The things people ask before they trust it.

Responses take time because a real search is several round trips, not one. Elute reads your question, queries its databases, fetches the abstracts that came back missing one, scores every candidate for relevance, and then writes the summary for each paper it kept. Most of that runs against services it doesn't control.

PubMed in particular rate-limits multiple queries coming in at once, so those queries run one after another with a deliberate gap rather than all at once. Going faster there would mean getting blocked.

The tradeoff is deliberate: the alternative to waiting is a tool that answers instantly from memory, which is exactly how invented citations happen.

Every paper Elute shows you was actually retrieved from PubMed or OpenAlex during your search. It cannot produce a reference that doesn't exist, because it can only cite what it found. The citation list is built from the retrieved records themselves, not written out by the model.

What you should still check is the fit. Elute reads abstracts, not full papers, so it can tell you what a study is about and whether it's relevant; it cannot tell you the sample size, the incubation temperature, or what the error bars looked like. When a question needs something an abstract doesn't contain, it says so rather than guessing. Papers with no abstract anywhere are labelled as title-only instead of being quietly filled in.

Open the papers before you cite them. That is the part that is still your job, and it is the part worth doing.

Free, with no account and no card. You get five searches a day. If you leave an email address you get fifteen a day instead. We send one welcome email when you register; unsubscribe any time and your higher limit stays. Your allowance refills 24 hours after you start using it.

The daily limit keeps Elute sustainable: every search costs money to run, and Elute is an independent project paid for by its founder. Follow-up questions about papers already on screen don't count. Only new searches do.

There is no account, so there is nothing to sign up for and no password to lose. The first time you open Elute, the server gives your browser a random, signed ID, kept in local storage. That ID is what separates your sessions from everyone else's. It identifies a browser, not a person, and it isn't linked to your name or anything about you.

Your questions and the papers Elute returned are stored so your sessions are still there when you come back. Clearing your browser storage or switching browsers starts a fresh, empty history.

An email address is stored only if you volunteer one, to raise your daily limit; it is also used for one welcome email (unsubscribe any time; your higher limit stays). No names, no passwords, no payment details, and no tracking across other sites.

To delete your data, delete any chat from the sidebar, which removes its messages and saved citations. If you gave an email address, write from it to admin@eluteai.app and your data will be deleted; without one, Elute can't tell which chats are yours, so you would have to delete them yourself. The full version is on the privacy page.

Elute is built with security in mind from the start, not added afterward. We test the site for security problems before changes go live, and we mask email addresses and other personal details in our server logs. We don't publish the technical details of how Elute is protected, because that would help attackers more than it helps you. If you find a security problem, please tell us through the Contact link.