Privacy & Data

What we store

Your account information: the email address and password credentials you sign up with, managed by Supabase Auth (passwords are hashed and never visible to us).

Publications you upload: the PDF files themselves, and the research profile extracted from each one (research question, methods, findings, keywords, and similar fields), including any edits you make to it.

Publications you import: when you import from an ORCID iD or an author-name search, we store the bibliographic record returned by those public services (title, authors, journal, year, DOI, PubMed ID, link, and the published abstract) plus any research profile built from it. No file is stored, because an import never includes the full text.

Content you generate: every visibility asset created from your profile (summaries, social posts, briefs, statements), and the funding matches found for it, including which you saved or dismissed. Deleting a publication deletes all of it.

Images tied to a publication: figures and tables lifted from your PDF, anything you upload yourself, and any AI illustration you ask for. Removing an image deletes the file.

Your brand kit: the colours you choose and the logo you upload, so your visuals come out in your own branding rather than ours.

Your research agenda: if you merge your publications into a combined program, or export a promotion and tenure narrative, we store the result so you can come back to it. The annual impact report is assembled from these stored pieces when you download it and is not itself stored.

Funders you choose to watch (a list of funder identifiers), and fit memos you ask for (a short positioning note per paper and funder, kept so it is written once).

Your account plan and a count of how many generations you have run this month, so free and paid limits can be applied. If you subscribe to Researcher Pro, the purchase is made through Stripe acting as the merchant of record, using its Link checkout: your card details and billing address go only to Stripe, which uses the address to calculate any sales tax owed and handles receipts and payment support. What we keep is Stripe's identifier for you, your subscription's identifier, its billing interval and renewal date. Your card never touches our servers.

We never store or process student records, grades, advising notes, disability information, financial aid information, or protected health information. The platform only ever touches published research and public grant and funding data.

We run no third-party analytics and no tracking. There is no Google Analytics, no advertising pixel, and no third-party script anywhere on this site. The only cookies we set are the ones that keep you signed in.

What we do count, ourselves: how many times each public page was viewed each day, and how many searches were run, as plain totals. A count is three things: the page, the day, and a number. It contains nothing about you — no IP address, no cookie, no device or browser detail, no identifier of any kind — so there is no way to connect one view to another, or any view to a person.

One narrow exception, stated so the sentence above stays true: to stop a script hammering the support chat, we count how many chat messages come from a given caller each hour. Your IP address is not stored. It is turned into an unreadable hash with a secret key, and those rows are deleted after two hours. It is a counter, not a record of you.

The try-it box on the home page works the same way. If you drop a paper there without an account, the PDF is read once and not stored; what is kept is the extracted profile and the paper's title, for seven days, together with the same kind of hashed caller count (three tries a day per caller). A cookie in your browser remembers which try was yours so that, if you create an account within the week, one click keeps the paper. After seven days, or as soon as you keep it or dismiss it, the held copy is deleted.

Where it's stored

All data lives in a Supabase project (Postgres database and file storage) tied to this application. Uploaded PDFs are stored in a private Supabase Storage bucket accessible only to your account.

Retention and deletion

Your data is retained for as long as your account exists. Deleting a publication removes its file, its extracted profile, its figures, and everything generated from it, straight away.

Deleting the whole account is not yet a button you can press yourself. Email us at the address on your account emails and we will delete everything, files included.

AI processing

PaperOrbit generates content when you ask it to, which means a language model runs at that moment rather than in advance. That model is Anthropic's Claude, called only from our servers, never from your browser. Anthropic states that it does not use content submitted through its API to train its models, and we do not use your content to train models of any kind.

Here is everything we send to Claude, in full. When you ask for a fit memo, your saved profile fields for that paper and the funder's public IRS grant record. When you upload a PDF, the file itself, so the profile can be extracted from it. When you import a record instead, its title, authors, journal, year and published abstract, because no file exists. When you generate any written piece, card copy, or funding explanation, your saved profile fields. When you scan a paper for figures and tables, images of your figures and images of the rendered pages, up to twelve at a time. When you build a combined research program or a promotion and tenure narrative, the profiles of the publications you include. When you request a funding brief, the public IRS and federal award records the brief analyses. And anything you type into the support chat on this site.

One other AI service is used, and only if you ask for an illustration: Recraft, which draws it. Recraft receives a short scene description written by Claude and nothing else. No part of your paper, your profile, or your account reaches it.

Nothing you generate is published anywhere automatically. PaperOrbit has no connection to any social network. Every output is a draft that stays in your account until you review it and copy it out yourself.

Funding data is read from public U.S. government sources: grants.gov, NIH RePORTER, the National Science Foundation awards API, and USAspending. Keywords derived from your profile are sent to them as search terms; your publication itself is never uploaded to any of them. Foundation and charity data comes from IRS Form 990-PF and Form 990 Schedule I filings that we have already downloaded, so searching it sends nothing anywhere.

Publication import reads three free public scholarly APIs: ORCID, Crossref, and PubMed (NCBI E-utilities). We send only what identifies the work you asked for, which is the ORCID iD or author name you typed and the DOIs or PubMed IDs of the records it returned. Nothing about your account is sent, and we do not use Google Scholar, which has no public API.

Security

Access to your data is protected by Supabase authentication and row-level security, so your records and files are only readable by your own account. All traffic is encrypted in transit over HTTPS.