Step 1

Compare registrations with papers

RegCheck extracts and compares text from your registration and paper, so you can see if and how a study deviated from its plan.

What are you comparing?

Choose the kind of documents RegCheck should check against each other.

Choose a parser

This determines how RegCheck reads your paper. PyMuPDF (recommended) runs entirely on our server and keeps all selectable text, including author notes and footnotes. GROBID is tuned for typeset journal PDFs but can fail on non-standard layouts. DPT-2 handles less standard academic documents.

Choose a model

ChatGPT and Claude Opus 4.8 are generally most accurate, but can be slower. DeepSeek can perform well but may be slower. Qwen3.6 27B is a fast, open-weight option.

Comparison dimensions

RegCheck compares the registration and paper on each of these dimensions in the order you specify.

Append previous outputs?

Should RegCheck include earlier dimension responses when prompting for later dimensions? “Yes” can improve performance but introduces dependence between decisions; “No” keeps outputs independent.

Does the paper report multiple studies?

This helps RegCheck know which study to focus on.

Which field is your study from?

We'll use RegCheck's default comparison dimensions for that field. To fine-tune individual dimensions, use Customize settings instead.

Where is your preregistration?

Upload files, provide a ClinicalTrials.gov identifier, or paste a link to the Open Science Framework.

Upload your files

Supported file types: PDF, DOCX, TXT, HTML.

Not signed in: this report is public, shareable by link, and auto-deletes after 7 days. Log in to keep reports and control visibility.