Import from ORCID or BibTeX, pick a few representative papers, and state your hypotheses, methods, and boundaries. You confirm every word.
Every week PaperCurrent searches multiple sources, deduplicates, and evaluates each candidate against your profile — with a stricter second pass for possible challenges.
At most five hits per digest, each with a score, a one-sentence why, and a verbatim quote from the abstract. Quiet weeks are verified, not empty.
Every claim in an assessment quotes the title or abstract verbatim — and the quote is verified programmatically. No fragment, no hit.
A “possible challenge” to one of your positions survives only after a second, stricter model pass. Suppressed borderline cases stay visible as weak signals.
An empty week shows the work: how many candidates were checked and that none crossed your threshold. A quiet digest is a verified quiet week.
Plenty of tools tell you what's new, or recommend what's like what you've read. None of them check a new paper against the hypotheses you're actually testing.
| What it gives you | What PaperCurrent adds | |
|---|---|---|
| Scholar alerts | New matches for keywords and authors — every week, regardless of relevance. | An explicit check against your hypotheses and methods, scored and evidenced, capped at five. |
| Personalized digests Scholar Inbox, R Discovery |
A weekly feed of papers like the ones you've read and rated — ranked by interest. | Judged against the hypotheses you stated, not inferred from clicks — with the reason and the evidence shown. |
| AI research assistants Elicit, Undermind |
Alerts on new papers matching a saved research question, with a relevance score. | A versioned profile of several positions and boundaries, a verified challenge flag, calibrated confidence, and a verbatim evidence quote. |
| scite | Supporting and contrasting citations — after the field has cited a paper. | A possible challenge to your position at publication time — no citation needed, before anyone has reacted. |
| A scheduled LLM prompt | A plausible-sounding weekly summary with no memory and no guarantees. | A versioned profile, reproducible evaluations with pinned model and prompt versions, cross-source deduplication, and documented precision. |
OpenAlex, Crossref, Europe PMC, and Unpaywall — openly documented, with per-source provenance stored for every paper.
Every evaluation pins the profile version, criteria version, prompt version, and model version it was made with. Nothing is rescored silently.
Judgements rest on titles and abstracts, never on full texts — and each card shows which. Missing abstracts cap the confidence label.
Free feeds recommend. This one evaluates.
One profile, one monitor, a monthly digest — judged against your hypotheses, not your reading history.
Your whole research program, watched weekly.
Five monitors, weekly digests, custom criteria, exports.
Shared monitors for a small group.
Up to five people, shared feedback. After Pro proves itself.