Trust
How CiteMe Handles Your Files
Theses, drafts, and unpublished manuscripts are sensitive documents. This page explains exactly what happens to a file you give CiteMe — what stays in your browser, what reaches our servers, what is stored, and for how long. No badges, just the actual mechanics.
The short version
- ✓Uploaded PDFs are processed in memory and discarded — the file is never written to our storage unless you explicitly save it to your Pro library
- ✓When we only need to identify a paper (DOI, ISBN, PMID, title), the scan runs entirely in your browser — the PDF never leaves your device
- ✓Everything travels over HTTPS (TLS); library files are encrypted at rest
- ✓We never use your documents to train models, and we never sell your data
- ✓You can delete your data at any time — or ask us to
1. What happens to your file, flow by flow
| What you do | Where it runs | What leaves your browser | What we keep |
|---|---|---|---|
| Search a citation (text, DOI, ISBN, URL) | Our servers + academic databases | The query you typed | Request logs with hashed IPs, up to 90 days |
| Turn a PDF into a citation (identify the paper) | Your browser | Only the detected DOI, ISBN, PMID, or title — never the file | Nothing from the file |
| Extract a bibliography (PDF to BibTeX/RIS, Reference Checker, the free reader) | Our extraction server | The PDF or Word file, over HTTPS | The file is not written to storage. For PDFs, we keep the extracted references in a processing cache for up to 90 days, and a SHA-256 fingerprint of the file (which cannot reconstruct it) for up to 12 months for quota accounting. For Word (.docx) files, only the extracted text of the document is kept, in a processing cache for up to 6 hours so a retry does not re-parse it, keyed by a SHA-256 hash of the file that expires with it; there is no reference cache and no 12-month quota fingerprint for Word files. |
| Upload a PDF whose bibliography our parser cannot extract (including scanned and unusual-layout PDFs) | Google Gemini API (fallback only) | The PDF is sent to Google for structured extraction when our parser returns no references after a longer processing attempt, or when you explicitly request OCR | Same as above — the reference cache and fingerprint, never the file |
| Save a PDF to your library (Pro, opt-in) | Supabase Storage | The file you chose to save | Stored encrypted at rest until you delete it |
2. Who touches your file
- Vercel — hosts the application and terminates HTTPS. Your file passes through during upload; it is not persisted there.
- Our extraction server (hosted at Hetzner) — a dedicated server we operate, running the open-source GROBID engine that parses bibliography sections. The PDF is processed transiently and not written to disk storage.
- Google Gemini API — receives a PDF only as an extraction fallback when our parser returns no references after a longer processing attempt (including scanned and unusual-layout files), or when you explicitly request OCR. It is used solely for structured text extraction.
- Supabase — stores your account, citations, and (Pro, opt-in) library PDFs, encrypted at rest.
- Upstash (Redis) — rate limiting and quota counters using hashed identifiers, plus the extracted text of Word uploads for up to 6 hours; never the file itself.
The full subprocessor list, including analytics and payments, is in our Privacy Policy.
3. Encryption
- In transit: every request — browser to CiteMe, CiteMe to the extraction server, CiteMe to third-party APIs — travels over HTTPS/TLS.
- At rest: database records and library PDFs are encrypted at rest by our storage provider (Supabase, AES-256).
4. Deleting your data
- Library items and uploaded PDFs can be deleted from your account at any time; deleting your account permanently deletes all of them.
- Processing caches and fingerprints expire automatically on the schedules listed above — no action needed.
- For anything else — including deletion requests for data tied to anonymous usage — email privacy@citeme.app.
5. What we don't do
- We don't use your documents or references to train models.
- We don't sell or share your personal data with advertisers.
- We don't keep uploaded files after processing — not even "temporarily for support purposes".
- We don't read your documents. Extraction is automated; no human looks at your files.
6. Reporting a security issue
If you believe you have found a vulnerability in CiteMe, email support@citeme.app with the details. We read every report and will respond as quickly as we can.
Related: Privacy Policy · Terms of Service