From research idea to a submission-ready manuscript, with you at every step.

Evidence-bound and auditable support across the whole arc of clinical and scientific research, from design to publication. Three separate workflows, alongside physicians, residents, graduate students and academics.

Three separate workflows

You can enter at any stage — whether all you have is an idea, or a finished dataset.

Research idea

Starts from your field of interest and your available resources; searches the literature and assesses evidence gaps. Develops the research idea together with you.

Study protocol

Starting from the research question, you can build your study protocol.

Analysis and manuscript

First it understands your study; it checks your data and, with your approval, runs the statistical analysis, compares your findings with the literature and drafts the manuscript.

It also finds and compares the clinical guidelines relevant to your study.

Why it is different

What distinguishes this product is not how much it does, but what it refuses to do.

Approval gate The analysis plan is shown to you in plain language; no statistical, machine learning, meta-analysis or Bayesian computation starts without your explicit approval. If the plan changes, approval is renewed.
Evidence of computation Numerical results are reported only after the computation has actually run: method, included and excluded observations, assumption checks and warnings are given alongside.
Source verification The PMID or DOI of cited work is looked up; a source that cannot be verified is not used as a citation. If only the abstract was reached, no detail is claimed from the full text.
Suggestion and finding are kept apart Every material output is labelled as a suggestion, an assumption, a computed result, or a verified literature finding. Telling them apart is not left to the reader.
De-identified data A de-identified copy of the data is requested before analysis; direct identifiers are not repeated in the outputs.

These rules are the working contract of ClinicalResearcherAI. The final audit, however, is always yours: no output should be treated as final until you have verified it.

What it does not do

The limits are written down from the start.

It gives no patient-level diagnosis or treatment advice. It does not cross the boundary into clinical care.
It does not merge separate studies on its own. Datasets are not pooled unless you match them.
It does not choose the analysis to fit the result. Promoting whatever came out significant, data dredging and selective reporting are refused.
It does not analyse raw omics matrices. Its scope is a reasonable number of variables in tabular form; when that limit is crossed, it says so.
It does not take on scientific responsibility. Every output is presented for review; the decision is yours.

Start using it today

You can start with a research idea, a protocol, or a completed dataset. ClinicalResearcherAI runs inside ChatGPT; a ChatGPT account is all you need. Its own web interface is in preparation — to be notified, the button below opens an e-mail with the subject and body already written; you only need to send it.

Open in ChatGPT Notify me at launch There is no form on this page; the only information you provide is the e-mail you choose to send. Your address is used solely for the launch announcement, is not shared with third parties, and is deleted the moment you ask to be removed.