Cento is an AI co-writer for medical research. It can only cite papers it actually retrieved — so every reference opens, every claim is graded against its source, and nothing is invented.
Early access for clinician-researchers. Ophthalmology first.
Intravitreal anti-VEGF therapy reduced the risk of moderate vision loss in neovascular AMD over 24 months, though sustained benefit depended on adherence to monthly dosing.
A language model predicts the next plausible token. A reference is just more tokens to predict — so it will produce a flawlessly formatted author list, journal, year, and DOI for a paper that was never written. In most fields that's an embarrassment. In medicine, a single fabricated citation is a retraction risk.
"You can't stand behind a citation you can't open."
Searches PubMed, OpenAlex, Semantic Scholar and Europe PMC, and builds a candidate set of real, retrieved sources before a single word is written.
The model writes prose plus citation slots as structured data. It cannot name a paper — only reference a retrieved candidate by ID.
The backend checks every citation against the candidate set. No match means the claim is marked [UNCITED] — never invented.
As you edit, Cento re-scores each claim against its source. Move away from the evidence and it flags the claim as Drifted.
Every reference resolves to a paper you can open. When a reviewer questions a claim, you don't reach for an explanation — you reach for the source.
Macular thickness predicted treatment response5 across both cohorts6.
Manuscripts rot during revision: you sharpen a sentence and quietly outrun its source. Cento re-scores each claim as you edit, so drift surfaces while it's still yours to fix.
Retrieval, formatting, and verification happen as you write. You stay in the argument; the reference apparatus keeps itself honest underneath.
One draft, one reference list — every citation stays verified as it changes hands.
Co-authors draft, cite, and revise in the same manuscript — no emailed versions, no reconciling reference lists by hand. Everyone works from one source of truth, and every citation stays verified no matter whose turn it is to write.
Every reference binds to a paper Cento actually retrieved. If it can't bind, it isn't cited — it's flagged. No exceptions, no hallucinated DOIs.
Claims re-scored against their source on every edit.
Level-of-evidence on every source, so you can weigh it.
Drafts that already know what a methods section is for.
Captioned, numbered, and cross-referenced in text.
Word document plus an audit appendix of every source.
The paper exists and was retrieved from the open literature. Open it and read it yourself.
Its level of evidence is shown alongside it, so a case report never poses as a trial.
Your sentence is scored against the source, and drift is flagged the moment it appears.
We don't put a single accuracy number on this page. The real test is the one you run yourself — and as clinician users come on board, their published work, not our marketing, will be the proof.
Shared libraries, team controls, and verification at the point of submission — for departments and journals.
Join the early-access waitlist. Built for clinician-researchers — starting in ophthalmology.
No spam. We're honest about the stage — this is early access.