scite.ai
Shows whether the papers citing a given study actually supported it or contradicted it
- Paid
- New
scite.ai is a research tool aimed at a problem citation counters ignore: the number of citations a paper has says nothing about their substance. A paper can rack up thousands of citations because researchers kept refuting it — not because it held up. Smart Citations fixes this by classifying each citation in a database of over 1.6 billion as supporting, contrasting, or merely mentioning, and showing the actual passage from the citing paper so you can judge for yourself.
Where Consensus or Elicit search the literature for an answer, scite interrogates the credibility of a paper you already have through everything published after it.
What sets it apart
- Smart Citations: every citation labeled supporting, contrasting, or mentioning — with the citing text right there
- Superseded-paper detection: see when a paper was later contradicted or overturned, no matter how impressive its citation count looks
- Citation Alerts: instant notification when a new paper cites or contradicts research you’re following
- General-assistant integrations: plug it into ChatGPT or Claude so any AI-assisted research comes back with real citations attached
How to use it
- Create an account at
scite.aiand start the 7-day trial - Look up a paper or topic and read its Smart Citation report (supporting / contrasting / mentioning)
- Turn on Citation Alerts for any paper or topic worth tracking over time
- Connect the ChatGPT or Claude integration to bring citations into your general AI research
- Move to Pro if you need API access, larger collections, or patent and clinical-trial data
Who it’s for
A good fit if you’re a researcher or graduate student who needs to vet a paper’s standing before citing it, a journal editor or reviewer checking a manuscript’s sources, or anyone tracking a research thread who wants to know the moment a contradiction appears.
Not a good fit if you want a tool that answers research questions directly (Consensus or Elicit), you can’t justify a dedicated citation-analysis subscription, or your research is primarily in languages other than English.
Limits and warnings
No permanent free plan. Seven days of trial, then a paid subscription from day one.
The classification is automated, not a ruling. “Supporting” and “contrasting” labels come from automated text analysis — for anything consequential, read the citing passages and the original paper in full.
English-language corpus. Coverage of non-English academic publishing is thin.
Alternatives
Consensus is better when you want a quick evidence-backed answer to a specific research question. Elicit is better when the job is a full literature review or structured extraction across dozens of papers.
✅ Pros
- Smart Citations classifies every citation in its 1.6-billion-citation database as supporting, contrasting, or merely mentioning — with the actual text from the citing paper
- Surfaces papers later contradicted or superseded by subsequent research, which a raw citation count never tells you
- Citation Alerts notify you the moment a new paper cites or contradicts research you're tracking
- Direct ChatGPT and Claude integrations back any general-assistant research with real citations
❌ Cons
- No permanent free plan — only a 7-day trial before you commit
- Pro ($50/month) adds API access and larger collections, a steep jump for an individual user
- The corpus is English-language academic publishing, so other-language research is thin
- Its strength is citation classification specifically — not general search or summarization like Elicit or Consensus
💰 Pricing
| Plan | Price | Features |
|---|---|---|
| Basic | $20 / month (annual billing, saves up to $96) | Unlimited assistant use, full-text search, Smart Citation reports, citation alerts, 250 monthly MCP credits, ChatGPT and Claude integrations, collections up to 1,000 papers |
| Pro | $50 / month (annual billing, saves up to $120) | Everything in Basic plus API access, 2,500 monthly MCP credits, collections up to 10,000 papers, and patent, clinical-trial, and grant data |
🔄 Similar Alternatives
📘 Tutorials that use this tool
❓ Frequently Asked Questions
How is it different from Consensus or Elicit?
Consensus and Elicit answer research questions or extract data across many papers. scite does something fundamentally different: it shows how a specific paper was received after publication — did later studies support its findings, contradict them, or just mention it? That's how you catch important papers that were later debunked or superseded, something a raw citation count on Google Scholar will never reveal.
Is there a free plan?
No permanent free tier — just a 7-day trial. After that you need at least the Basic plan to keep using it.
Do I need Pro, or is Basic enough?
Basic covers most individual researchers: unlimited search, Smart Citation reports, and alerts. You only need Pro if you're building something on their API, need collections beyond 1,000 papers, or want patent and clinical-trial data.
Can it tell me if a paper is trustworthy before I cite it?
That's precisely its use case. A paper with thousands of citations may have been cited mostly to contradict it — scite shows that breakdown, which a plain citation counter hides. Treat the classification as a strong signal, then read the contrasting citations yourself.