Models
A mixture of models, cross-checked.
Every frontier model (Claude, GPT, Gemini, Mistral, Grok) alongside our own, routed per sub-task and made to check each other. No single-model bias.
Explore Science vs Writefull
Writefull runs on language models trained on published journal articles, inside Word, inside Overleaf, and inside the systems publishers use to sort incoming manuscripts. Explore Science reads the research itself and scores it.
The short answer
Most researchers meet Writefull because a university licence or an Overleaf setting put it there, and for repairing academic English that is a fair place to leave it. What it will not tell you is whether the study behind the sentences holds: the score it produces grades language quality so a publisher can route your file to the right copyeditor. Explore Science gives you a Calibre score out of 100 across eight fixed categories covering design, claims and evidence, with every reference resolved and tested against the claim it carries. If the draft is finished and the open question is whether a reviewer will believe it, run it through Explore Science for $49 before you submit.
Compared July 2026 · from Writefull's own published materials
Choose Explore Science when
- Calibre scores the research out of 100 across eight fixed categories, with the method published at /scoring; Writefull's published score grades language quality for triage.
- Every reference is resolved against live scholarly databases and flagged when it does not support the claim it is attached to.
- Novelty Check searches the live literature through Semantic Scholar and OpenAlex, so you learn what has already been published before an editor tells you.
- Several frontier models and our in-house Explorer One cross-check each other over a single pass of up to 2 hours.
- Every revision is scored and the movement is shown category by category, in one workspace your co-authors can open.
Choose Writefull when
- Language models trained on millions of published journal articles, which is why its corrections sound like the field rather than like generic grammar software.
- It works where you already write: a Word add-in and an Overleaf extension you switch on in your account settings.
- For an author writing in a second language, it fixes word choice, phrasing and word order sentence by sentence; Explore Science does not touch your prose.
- Free tools on Writefull X (Academizer, Paraphraser, Title Generator, Abstract Generator, TeXGPT, GPT Detector) with no account cost attached.
- Over 1,500 institutions hold licences, Stanford, Oxford, Monash and the University of Tokyo among them, so it is often already paid for.
Side by side
| Capability |  | Writefull |
|---|---|---|
| What it does | Reviews the research: design, claims, evidence, references | Corrects and rewrites academic English; scores language quality |
| Where it runsExplore Science loses this one | Its own browser workspace; nothing to install in Word or Overleaf | Word add-in, Overleaf extension, and publisher systems via API |
| Turnaround | One pass of up to 2 hours on a finished draft; you start it and come back to the report | Suggestions appear as you write; Revise returns an uploaded file marked up in Track Changes |
| Review engine | Claude, GPT, Gemini, Mistral and Grok run alongside Explorer One, our own model, each pass checked against the others | Writefull's own language models, trained on millions of published journal articles |
| What the score means | Calibre: research quality out of 100 across eight fixed categories, method published at /scoring | A language quality score or category, used to triage manuscripts and assign copyeditors |
| References | Every reference resolved against live databases, and flagged when it does not support the claim attached to it | Writefull Cite flags sentences that need a citation and have none; the published tooling does not check the references already in your bibliography |
| Literature and novelty | Novelty Check searches Semantic Scholar and OpenAlex for prior work against your contribution | No literature search in the published feature set |
| Language editingExplore Science loses this one | Not offered; we do not rewrite your prose | Grammar, word use, phrasing, word order, style changes, paraphrasing, split and join |
| Revisions | Every resubmission scored, so you see which categories improved | Revise returns Track Changes you accept or reject; no draft-to-draft scoring published |
| Paper-grounded chat | Rosa opens already knowing the manuscript, the review it just received, and the rest of your published work | An AI assistant inside Word that answers using the surrounding document context |
| Integrity checks | Plagiarism, AI-text and AI-figure detection available as plug-ins on the scored review | Free GPT Detector on Writefull X; no similarity or plagiarism check in the published product line |
| Training on your text | We do not train on your manuscript | States that texts and searches are neither stored nor used for training |
| Pricing | Nothing to pay to start; $49 buys one full-depth review with no plan attached, and monthly plans open at $99 | Free tier with daily quotas; Premium lifts them per seat; Revise is licensed to institutions and publishers only |
90% rate a review on par with or better than the human peer review they have had.
Two in three rate it better. From 304 head-to-head ratings by researchers who had received peer review on the same paper. These are our own users, self-reported, which is worth knowing when you weigh them.
Depth
Why a thorough review takes the time it takes.
Most tools optimize for speed. Explore Science runs the engine we built to conduct original research autonomously, and points it at your manuscript: it reads recent literature in your field, checks every reference, and works through the paper the way a careful reviewer does.
The time is the method, not an overhead on it.
How a paper is scoredTime spent thinking about your paper
How long a review actually takes
Single-pass LLM
~2 min
Other AI reviewers
5 to 20 min
Human peer review
(active reading time)
~30 min
Explore Science
up to 2 hrs
Industry observation; self-reported vendor figures; Publons peer-review time survey, 2018.
The engine
What actually reads your paper.
Three pieces of the architecture do the work behind every review.
Agents
Scientific agents that reason like the field.
Dozens of specialised agents work each review in parallel, each tuned to one dimension of analysis, spawning more when a section needs them.
Infrastructure
An in-house science stack.
Memory that holds your paper in context for the whole review, live verification of every citation, and novelty assessment. Built over years of autonomous science.
How Writefull gets into your workflow
Writefull is owned by Digital Science, which also owns Overleaf, and the Overleaf integration is a switch in your account settings rather than a purchase. More than 1,500 institutions hold licences covering their students and staff. On the publisher side, its Language API, Manuscript Categorization API and Metadata API sit inside submission and production systems: the American Chemical Society and the Royal Society of Chemistry have integrated it, Cambridge University Press used it on COVID-19 papers, and the language-editing firm Charlesworth runs the Language API.
So a lot of researchers are using Writefull without ever having compared it to anything. That is worth knowing before you assume it covers ground it does not.
What the language quality score is for
Writefull's Manuscript Categorization API assigns a language quality score or category so a publisher can route incoming manuscripts: which files need heavy copyediting, which editor gets them, which can go straight through. Benchmarked against the American Chemical Society's own quality analysts it reached over 95 percent agreement, and applying its editing service before copyediting moved the share of well-written papers from 58 percent to 81 percent. Those are real numbers about real work, and none of that work is about whether the study is sound.
Calibre answers the other question. It scores the manuscript out of 100 across eight fixed categories covering the design, the analysis and the evidence behind each claim, with the method published at /scoring, so an underpowered comparison or a conclusion the data will not carry arrives as a number you can argue with rather than a note about phrasing.
The references already in your bibliography
Writefull Cite works forward: it screens your text for statements that present non-general knowledge with no citation behind them, and flags the gaps. Useful while drafting, and nothing in Writefull's published feature set then goes back over the citations you have and checks them.
Explore Science resolves every reference against live scholarly databases and reports the ones that fail to resolve or do not say what you attributed to them, next to the claim they were meant to support. Novelty Check runs the other direction, searching Semantic Scholar and OpenAlex for work that already covers what you are presenting as new.
Common questions
My university licence already covers Writefull. Where does a $49 review fit?
The licence buys you unlimited language help across Word, Overleaf and Revise, and it is worth using. It does not buy a judgement on the research. Explore Science charges $49 for one full-depth review with no subscription, which is the cheapest way to find out whether the science survives a reviewer before you spend three months finding out from the reviewer.
Can I get Writefull Revise as an individual?
No. Revise, which takes an uploaded document and returns it with edits in Track Changes, is licensed to institutions and publishers; your university, research centre or company needs a licence for you to have access. The Word add-in, the Overleaf extension and the Writefull X tools are open to individuals.
Does either tool train on my manuscript?
Neither does. Writefull states that texts and searches processed through it are not stored or used for training, and we do not train on your manuscript. If your work is unpublished and sensitive, both are defensible on that point.
My last rejection was about my English, not my methods. What should I do?
Fix the English with the tool built for it, which is very likely already licensed to you. Then run the paper through a review before the resubmission, because a desk rejection framed as a language problem is sometimes an editor being polite about a weak evidence chain, and Calibre will tell you which of the two you are dealing with.
Claims about Writefull reflect its own published materials; everything about Explore Science reflects what the platform does today. Sources: Writefull, Writefull for Word, Writefull for Overleaf, Writefull Revise, Writefull Revise user guide (institutional access), Writefull Cite, Writefull X free tools and GPT Detector, Writefull for Publishers, Manuscript Categorization API, Writefull Language API, American Chemical Society case study, Charlesworth integrates the Language API, Writefull for Institutions, Writefull for Enterprise, Digital Science acquires Writefull (November 2023), Writefull plans. Something out of date? and we will fix it.
Already using Writefull? Put the same paper through a scored review.
Get started free