Skip to content
Why DeskReview exists

AI in the editorial process, where it belongs.

Somewhere right now, a researcher is submitting a manuscript that took two years to write. It may come back within days, without a single reviewer having read it. The reason often has nothing to do with the science. A reporting item is missing, a statement is incomplete, or a checklist wasn't followed. On the other side of the desk, an editor is triaging a queue that grows faster than anyone can read it. Reviewers are giving up their evenings to papers that were never ready for review.

This is a quiet crisis in science communication. It rarely makes headlines, but it costs careers, delays findings, and wears out the people the whole system depends on.

A loud answer is taking shape: let AI read the papers, write the reviews, and make the calls. We think that answer is wrong. DeskReview exists because we believe there is a better one.

AI supports review. People make the decisions.

No AI system should decide whether research gets published. Every decision that shapes a manuscript's fate belongs to people, and so does the accountability that comes with it. That doesn't mean human review is flawless. It isn't, and pretending otherwise helps no one. What matters is that people and machines make different kinds of mistakes.

A person reading their fortieth manuscript of the week can overlook a missing item that a defined check will always catch. A machine will never grasp why an unconventional study matters. Neither is enough on its own. Each becomes valuable where it covers what the other misses.

So our goal is not to automate scholarly judgment. It is to give people narrow, well-tested tools that help them handle specific parts of an overwhelming process with more consistency and better information. We apply that to one specific moment: when a manuscript is ready to leave the author's hands.

What problem are we solving?

Missing details and routine checks take time away from the research. DeskReview helps authors identify reporting gaps before submission and prepare clearer, more complete manuscripts. It helps editors check submissions more efficiently and focus their attention on editorial decisions. And it helps reviewers spot overlooked details and follow findings back to the manuscript, leaving more time for a thorough assessment of the science. Across all three roles, the aim is to save time and support better-quality submissions and reviews.

What DeskReview does

DeskReview checks a manuscript against established reporting guidelines and the requirements of the journal it is headed for. Each finding is tied to a defined criterion and traced back to the passage in the manuscript that supports it. That way, authors and editors can check the basis for each finding themselves.

It also generates an AI review report: a source of questions and observations to help authors, editors, and reviewers think about the paper and spot items they might have missed. It offers material for reflection and independent assessment.

Retrieved papers and original abstracts give readers a starting point for exploring the research context. The reader judges how the studies relate; a bounded search cannot establish originality or completeness of the literature.

What DeskReview does not do

An AI review report is not a peer-review report. It cannot replace an expert’s reading of the manuscript or settle its novelty, significance, or scientific merit. DeskReview does not predict acceptance, rank submissions, or make editorial decisions. It never rejects anything. A flag is a question for a person, not a verdict.

Where the human decision sits

Always with people. The author decides what to revise. The editor decides what goes to review. Reviewers judge the science. DeskReview helps make gaps visible and offers questions to consider, in an interface crafted to make checking each one practical, not just possible. People decide which observations matter, which need further investigation, and which to set aside.

Research in European hands

Europe's strategy for AI in science aims to make Europe a leader in AI-driven research without giving up control of the tools it depends on. DeskReview is built in the Netherlands, under European law, and our hosted analysis uses a European AI provider with zero data retention and no model training. The research belongs to the people who did it.

Manuscript confidentiality

Many publishers promise authors that submitted manuscripts will be treated confidentially. DeskReview Local supports editors and reviewers in honoring that commitment: manuscript analysis runs entirely on your own Mac, without sending the manuscript to a cloud AI service.

Limitations - A scientific instrument, not an oracle

We call DeskReview a scientific instrument, and we mean it as a standard to live up to, not a slogan. An instrument earns trust by being examinable. It needs a consistent basis, a clear separation between fact and opinion, evidence you can verify at a glance, and criteria the research community can debate.

DeskReview is built by people who have spent their careers inside scholarly publishing, on the systems that move research from manuscript to the published record. We have seen where the process breaks, and we know it depends on human trust. Meet the team.

Let the machine keep the checklist. Let people keep the judgment.