scite

Scite.ai is an AI-powered research platform designed to help scholars, students and institutions discover and evaluate scientific literature using “smart citations”. It doesn’t just count citations — it classifies them according to whether a cited work is supported, contradicted, or simply mentioned. Scite+2Wikipedia+2
The platform was founded by Josh Nicholson and Anand Desai in 2018. 

• Traditional citation counts treat all citations the same (“X was cited”), but scite.ai adds nuance by showing how a work is cited — e.g., if others replicate, support, or challenge the finding. This adds depth to bibliometric analysis. For research evaluation, it allows institutions and individuals to identify articles with strong supportive evidence or those that are controversial / contradicted — improving insight into research quality and impact. It aligns with increasing emphasis on transparency, research integrity, and reproducibility — especially as the scientific community grapples with issues like retractions, corrections, and misuse of findings. 
• For literature reviews or meta-analysis, instead of just “how many times has this been cited”, you can ask “how many times has this been supported or contradicted?” which helps in assessing robustness.

Key features & advantages

• Smart citations: Each citation is classified into categories: “supports”, “contradicts”, or “mentions”. 
• Search and discovery: scite.ai allows searching for articles, exploring citation context (snippets of text where the citation occurs), seeing trends of citation types over time. 
• Browser extension: There’s a browser plugin that overlays smart-citation info when you view a paper, helping you at the point of reading. 
• Data & services: For institutions, there are APIs and data-feeds so bibliometric teams can integrate smart citation metrics into dashboards or reporting. 
• Free and paid tiers: The platform offers free access (limited) and paid plans for enhanced features or higher volume usage. 

Limitations / things to watch

• Coverage: Although the classification is sophisticated, not every citation is yet classified, and contexts vary by domain — so one should check how much of a paper’s citations have been analysed.
• Interpretation: A citation classified as “contradicts” doesn’t always mean the original finding is wrong — it might mean the later paper challenges it in some way, users need to inspect the context.
• Cost / Access: For heavy usage (institutional, large-scale bibliometric analysis) one may need a paid plan.
• Domain variation: The quality and completeness of classification may vary across disciplines and less-represented journals, newer/less accessible papers may have less metadata.

Why your lab/institution might use it

Given your lab-context (and you preferring structured documentation, open science, reproducibility) here are ways scite.ai could be relevant:

• When writing a literature review: You can use scite.ai to identify key papers that have substantial supportive evidence vs those that have been challenged — helps in framing the strength of your argument.
• When selecting references: Instead of blindly citing highly-cited works, you can check whether those citations are positive (supports) or negative/contradicting — this can improve your scholarship.
• When tracking your own outputs: After you publish, you can use scite.ai to monitor how your papers are cited (supportively or otherwise) and include that in reports or pitches.
• In preparing a pitch deck: If you are seeking funding or institutional support, showing citation-quality (not just quantity) of previous work may strengthen your narrative of impact / validity.
• For reproducibility / integrity: If some key background studies of your project have been contradicted, scite.ai can help you spot that risk early and adjust your assumptions / design.

Quick user-journey example

• You’re preparing your next manuscript and have compiled ~20 key references.
• Use scite.ai to look up each reference and see: how many citations it has, how many “supports” vs “contradicts” vs “mentions”.
• You find one highly-cited paper but many of its citations are “contradicts” — you flag it and investigate whether your method should account for that.
• You also check your own prior work: it has 5 “supports”, 1 “contradict” citation — you include that metric in your institutional impact report.
• Going forward, you install the browser extension so when you browse new pre-prints or articles you can immediately see smart-citation metrics and decide whether to dig into them.