Privacy Policy
Effective October 7, 2026
Summary
- ResearchPDF has no account and no server. The developer cannot collect or see any of your data.
- PDF files are read only inside your browser. Their contents are not sent anywhere.
- Web PDFs you have opened (and local PDFs that carry an arXiv stamp) are kept in the extension's storage on this device so they reopen quickly (up to 1 GB; you can turn this off or clear it in settings). These copies never leave the device.
- Only if you connect a Google account are your drawings, reading positions and the list of PDFs you opened (titles and addresses, pinned or not) saved in an app-only folder in your own Google Drive.
- With paper info on, only paper identifiers (DOI, arXiv ID, title) are looked up in scholarly databases.
Data stored in your browser
The following stays in the extension's local storage (IndexedDB, chrome.storage) and is deleted with the extension.
- Document identifiers (the PDF's internal ID and a hash of the file), the last page and zoom, the file name or URL
- The documents opened in a PDF tab: address, file name, PDF title and recognised paper title, when last opened, pinned or not
- Your annotations (pen, highlight, text, stamp)
- Projects: name, documents, pinned documents, the tabs open when closed, icon and color, folders and order
- A cache of paper lookups
- Settings, and the Semantic Scholar and OpenAlex API keys if you entered any
Google Drive sync (optional)
When you connect a Google account in settings, ResearchPDF asks only for the Google Drive API's drive.appdata scope. It reaches only a hidden app-only folder; ResearchPDF cannot even list your other Drive files.
Only if you also turn on "Keep PDF files in my Drive too", ResearchPDF asks for the drive.file scope as well. It then creates a visible folder named ResearchPDF in your Drive and uploads the PDF files you choose (or that a project's rule says to keep) there; with this scope it can see only the folder and files it created itself, never your other files. Turning it off stops uploads; files already there stay in your Drive, under your control.
- What is synced: document identifiers, reading page and zoom, file name or URL, your annotations, the list of PDFs you opened (address, title, when last opened, pinned or not, kind of paper), projects (name, documents, pins, the addresses of the tabs that were open, icon and color, folders and order). One gzip-compressed JSON file.
- Also synced: the names, notes and links you give documents, and your preferences (tab names and icons, language, paper strip, figure capture, which projects keep files in Drive).
- What is not synced: the PDF files themselves (unless you keep them in your Drive folder as above), the paper-lookup cache, API keys, and this device's access and storage settings.
- The data is stored only in your Google account. There is no developer server, and the developer has no access to it.
- Sign-in happens on Google's OAuth screen. ResearchPDF never handles your password, stores no long-lived (refresh) token, and keeps only an access token of about an hour in the browser session's memory.
- "Disconnect" revokes the token and removes the account details from this device. You can delete the data in Drive yourself under Google Drive settings → Manage apps.
ResearchPDF's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
Paper info (optional, on by default)
When the viewer recognises a DOI, arXiv ID or title in an open PDF, it looks up only that identifier in Semantic Scholar, OpenAlex, Crossref and arXiv. If the databases have no reference list, the reference list printed in the PDF is read on the device, and only each entry's DOI, arXiv ID or title is looked up in OpenAlex to add citation counts. Documents opened with an earlier version are checked once the same way while the PDF tab is idle, reading only the copy stored on the device and never downloading the PDF again. No other PDF text is sent. Figure and table detection, and reading a formula as LaTeX, run on the device with models shipped inside the extension; page images are not sent anywhere. If you enter an API key, it is sent to that service with the requests. You can turn paper info off in settings.
Permissions
- storage, alarms: storing the data above locally, and syncing periodically.
- webNavigation: noticing when a tab goes to a PDF address or a local PDF file, to open it in the PDF tab. Page contents are not read.
- declarativeNetRequestWithHostAccess: only when "Open web PDFs in ResearchPDF" is on and site access is granted, six rules send responses that are PDFs to the built-in viewer — top-level documents to the PDF tab, PDFs embedded in a page (iframe, embed/object) to the viewer in their place — each for a PDF content type, an
application/octet-streamresponse whose address ends in.pdf, or an inline.pdffile name. Downloads (attachments) and POST responses are left alone. - identity: opening Google's sign-in window.
- Site access (optional): requested only when you turn on "Open web PDFs in ResearchPDF" (or allow it from the viewer). Used to send PDF responses to the viewer as above; for the viewer to download the PDF, show it and keep its copy on this device; and to read the addresses and titles of tabs showing a PDF in Chrome's own viewer, so that home, settings and the welcome guide can offer to gather them into the PDF tab. Those addresses and titles stay on the device.
- File URL access (you turn it on in Chrome): reading PDF files on your computer in the viewer, and finding local PDFs open in Chrome's viewer to offer gathering them.
What is not collected
There are no analytics, ads, tracking or remote logging. The developer receives no data at all, usage statistics included.
Changes and contact
If this policy changes, the effective date on this page is updated. Contact: croksuter@gmail.com