustive reviews. In the ongoing “Consensus vs Perplexity” debate, Consensus wins on academic depth and rigour every time, even if Perplexity is the better generalist.

Consensus pricing

Consensus runs a freemium model. It restructured its plans in 2026 after its Series A, so older roundups quoting an $8.99 or $11.99 “Premium” tier are out of date — always check the live pricing page.

Students get roughly 40% off Pro (around $6/month with a .edu or .ac email) — among the cheapest serious academic AI tools going. Practising clinicians can get about 25% off with a verified NPI.

Where it falls short

Narrow by design

Consensus only searches peer-reviewed papers — which is the source of its strength and its biggest limitation. It’s not a general search engine and won’t help with everyday questions, current events or anything outside the academic literature. It’s a specialist instrument, not an all-rounder, so most people will use it alongside a generalist like ChatGPT rather than instead of one.

A summary is not a substitute for reading

For rigorous work, Consensus is a starting point, not the finish line. Its synthesis is excellent, but the Consensus Meter can flatten genuinely complex questions into a yes/no, and coverage is stronger in heavily-published fields (medicine, biology) than in niche areas. It can also lag on very recent or breaking findings. Use it to map the literature fast — then read the key papers yourself before you rely on them.

Pros and cons

Who should use Consensus?

Consensus is close to essential for PhD candidates, academic researchers, clinicians running literature searches, science journalists and policymakers — anyone who needs to know what the evidence says, fast, with citations to back it. Students get the best deal of all, and even the free tier is enough to feel the quality before paying.

It’s the wrong tool if you want a general assistant for everyday questions, the latest news, or open-web research — that’s where Perplexity and You.com fit better. Most serious researchers will happily run Consensus alongside one of those.

Verdict

Consensus is the best AI academic search engine available, and it’s not especially close. It does one thing — evidence-based research from peer-reviewed literature — and does it with a rigour, clarity and trustworthiness that general chatbots can’t touch. The Consensus Meter is a genuinely original idea, the citations are watertight, and the student pricing is a gift.

Its limits are simply the flip side of its focus: it’s narrow, summaries shouldn’t replace reading the originals, and niche fields get thinner coverage. None of that dents what it is — the most trustworthy way to find out what science actually says, and a tool serious researchers will wonder how they managed without. Score: 8.3/10.

Frequently asked questions