Skip to content

Find out what a reviewer would tell you. Before they tell you.

Can I Publish reads your manuscript with the same criteria real reviewers apply, and returns every point anchored to a verbatim quote from your text.

The file is deleted immediately once the text is extracted. The text, after 30 days. Nobody trains models on your manuscript.

The same manuscript, reviewed twice

On the left, what the two reviewers of a real article wrote. On the right, what the diagnosis flagged on that same manuscript in ninety seconds. The three entries match.

What came back · 2 reviewers

Review reports

Open review · verbatim

Reject · Approve with reservations

What you would have seen · minute 2

Diagnosis

Every point, with its verbatim quote

Major revision

The numbers match: three of the points their reviewers raised are three of the points the diagnosis flagged. Hover or tab through one to see both of its sides. Real case: «Universal mental health training for frontline professionals», Open Research Europe, open peer review under CC BY 4.0. The points come from a real diagnostic run on that same manuscript.

Skip the wait, not the truth.

Most rejections are decided in the first few minutes of editorial reading. Not on the science, but on signs of an unfinished manuscript. You were always going to hear it. Can I Publish just makes sure you hear it today, not in eight months.

Built on real reviews. Not AI guesswork.

To know what a reviewer looks at, you have to read what reviewers write. We gathered a corpus of 10,749 open peer review reports on 3,703 published articles: public, citable reports, with no private material. Our criteria came out of that: the checklists, one per dimension of a manuscript, that the diagnosis applies to yours.

We tested it against the alternative: pasting your paper into a generic LLM. And every point comes anchored to a verbatim quote from your manuscript, verified by code: 90% of them, against 52% for human reviewers quoting from memory.

Corpus and coding verified on 29 July 2026. The full method, with interrater reliability.

Can I Publish vs other LLMs

Can I Publish73%ChatGPT67%Claude58%% of what reviewers raise, detected
Can I Publishthe best generic configuration from each company, with a good prompt

How it was measured: prompts frozen before seeing a single result, half the articles held back, and the matching judged by a model from a different family.

Do you coordinate a group? The first pass is already done.

What you find in every draft before you even open it (the missing citations, the method that never says on what criterion it stopped, the conclusion that reaches past the data) is mechanical work that repeats with every new person. The diagnosis does it first, and the group works on top of it: each point is assigned to whoever will fix it, the assignment survives into the next version, and your conventions go into every review.

Five axes, each with a name.

First it detects your study design (quantitative, qualitative or mixed) and applies the matching criteria. Then it reviews these five axes.

01
Methodological rigour
Qualitative, quantitative and mixed designs, each with its own criteria. Half-described samples, missing saturation criteria, analyses the tables do not support, gaps between what the data show and what the text claims.
02
Citations and references
Orphan citations, phantom references, in-text calls that never reach the list, sources cited second-hand. The bibliographic hygiene an editor checks in the first few minutes.
03
Theoretical argument
Inferential leaps between what the source says and what the manuscript claims, a fragmented theoretical framework, conclusions that go beyond the available evidence.
04
Presentation
Structure and organisation, writing and clarity, editorial formatting, length and density. What makes a text read as finished or as a draft.
05
Contribution
Novelty and differentiation from what is already published, relevance of the contribution, and the overall assessment. The axis that answers «so why does this matter?».

And one separate output

Journal fit

Which journals take your kind of study and why: topic, method, article type. A ranking with the real reason for the fit, not a list of impact factors. It is not one of the five axes because it does not assess your manuscript, but its destination.

What it doesn’t do, said before you ask.

  • It does not rewrite your text or fix your style. There are better and cheaper tools for that.
  • It does not replace peer review: it anticipates it. Review assesses your scientific contribution; this points out what stops you reaching that assessment.
  • It does not guarantee you will be accepted. Nobody can, and anyone who promises you that is selling you something else.
  • It does not decide for you. It gives you the list and the verdict; sending or fixing is your call.

Request access for your group.

We are opening up by research group, not one person at a time. Leave us your email and what you work on, and we’ll let you know when your area comes up.

What is your role in your group?

We use your email only to handle your access request (pre-contractual measures at the data subject’s request, Art. 6(1)(b) GDPR). More in our privacy policy.

Questions.

How does Can I Publish work?
It first detects whether your study is quantitative, qualitative or mixed and applies the matching criteria. Then it reviews five axes: methodological rigour, citations and references, theoretical argument, presentation and contribution. Every point is anchored to a verbatim quote from your manuscript, so you can check it in ten seconds. The criteria are built and tested against an open peer review corpus of 3,703 published articles and 10,749 real reviewer reports, with 28,201 critiques followed to what happened in the next version. How it works, in detail.
What does Can I Publish NOT do?
It does not rewrite your text, it does not fix style or grammar, and it does not replace peer review. It does not guarantee acceptance either. It tells you what an editor will look at before you send it to them. That is all.
Does it replace peer review?
No. It is a diagnosis beforehand. Peer review assesses your scientific contribution; we point out the problems that get a manuscript rejected before it reaches that assessment.
What languages does it work in?
It analyses manuscripts in English or Spanish. Reports are delivered in English.
What happens to my text?
The original file is deleted once the text is extracted; the extracted text is removed after 30 days; nobody trains models on your manuscript.
How long does it take?
Ninety seconds per manuscript, timed. The report gets straight to it: concrete points you can mark as resolved, no filler.
Who is behind this?
Researchers in the social sciences and education. What we know comes from a corpus of 10,749 real review reports on 3,703 open-review articles, 1,423 of their comments coded one by one, and from having received a few ourselves.