Scientists have developed a new artificial intelligence tool called DinoTracker that helps identify which dinosaurs made ancient footprints by analyzing their shapes. Unlike earlier systems that learned from pre-labeled data — which could introduce errors — this AI was trained on thousands of unlabeled prints and learned to recognize meaningful variations on its own. By focusing on features such as toe spread, heel position, and ground contact area, the system can group and compare footprints in a way that closely mirrors expert classification.
The AI model behind DinoTracker was trained using nearly 2,000 real fossil footprints along with millions of simulated variations that mimic natural distortions caused by sediment, compression, and erosion. By identifying eight key shape characteristics, the system can compare new footprints with known examples and determine which dinosaur group likely made them. When tested, its results agreed with human expert classifications around 90% of the time, even in challenging or disputed cases.
One of the intriguing findings from the system’s analyses involves very old tracks that show bird-like features despite being far older than the earliest known bird fossils. This has raised questions about whether birds evolved earlier than currently believed or whether some dinosaurs simply had very bird-like feet. The AI also provided fresh insight into mysterious footprints from Scotland’s Isle of Skye, suggesting they may have been left by some of the earliest relatives of duck-billed dinosaurs.
Beyond scientific research, DinoTracker has been made into a public app that allows anyone to upload a footprint photo or sketch and receive an instant analysis. This broad accessibility could help both professionals and enthusiasts explore fossil evidence more easily, while supporting paleontological study and public engagement with ancient life on Earth.
