The Gut Health Index
What 185,000+ analyzed stools actually look like
Every medical site tells you what healthy stool should look like. None of them can tell you what it actually looks like across a real population. PoopCheck has analyzed 185,000+. This page breaks down what they look like — every Bristol type, color, consistency and flag, aggregated and published in full.
Updated 2026-09-04 · Aggregate counts only — no individual record is published
Explore by Bristol type
Select a type to see how every other measurement changes. Each bar's height is that type's share of every stool analyzed, scaled against the most common type.
Color
Consistency
Shape
Quantity
Where does yours fall?
Pick the shape closest to your last one and see how it sits against the 185,000+ analyzed stools.
Not sure which one is yours? Take a photo and the same AI that classified all 185,000+ stools will place yours on the scale — right here, in seconds.
Your photo reads as .
Placed on the scale above. The photo was analyzed and not stored — the app keeps your history so you can see which way yours is trending.
The data rediscovers the Bristol scale on its own
The AI classifies a stool's type and judges its health independently. Plot one against the other and a U-curve appears: type 4 is flagged least often, at 42.6%, with the rate climbing toward both extremes. Nobody told the model that type 4 is the ideal — the 185,000+ analyzed stools say so.
How this is measured
Source
185,000+ stools submitted to PoopCheck and classified by its image model, each carrying a Bristol type, color, consistency, shape, quantity, blood and mucus level, and buoyancy. Figures are given as shares of the analyzed set.
Privacy
Only aggregate counts are published. No image, account identifier, or individual record leaves the database, and any bucket under 50 records is suppressed so nothing here is re-identifiable.
Limits
This is a self-selected app population, not a clinical sample — people log when something seems off, which skews it away from the general public. Health flags are the model's judgment, not a diagnosis. No time-based trends are shown, because record timestamps are backfilled and would not reflect real logging behavior.
One stool is a data point
The pattern is what matters. PoopCheck logs every scan against this scale, so you see which way yours is trending rather than guessing from a single day.
Start trackingLicense the underlying data
The full record-level dataset — 185,000+ labelled stool images with structured annotations — is available for research and commercial ML under licence.
Dataset licensing