How AnteSum works, exactly

The front page makes some confident claims. This page substantiates them: which journals we monitor, what we fetch and what we exclude, what the AI reads and writes, and what AnteSum deliberately does not do. (The exact recipes — the prompts, the rules, the tuning — we keep to ourselves.)

Which journals we cover

Each specialty edition monitors a curated list of journals for that field. Right now that is 61 anaesthesia and pain journals and 52 orthopaedic journals. The full lists are below; they are revised as journals launch, merge or fold. If a journal you rely on is missing, tell us at hello@antesum.com.

What we fetch, and what we exclude

Once a week, AnteSum checks the official records that publishers register for everything they publish, for each journal on the list, and takes every recently published item registered as a journal article. In practice that means:

• Included: original research, reviews, guidelines, editorials, case reports and correspondence, whether published online-first or in an issue. Each is labelled by section so you can tell an editorial from a trial at a glance.
• Excluded: journal front matter (contents pages, editorial boards, cover images, indexes), items the publisher registers as something other than an article, and anything the publisher has not registered at all.
• Not covered: preprints. AnteSum only monitors peer-reviewed journals. Supplement content appears only when the publisher registers it under the journal as a normal article.

So when the front page says AnteSum gathers "the new papers", the precise claim is: every item registered as a journal article by these journals' publishers, minus front matter. Our coverage is only as complete as the publishers' own records, which for these journals are very good but not something we can independently guarantee.

Timing, dates and months

Publishers usually register a paper when it goes online, which for many journals is weeks or months before it is assigned to a print issue. That is the basis of the "often weeks before print" claim: we pick papers up from their online registration, and our fetch runs weekly, so a new paper typically appears in AnteSum within days of publication.

A paper belongs to the month AnteSum first saw it, permanently. Publisher dates (online date, issue date) are shown for reference but never move a paper between months. This keeps every monthly archive stable: what you saw in a month's digest is what stays there.

Duplicates

Every paper is identified by its DOI. A DOI enters AnteSum once; if the same paper turns up again in a later fetch (for example when it moves from online-first into an issue), it is recognised and left where it was, in its original month with its original classification.

How papers are classified

Two separate steps, and only one of them uses AI:

• Section (Reviews, Original Articles, Guidelines, Editorials, Case Reports, Correspondence): assigned automatically from how the journal itself presents the article. No AI involved.
• Subspecialty (e.g. obstetric, regional, paediatric): assigned by AI, within guardrails we set for each journal.

Classification happens once, when a paper is first seen, and does not silently change afterwards. Like any classifier it will occasionally mislabel a paper; corrections are applied deliberately, not by re-running the AI over old months.

What the AI reads, and what it writes

The monthly digest for each subspecialty is written by a frontier AI model and refreshed weekly as new papers land. Its input is the month's papers: their titles, journal names and section labels. It does not read full text.

Where an abstract is available, the model sees only what the study set out to do — never its results or conclusions, which are removed before it reads anything. What a study found is always yours to read, and abstract availability has no effect on a paper's inclusion or prominence in the digest.

Is there human review?

No. AnteSum is built and run by a practising doctor, but each weekly digest update is generated and published automatically; a clinician does not approve it first. That is exactly why the digest is constrained the way it is: because no one signs off each edition, it is only allowed to describe what has been published, never to report results or draw clinical conclusions. If you spot an error, a mislabelled paper or a badly framed digest entry, please tell us at hello@antesum.com and we will fix it.

What AnteSum does not claim

• It does not assess study quality, risk of bias or strength of evidence. Inclusion in the feed or the digest is not an endorsement of a paper's methods or findings.
• It does not give clinical advice, and nothing in it should change your practice on its own. Read the paper; apply your judgement.
• It is not a systematic review or a complete record of the literature. It covers the journals listed above, as registered by their publishers.
• It does not replace reading. It exists to make sure the papers worth your time actually reach you.