Delv
Researchby Scite★ 4.3

Scite

Citation-aware research agent — shows whether a paper has been supported, contrasted, or merely mentioned. Surfaces hidden disagreements.

Try it Also in Tools: AI Research →
B
Safety & Trust

Delv Safety Grade: B

Score 72/100 · assessed 2026-04-18

Maintainer75
Permissions85
Supply chain55
Transparency50
Incidents100

Scite is a commercial research agent from a legitimate academic technology company that has operated since 2018. The service provides citation analysis through a web interface, requiring only network access to their API. Maintainer legitimacy is solid: Scite is a funded startup with academic partnerships and transparent leadership. Permissions are well-scoped to read-only research queries. However, supply chain visibility is limited because this is a closed proprietary service with no open source components or package distribution to audit. Transparency is moderate: the company is public about its methods and partnerships, but the underlying models and data processing remain opaque. No security incidents are known. The main risk is vendor lock-in and data handling practices for research queries, which may contain sensitive academic work. Suitable for general research but consider data sensitivity before submitting unpublished work.

Green flags

  • Established company (2018) with academic institutional partnerships
  • Read-only research queries with no write permissions required
  • No filesystem, shell, or desktop access needed
  • Transparent leadership and public company information
  • No known security incidents or data breaches

Red flags

  • Closed source proprietary system with no code audit possible
  • No public API documentation or integration transparency
  • Research queries sent to external service may expose unpublished work
  • Paid service creates financial dependency for continued access
  • No visibility into citation extraction algorithms or accuracy guarantees

Permissions requested

Outbound networkExternal LLM call
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

PAID

Platforms

web

Review

Scite does one thing that standard literature search tools don't: it tells you whether subsequent papers actually agreed with the claims you're reading. Instead of just counting citations, it classifies them as supporting, contrasting, or mentioning. That matters when you're building an argument on someone else's result and need to know if it's been quietly debunked. I've used it to audit systematic reviews where the autonomy shows up in how it crawls citation networks without you having to manually chase each reference. You feed it a paper, and it surfaces not just who cited it, but what they said about it. The contrast citations are gold when you're trying to steelman an opposing view or spot methodological disputes that abstract databases miss. The agent works best when you're verifying a specific claim rather than exploring a broad topic. If you're checking whether a 2019 finding about protein folding still holds water, Scite will show you the three papers that replicated it and the one that found boundary conditions. That's faster and more reliable than skimming abstracts yourself. It also flags retractions, which sounds basic but catches things that slip through PubMed updates. Failure mode: it's only as good as its corpus. Niche fields or very recent preprints won't have enough citation data for the classification to mean much. And it doesn't write the literature review for you, it just arms you with better evidence. You still need to interpret why a paper contrasted the original claim. Compared to something like Elicit, Scite is narrower but more rigorous. Elicit will summarise papers and answer questions; Scite tells you what the scientific community actually thinks about a result over time. If you're writing a grant or a rebuttal, that's the difference between a defensible claim and a citation that looked good until someone checked it. The paid tier is necessary for bulk citation analysis. Free access gives you a taste but throttles the network exploration that makes it useful. For PhD students or anyone writing systematic reviews, it's worth the cost. For casual literature browsing, probably not.
Verdict

Pay for it if you're writing anything that needs to survive peer review or if you regularly need to verify whether a cited result is actually consensus. Skip it if you just need paper summaries or broad topic overviews.

Good at

  • Classifies citations as supporting, contrasting, or mentioning, surfacing hidden disagreements
  • Flags retractions and contradicted findings that standard databases miss
  • Crawls citation networks autonomously, saving manual reference-chasing
  • Particularly strong for verifying specific claims in established fields
  • More rigorous than summarisation tools when building defensible arguments

Watch out

  • Limited value in niche fields or for very recent preprints with sparse citations
  • Doesn't write or synthesise the review, just provides better evidence
  • Paid tier required for bulk analysis and full network exploration
  • Narrower scope than general research agents like Elicit
  • Classification quality depends on corpus coverage and citation context

Use cases

  • Verifying what other papers say about a result
  • Spotting retracted or contradicted findings
  • Citation network exploration
  • Building defensible literature reviews