We score for clickbait today. AI slop is next.
Honestly? The score's pretty naive right now. It pattern-matches on length, sources, link-richness, and a handful of clickbait and buzzword-stuffing signals, which catches a fair bit of obvious slop but misses the subtler stuff. It also can't really tell if a piece was written by AI, and that's something we'd like to help solve. Doing it properly means training a real model, building evals for it, and paying for the infrastructure to run all of that, which gets expensive fast.
If you care about the open web staying readable, help us build it. Sponsor the work, refer someone who would, or send an idea for how the bagelometer should grade AI content.
We're a small team. Every introduction or contribution moves this work forward.
The five tiers
The number maps to a tier so you can read it at a glance:
- Golden Brown (80–100): properly toasted. Long, well-sourced, link-rich.
- Toasty (60–79): a solid read.
- Doughy (40–59): edible, a bit underdone. Often shorter pieces or news briefs.
- Stale (20–39): thin on substance.
- Burnt (under 20): probably skip it.
What goes into the score?
Three signals from the article itself, weighted roughly like this:
- Reading time (about half the score). Longer pieces tend to do more than skim a topic, so they earn more.
- Outbound links (about a third). Articles that cite sources, link to related work, or send you elsewhere score higher than ones that link nowhere.
- Word count (the rest). A backstop for cases where reading time is missing or off.
Articles that are both substantial and well-sourced get a small bonus on top. Very short pieces with no outbound links get a small penalty.
What it doesn't measure
The bagel score is a shape-of-the-article check, not a verdict on the writing. It can't tell you whether a piece is original, accurate, funny, or worth your time in the way a human reader would. A long, link-rich article can still be filler. A 400-word post can still be brilliant.
Use the badge as a quick filter, not as a recommendation.
Does the badge decide what shows up on the homepage?
No. The homepage and the "hot off the press" rail use a separate quality signal that combines a closer read of the writing with how recently it was published. The badge tells you about the article in front of you. Where it ranks is a different question, covered in What's popular.
Why doesn't this article have a badge?
If we couldn't fetch the full article (paywall, login wall, broken page) we don't have enough signal to score it, so the badge is hidden. The article still appears in feeds and search, just without the score.