Reviewers read for patterns. So does Clasr.
The gap between what a paper claims and what its evidence actually supports. That’s what Clasr finds.
The core function.
Clasr reads the whole manuscript; how the topic gets framed, what the methods actually show, where the figures land, what’s left uncertain. Eleven sections, each read the same disciplined way.
What comes back isn’t a summary. It’s every observation, labeled: where it’s located, what kind of read it is, how serious it looks. You get the map, not someone’s opinion of the map.
Every manuscript opens with a promise. The question is whether that promise survives.
The Argument Chain follows the central claim from introduction to conclusion, watching whether the manuscript stays inside the argument it set up, or quietly redefines it once it reaches the discussion.
Claim scope
What’s being claimed, where, and at what level.
Premise visibility
Are the argument’s foundations shown, or just assumed?
Method-claim fit
Can the design actually carry the claim being made?
Finding alignment
Do the reported findings match how they’re framed?
Conclusion scope
Does the conclusion stay inside what the evidence allows?
A pattern this catches often: a manuscript that’s correct section by section, but quietly redefines its claim by the time it reaches the discussion. Reviewers notice this. So does the Argument Chain.
Editorial concern rarely comes from a single flaw.
It comes from clusters: a scope mismatch next to an over-promising abstract, opaque methods, a thin reproducibility profile. Any one might pass alone. Together, they often don’t.
The Desk-Reject Profile maps where these risks cluster before submission happens, and flags it when two critical gaps combine into something worse than either alone.
This is visibility, not prediction. Clasr doesn’t estimate whether a manuscript gets rejected, since that depends on things no system can see, like journal strategy or submission load. It surfaces the patterns that tend to show up together right before one does.
A Q1 journal reads differently than a Q3 journal.
You choose the tier you’re targeting, and Clasr reads accordingly. If you’d rather not choose, Auto is there as a backup, not the default. Most manuscript tools skip this step. They apply the same scrutiny no matter the target, so a borderline paper gets graded too harshly and a strong one gets let off too easy.
Tier and field work together too: a Q1 legal manuscript isn’t read by the same criteria as a Q1 clinical trial, because their review cultures don’t share the same criteria. Reporting standards like CONSORT, STROBE, PRISMA, STARD, ARRIVE, COREQ, and APA 7 are applied where relevant.
One detection engine. Three audiences.
Author Mode
Plain-language explanations, plus a sense of what a reviewer might ask. Built for authors who don’t speak the full signal vocabulary yet, and don’t need to.
Reviewer Mode
Just labels and locations. No explanation, no prose. Built for readers who want density, not context.
Advisor Mode
Signals ranked globally, for supervisors, writing centers, and institutional support teams. Built for triage, not deep reading.
Detection doesn’t change between modes. The signal layer underneath stays the same. Only the presentation does.