Skip to content
ChatGPTAppsRank

Scite ChatGPT App Review

Ground answers in peer-reviewed literature, and see whether a finding is supported or contradicted by later research.

Editorial79Doc-verifiedResearchDoc-checked Aug 24, 2026

Reviewed by the ChatGPTAppsRank editorial team · Verified against primary documentation · How we test · No paid placements

Doc-verified. Verified against primary documentation (OpenAI's release notes and the provider's own docs), but not yet hands-on tested inside ChatGPT. Capability claims here rest on those sources — see Sources & verification below.
TL;DR

Quick answer

Worth using?
YesSmart Citations are the reason this scores where it does. Every research tool can tell you a paper exists and most can tell you it is well cited; classifying citations as supporting, mentioning or contrasting answers the question researchers actually have, which is whether a finding survived contact with later work. That is a meaningful step beyond retrieval, and it is the correct antidote to a model confidently citing a paper that has since been contradicted. It sits just below Consensus, and the gap is value rather than capability: Consensus works on a free tier and this needs a subscription after a 7-day trial. Scite's own documentation notes ChatGPT sometimes disconnects tools silently between sessions, which is worth knowing before you trust a long research thread. A correction: we previously listed Scite as a vendor-hosted MCP server rather than a directory app. Both exist; the app was the part we missed.
Best for
Researchers and students who need claims traceable to literature, and want to know if it holds up
Skip if
Anyone without a Scite subscription, or topics with little published research
Editorial Score
79/100

Quick verdict

Smart Citations are the reason this scores where it does. Every research tool can tell you a paper exists and most can tell you it is well cited; classifying citations as supporting, mentioning or contrasting answers the question researchers actually have, which is whether a finding survived contact with later work. That is a meaningful step beyond retrieval, and it is the correct antidote to a model confidently citing a paper that has since been contradicted. It sits just below Consensus, and the gap is value rather than capability: Consensus works on a free tier and this needs a subscription after a 7-day trial. Scite's own documentation notes ChatGPT sometimes disconnects tools silently between sessions, which is worth knowing before you trust a long research thread. A correction: we previously listed Scite as a vendor-hosted MCP server rather than a directory app. Both exist; the app was the part we missed.

Who should use it

Researchers and students who need claims traceable to literature, and want to know if it holds up

Who should skip it

Anyone without a Scite subscription, or topics with little published research

Overview

Scite's ChatGPT app, launched 27 February 2026 and described by Scite as verified by OpenAI, answers research questions from a corpus of peer-reviewed literature and returns citations you can check. Its distinguishing feature is Smart Citations, which classify each citation as supporting, mentioning, or contrasting the finding it references — so instead of learning that a paper is cited a lot, you learn whether the citations agree with it. Type @Scite in a conversation to invoke it.

Pros

  • Smart Citations show whether later work supports or contradicts a finding
  • Answers are traceable to real, checkable citations
  • Described as verified by OpenAI, so authentication and permissions are handled
  • Invoked directly with @Scite, no configuration

Cons

  • Requires a paid Scite subscription after a 7-day trial
  • Scite documents that ChatGPT can silently disconnect tools between sessions
  • Thin on topics without much published literature
  • Consensus covers similar ground with a free tier

Test results

Pending hands-on test, which needs a Scite subscription. The launch date, the verified-app status, the Smart Citations classification, the @Scite invocation and the silent-disconnect caveat all come from Scite's own documentation.

Sources & verification

Availability and capability claims on this page are grounded in the primary sources below — OpenAI's own documentation and release notes, or the vendor's own announcement.

Setup experience

No configuration described beyond connecting your Scite account; type @Scite in a conversation to activate. Note the documented need to reconnect manually if ChatGPT drops the tool.

In-chat experience

Findings and citations render inline with their support-or-contrast classification, which is the part worth reading and the part that survives being flattened into text.

Privacy and permissions

Narrow disclosure: your queries and your Scite account, not personal records. Research questions can reveal what you are working on, which matters in a competitive or pre-publication sense.

Pricing and value

Value scores 66 because Consensus does much of this without a subscription. If Smart Citations matter to your work, the premium is defensible; if you just need cited answers, start with the free option.

Requires a Scite subscription. A 7-day free trial is documented.

Score breakdown

How we score apps →

Each dimension is scored 0–100 based on hands-on testing. Weight shows the share of the final editorial score.

  • Usefulness25%86

    Does it actually solve a real problem inside ChatGPT?

  • Reliability20%82

    Does it consistently return correct, complete results?

  • Ease of use15%80

    Is it discoverable, predictable, and forgiving?

  • Setup10%78

    How quickly can a real user get to first value?

  • In-chat experience10%78

    Does it feel native inside ChatGPT, or grafted on?

  • Privacy clarity10%78

    Are permissions, scopes, and data flows clearly disclosed?

  • Value10%66

    Pricing, free tier quality, and overall value for money.

Best alternatives

Also in research

Other top-scoring apps in the research category.

See all best ChatGPT apps for research

Scite compared with

  • AccuWeather vs Scite in ChatGPTBoth apps score 79/100. Pick by use case: any question where the answer depends on live weather rather than general knowledge → AccuWeather; researchers and students who need claims traceable to literature, and want to know if it holds up → Scite.
  • Consensus vs Scite in ChatGPTConsensus edges out Scite on overall editorial score (84 vs 79). Researchers, students, and analysts who need answers tied to citable peer-reviewed sources → start with Consensus.
  • Expedia vs Scite in ChatGPTScite edges out Expedia on overall editorial score (79 vs 76). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • Kayak vs Scite in ChatGPTScite edges out Kayak on overall editorial score (79 vs 74). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • MakeMyTrip vs Scite in ChatGPTScite edges out MakeMyTrip on overall editorial score (79 vs 66). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • ScholarAI vs Scite in ChatGPTScholarAI edges out Scite on overall editorial score (86 vs 79). Researchers and graduate students who need peer-reviewed sources with real DOI citations → start with ScholarAI.
  • Scite vs Skyscanner in ChatGPTScite edges out Skyscanner on overall editorial score (79 vs 78). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • Scite vs WebPilot in ChatGPTScite edges out WebPilot on overall editorial score (79 vs 74). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • Scite vs Wikipedia in ChatGPTScite edges out Wikipedia on overall editorial score (79 vs 73). Researchers and students who need claims traceable to literature, and want to know if it holds up → start with Scite.
  • Scite vs Wolfram in ChatGPTWolfram edges out Scite on overall editorial score (87 vs 79). Students, scientists, engineers → start with Wolfram.

Final verdict

Editorial79Worth using? Yes

Smart Citations are the reason this scores where it does. Every research tool can tell you a paper exists and most can tell you it is well cited; classifying citations as supporting, mentioning or contrasting answers the question researchers actually have, which is whether a finding survived contact with later work. That is a meaningful step beyond retrieval, and it is the correct antidote to a model confidently citing a paper that has since been contradicted. It sits just below Consensus, and the gap is value rather than capability: Consensus works on a free tier and this needs a subscription after a 7-day trial. Scite's own documentation notes ChatGPT sometimes disconnects tools silently between sessions, which is worth knowing before you trust a long research thread. A correction: we previously listed Scite as a vendor-hosted MCP server rather than a directory app. Both exist; the app was the part we missed.

Frequently asked questions

Scite or Consensus in ChatGPT?
Consensus if you want cited research answers without paying — it scores higher overall largely on value. Scite if you need to know whether a finding has been supported or contradicted by later work, which is the thing Smart Citations do and Consensus does not.
Do I need to pay?
Yes, after a documented 7-day free trial. A Scite subscription is required.
Is it actually in the ChatGPT directory?
Yes — Scite describes it as a verified app in the ChatGPT app directory, and OpenAI's curated plugin marketplace carries a Scite plugin manifest. Scite also runs a separate MCP server for other AI tools; the two are different things and we previously documented only the second.

Used this app inside ChatGPT? Share your experience

Community feedback coming soon. For now, contribute a correction or a missing app via the links below.

Use cases

  • Literature reviews with synthesised findings
  • Fact-checking a claim against published research
  • Finding papers from a plain-language description
  • Checking whether later work supports or contradicts a finding
  • Pre-meeting or pre-writing research briefings

Trust

Editorial Score, not user rating. Community feedback coming soon.

How we rank apps →