The same rubric that gates our builds, applied to a public corpus.
Every quarter, we score a public corpus of AI publications against the same credibility scanner that gates every build of this publication. The rubric is open. The thresholds are pre- declared. Right of reply is built into the publishing pipeline.
The rubric is live and reproducible. The first scored quarter ships only after the corpus selection is publicly announced (two-week dispute window) and the right-of-reply notices have been mailed. We don't ambush publications with this; we tell them what's being measured and when.
Read the methodology now — /slop-index/methodology/. Subscribe to the publication newsletter to be notified when the first corpus is announced.
What this is, and what it isn't
The Slop Index measures text-pattern signals that correlate with low-rigor AI-generated content. A publication can score high on pattern density and still be excellent. A publication can score low and still publish slop. The score is one signal among many, and we publish the rubric and raw scrapes so any reader can dispute it.
Status grammar matches the rest of this publication. Where a dimension or aggregate falls below the threshold, we say Not holding. We do not say "broken", "failing", or "garbage". The grammar is applied to the dimension being measured, not to the publication's overall quality — which we do not claim to measure.
Process at a glance
- Pre-declare corpus + selection criteria. Two-week dispute window.
- Snapshot each publication's public articles in the prior 90 days.
- Run the public credibility scanner; manually review every flag.
- Score each dimension; derive the aggregate.
- Issue 14-day right-of-reply notices with proposed scores + raw scan.
- Publish. Per-publication pages render any response in full.
Full detail at /slop-index/methodology/.
We hold ourselves to the same rubric. Every build of this publication runs the same credibility scanner before any push to production — see /standards/ for the build-time enforcement and /holding/ for the public claim ledger that catches what the scanner doesn't.