Privacy & Expectations

What We Collect

We store the information you provide, including your account details, research interests, searches, saved items, projects, and feedback. We also keep technical and email-delivery records to operate the service, troubleshoot problems, and improve the tool. Some activity is associated with your account.

We use public ORCID records for scientist discovery and suggested research interests.

We use a first-party cookie to keep you signed in on the website and iPhone app for up to 30 days. Sign-in expires after seven days without an authenticated request; background requests from an open app can count as activity. Signing out ends that session on that browser or device.

While you are signed in, we record limited activity such as pages visited, visible-page time, search outcomes, papers shown or opened, saves, and changes to relevance feedback. These records help us see whether recommendations are useful and where the site fails. They do not include copies of your search text, research interests, private notes, or page contents.

Paper links shared by signed-in users may contain a random referral ID associated with the paper and the sender's account. When someone opens a shared paper, we record a limited visit in the existing site session. If they sign in or create an account in that session, we can associate that visit with their account, signup, a save of that paper, or a later return for up to seven days after the most recent eligible referral. We do not show senders who opened their links. Forwarded links keep their original attribution.

Referral tracking uses no separate tracking cookie or browser fingerprint. It honors Do Not Track and Global Privacy Control signals. Links remain readable without tracking, including when a referral ID is invalid or expired. Recipient-level referral records have a 30-day retention target, and share-link records expire and have a 90-day retention target. Admin reports show aggregate counts, not lists of recipients.

How We Use It

We use this information to recommend papers, support collaboration, and understand what works and what needs improvement. Relevance scores are model estimates, not judgments of scientific quality.

Your research interests, one-paper reading goals, and relevance feedback are not public. Profile information you choose to publish is public, and shared projects are visible to their members.

Connecting an assistant is optional. When enabled, you can sign in and choose Allow to grant access to your research interests, feed criteria, public paper metadata, and existing relevance scores and statements. With separate search permission, an assistant can also submit queries and calculate new relevance scores using Jev, subject to usage limits. Existing read-only connections require new approval for this permission. This does not grant access to projects, private feedback notes, account email, or account changes. Developer API keys remain read-only.

For assistant connections, we store client names and callback addresses, account-linked authorization records, creation, expiry, last-used and revocation metadata, hashes of authorization-code and token identifiers, and bounded usage counters. The OAuth database does not store the raw codes or bearer tokens. Developer keys are stored as hashes with names, expiry and last-used times. Agent reads are not counted as human page visits. Expiry or revocation stops access; it does not immediately erase these operational records, which are subject to cleanup.

You can revoke an assistant connection or developer key in Settings > Connections. Revocation stops future access through that authorization, but does not erase data already received by the assistant, whose own data handling also applies.

Services We Use

To assess relevance, we send your research interests, paper-search query, or one-paper reading goal, together with paper information, to Typesafe's Jev service. A one-paper goal does not change your saved interests. Resend handles automated email, and Fastmail handles feedback emails and replies. These services process the information needed for those functions.

To write the short relevance statements shown with papers, we send the applicable interests, search query, or reading goal, the paper's title, authors and abstract, and its relevance score to OpenAI. We do not include your account identity or private project notes. Statements are cached privately for your assessment. We disable stored API responses; provider security and abuse-monitoring retention may still apply. These statements are generated summaries of the connection, not Jev's internal reasoning, and can be mistaken.

We do not currently train or fine-tune models using your private account data. We do not sell personal data.

Activity metrics stay on our servers. We do not use session recordings, advertising trackers, or email-open tracking. Raw activity events have a 90-day retention target; saved account information and earlier assessments are separate from these events.

Your Choices

Daily paper emails are opt-in, and you can unsubscribe at any time. You can edit your interests, manage public profile information, and remove saved items and projects through the site. Changing your interests does not automatically erase earlier assessments.

Please avoid entering confidential research, patient information, or other sensitive material.

For questions, feedback, or account and data-deletion requests, contact feedback@author.link.