AI helps paleontologists identify dinosaurs from their fossilized footprints

Paleontology and Artificial Intelligence: A New Perspective on Dinosaur Tracks

Paleontologists study not only skeletons but also “living” impressions—fossilized traces of the daily activities of prehistoric animals. However, recognizing such tracks can be difficult: they are often blurred, found in groups, or covered by sedimentary rocks, making precise identification of the trackmaker almost impossible.

German researchers solved this problem with artificial intelligence. They uploaded nearly 2,000 images of dinosaur tracks covering 150 million years of Earth’s history and trained an algorithm to highlight eight key characteristics of the impressions. These parameters allow not only the species to be identified but also the details of its movement to be reconstructed.

What exactly did the AI consider?
Feature Description
Overall weight load on the track How much weight the dinosaur exerted with each step
Shape and contact area How the limb touched the surface
Body placement during movement Balancing and mass distribution
Distance between toes Wide foot, important for species identification
Attachment of toes to the foot Structure of the “heel” and its influence on the track
Heel position and pressure level How much the rear part of the leg was pressed
Balance of pressure fore‑to‑hind Prioritization of weight distribution front vs. back
Differences between left and right sides Comparison of right and left track shapes

Why is this important?
Before AI, most tracks could only be identified with a high degree of uncertainty. The classic approach considered many variables: type of movement, soil moisture, its structure, long‑term coverage by sedimentary rocks, and subsequent erosion. All of this creates a complex “detective” task.

With the new algorithm, scientists obtained the “Cinderella’s technological shoe”—a tool that allows more accurate matching of tracks to specific dinosaur species. This is especially valuable because the skeleton of a particular animal is rarely found near its footprints.

Practical results
One striking example is the tracks of a three‑legged dinosaur found in South Africa and dated to about 210 million years ago. Previously it was thought that ancestors of modern birds appeared only about 60 million years later. AI confirmed the hypothesis that these small tridactyls could indeed have been early relatives of feathered animals, thereby revising the timeline for the evolution of avian traits.

Thus, artificial intelligence not only speeds up track analysis—it opens new horizons in understanding the behavior and biology of prehistoric creatures.