Why ScholarGate exists.
Sources, not fluency.
Almost anything can write a confident paragraph about your field. Almost nothing can tell you where each sentence came from. ScholarGate is built around that second problem.
What does a generic chatbot get wrong?
Generic AI chatbots are trained to sound right. When a source is missing they do not stop; they improvise, and an improvised reference looks exactly like a real one.
The person who pays for that is you, at review time. ScholarGate runs on the opposite rule: an answer is tied to the method pages it stands on, and when the library has nothing to stand on it says so instead of filling the gap.
- Answers cite the method pages they stand on.
- A missing source is reported as missing, never invented.
- Every cited page is one click away, so a claim takes seconds to check.
Where do the answers come from?
The library is built on 3 open catalogs. Method pages are assembled from records in OpenAlex, Semantic Scholar and Crossref, and each page keeps the trail back to what it was built from.
Open sources are a deliberate choice. Anyone can open the same records ScholarGate was built from, and nothing important rests on a catalog only we can see.
- OpenAlex, Semantic Scholar and Crossref, named in the open.
- Each method page carries the records it was built from.
- Chat answers are grounded in those same method pages.
What does that look like side by side?
Two design rules, not two feature lists. A general model is genuinely better at open ended conversation, and that is the same reason it cannot promise the column on the left.
- What a citation is tied toScholarGateThe method pages the answer stands onGeneral chatbotsNothing the reader can open
- When there is no evidenceScholarGateThe gap is stated in the answerGeneral chatbotsThe gap can be closed with fluent text
- Source records on the pageScholarGateIncludedGeneral chatbotsNot included
- Sections it will draft for youScholarGateIntroduction and Literature ReviewGeneral chatbotsWhatever you ask for
Where this sits in the market
Named tools, not a straw man. Most of them do something we do not, and a few do it better. The point of this section is to show which corner is actually ours.
The combination is the product, not any single part of it.
Paper search tools
General assistants
Writing and method references
Method pages you can reread
ScholarGateA catalog of method pages, each carrying its records
General assistantsExplained inside the reply, nothing to reopen later
Paper search toolsThey return papers, not method write ups
SciSpaceSearch and PDF chat, not a method catalog
What a chat answer stands on
ScholarGateThe method pages cited under it
General assistantsTrained text; fabricated references measured at double digit rates
Paper search toolsThe papers the query returned
SciSpaceThe paper you opened, or its own index
What a draft is allowed to cite
ScholarGateOnly the sources in your own pool
General assistantsWhatever the model produces
Paper search toolsThey do not draft sections
SciSpaceIts writer can add citations from its own index
Citation context on a claim
ScholarGateNot offered
General assistantsNot offered
Paper search toolsScite labels citations as supporting, contrasting or mentioning
SciSpaceNot offered
Screening or mapping a whole literature
ScholarGateNot offered
General assistantsNot offered
Paper search toolsElicit extracts across many papers, ResearchRabbit maps them
SciSpaceSearch across its own index
Breadth of the paper index
ScholarGateNarrow on purpose: a method catalog, not a full index
General assistantsNo index of its own
Paper search toolsHundreds of millions of records
SciSpaceHundreds of millions of records
Fabricated reference rates: 18 percent for GPT-4 and 55 percent for GPT-3.5 in a peer reviewed sample (Walters and Wilder 2023, Scientific Reports), and 28.6 to 39.6 percent in a systematic review setting (Chelli et al. 2024, JMIR). Those figures describe the 2023 to 2025 model generation, not a permanent property of any tool. Everything else in this table was read from each product's own public description.
What stays yours?
The Studio drafts Introduction and Literature Review from the sources on your desk. Methods, Results and Discussion are not on a roadmap somewhere; they are deliberately out of scope.
Those sections carry your data and your argument. A tool that writes them is writing your paper. That is not a missing feature, it is the line we drew.
- Two sections drafted with you, the rest untouched.
- Drafts come from the references you bring: your bookshelf, or your own .bib and .ris files.
- One research question keeps the whole review in focus.
How is the library kept honest?
The library was grown content first. It holds 8,100+ method pages and reads in 46 languages.
A page earns its place by explaining a method well enough for a person to use it.
So which side are you on?
Read first or write first. Both sides of ScholarGate run on the same sources and the same rule about naming them.
Questions people actually ask
No. The Studio drafts an Introduction and a Literature Review from the sources you bring, and stops there. Methods, Results and Discussion stay yours to write, which is also what your reviewers expect.
You get told. An answer that cannot be grounded in the library is returned as ungrounded rather than dressed up with a plausible citation. That is the whole point of the design.
From open scholarly catalogs: OpenAlex, Semantic Scholar and Crossref. Method pages keep the records they were built from, so anyone can follow a page back to its source.
Yes. The desk drafts from the bookshelf you collect inside ScholarGate and the .bib or .ris files you upload. The Review Studio can also run its own literature search across the open catalogs, and it asks you to approve the search plan and the outline before a single draft line is written.
ScholarGate is in early access and the plans are still settling. The pricing page shows where they stand today, and the waitlist exists so that you hear from us before anything changes.
See whether it holds up.
ScholarGate is in early access. Join the waitlist and we will write to you before anything changes.
Join the waitlist