Evidence Watch

A brief is a snapshot of what the literature said on the day it was written. Evidence Watch keeps checking after that day. Add the findings of a brief to a watchlist and BioSkepsis re-validates each one on a schedule, telling you only when a finding you have already cited stops holding.

#Why a standing watch

The screening that runs during a research run is a point-in-time check. It blocks retracted sources and flags corrections as they stand at that moment. But the literature keeps moving after your brief is filed: a paper you cited is retracted a year later, a cell line behind a key result is reported as misidentified, or a new systematic review reverses a conclusion the field had treated as settled.

Those changes are exactly the ones you would want to know about before a claim reaches a grant application, a manuscript, or a decision. They are also the ones nobody goes looking for, because re-reading the sources behind a finished brief is work that never gets scheduled. Evidence Watch schedules it.

#Setting up a watchlist

A watchlist is a named set of claims being watched on a cadence. You build one from a finished brief rather than by typing claims in by hand.

1. Open a completed brief and choose Add to Evidence Watch beneath it.
2. Name the watchlist and pick a cadence (monthly or quarterly). You can create it fresh or add the brief's findings to a watchlist you already have.
3. The brief is decomposed into individual claims. Each one is stored with the sources it cited, so later checks can go back to those exact papers. This takes a few seconds.

From then on the watchlist runs itself. You can add findings from other briefs to the same watchlist, remove individual claims, rename it, change its cadence, or pause it.

#What each check looks at

Every claim on a watchlist goes through three independent checks on each cycle. Two examine the claim's own cited sources, and one looks at what has been published since.

• Retractions and corrections. Every source behind the claim is checked against the record for a retraction, an expression of concern, or a published erratum. This check reads PubMed as the authority rather than trusting a secondary index, so a source is only called retracted when the authoritative record says so.
• Cell-line authentication. Sources are screened against reported misidentified cell lines, so a finding resting on a questioned line is surfaced rather than quietly carried forward.
• New evidence since the brief. BioSkepsis gathers work published after the brief was generated, judges whether each paper supports or contradicts the claim, and counts those positions by independent research group so one lab publishing repeatedly cannot look like a consensus. It then runs a separate check for an authoritative synthesis, a systematic review, meta-analysis, or clinical practice guideline, that contradicts the claim.

The window for “new” is anchored to when the brief was generated, not to the age of the papers it cited. A brief built on a decade of literature is not treated as having a decade of new evidence waiting for it the moment you start watching it.

#The statuses

Each cycle leaves every watched claim in one of three bands. Only the third band alerts you.

The split between the second and third band is deliberate. A handful of primary studies disagreeing is normal science, and treating it as an emergency would train you to ignore the alerts that matter. So conflicting primary literature moves a claim to Now contestedand waits there, visible in the report but silent. It takes a formal retraction, a flagged cell line, or an authoritative synthesis to actually notify you.

#How you are told

The first check after a claim is added establishes its baseline and never notifies, whatever it finds. Every later cycle is compared against that history.

• Only moves into an alert state notify. Improvements, unchanged statuses, and moves within the watching band never send anything. A claim already alerting does not re-notify every cycle unless its status changes again.
• One email per watchlist per cycle. Everything that changed in a single pass arrives together, with each finding, its new status, and why it changed.
• In-app notifications. A bell in the header collects findings as they land, with an unread count, so nothing depends on email getting through.
• Browser notifications are deliberately content-free. They tell you something changed without putting a claim or a watchlist name on a lock screen.

Evidence Watch email has its own opt-out, separate from every other kind of BioSkepsis mail, so turning it off does not affect anything else. Muting the email does not stop the re-checks: findings still appear in the app. To stop the checks themselves, switch off Scheduled re-checks on the watchlist.

#Cadence and control

Watchlists run monthly or quarterly. Shorter cadences are not offered on purpose: the things Evidence Watch is looking for, a retraction notice, a new meta-analysis, turn over on a scale of months, so a daily re-check would spend your usage re-deriving the same answer.

• Check now runs a pass immediately without waiting for the schedule. It updates the report in place and does not email you, since you are already looking at it.
• Scheduled re-checks pauses a whole watchlist when switched off. A paused watchlist is not re-checked and sends no alerts, and Check now still works.
• The report shows how many claims still hold, a breakdown by status, and a card for each claim needing attention, with the status it moved from, why it changed, and the papers that drove the change.

#Plans and usage

How many watchlists you can keep depends on your plan.

Re-checks draw on the same monthly usage as your research runs, and the Usage page breaks that spend out on its own line so background checking is never an unexplained part of your total. If you reach your monthly limit, re-checks pause rather than failing silently, and resume on their own when your usage resets.