WASHINGTON, Feb 6, 2026, 12:19 (EST)
- Researchers revealed a new artificial-intelligence technique designed to pinpoint which dinosaurs made specific fossil footprints.
- From 1,974 footprint silhouettes covering roughly 150 million years, the system extracted eight recurring shape traits.
- Scientists say this method could add much-needed consistency to a field where identifying the “trackmaker” often depends on subjective judgment.
Scientists have created an AI technique designed to pinpoint which dinosaur made a specific fossil footprint, addressing a persistent challenge in paleontology. Their study, featured in the Proceedings of the National Academy of Sciences, breaks down tracks into eight consistent traits for comparison across different sites. Reuters
Timing is key since footprints rank among the most common dinosaur fossils, frequently appearing without any bones or teeth close by. But depending on the animal’s activity and the substrate, the same creature might leave very different tracks. That variability sparks debate among researchers trying to decide if a print came from a carnivore, herbivore, or something bird-like.
That uncertainty matters a lot. Tracks help researchers determine which groups coexisted in an ecosystem and, sometimes, challenge claims that certain groups showed up earlier than the body fossil record indicates.
“This matters because it offers an objective method to classify and compare tracks, cutting down on subjective human judgment,” said physicist Gregor Hartmann from the Helmholtz-Zentrum Berlin research center in Germany, who spearheaded the study.
“Linking a track to its trackmaker is notoriously difficult, and paleontologists have been debating this for generations,” said Steve Brusatte, a paleontologist at the University of Edinburgh and senior author of the study.
The method was refined by applying an algorithm to 1,974 footprint silhouettes spanning about 150 million years of dinosaur evolution. The AI identified eight key features that most clearly account for variations in track shapes, such as toe spread, toe attachment to the foot, and heel placement.
The paper’s authors used “unsupervised” machine learning—software that finds patterns without preset labels—to cut down on bias from pre-labeled training data. By mapping tracks into a kind of shape space, they achieved 80% to 93% agreement with expert identifications from previous studies. They also referenced earlier supervised machine learning methods that grouped tracks into broad categories. Pnas Nih
The researchers also developed an app called DinoTracker and made the code available, allowing others to apply the method to fossil shapes altered by geological processes and time.
One standout case featured seven tiny, three-toed footprints from South Africa, dating back roughly 210 million years. The algorithm backed up earlier claims that these tracks look like bird prints, despite being much older than the earliest avian fossils. That’s especially notable since birds evolved from small, two-legged dinosaurs.
“This, of course, doesn’t prove they were made by birds,” Brusatte said. He added the tracks might belong to unknown dinosaurs that came before birds — or to unrelated dinosaurs with feet that looked similar. “We have to take this seriously and figure out what’s going on,” he said.
Hartmann warned that a footprint isn’t always straightforward—a trackmaker’s identity is “inherently uncertain.” The shape of a track shifts with behavior like walking, running, jumping, or swimming, plus factors like moisture, sediment, burial, and erosion. So, a single dinosaur might leave prints that look wildly different from each other.
Brusatte pointed out that it’s uncommon for tracks to come paired with a body at the trail’s end. Often, prints appear on their own, leaving researchers piecing together clues to match a footprint with a dinosaur foot. He said the new tool aims to bring more consistency to that detective work, not to wipe out uncertainty entirely.
The researchers describe their method as a tool to analyze tracks more consistently, not to replace traditional fieldwork or expert anatomical analysis. The core issue persists: the fossil record holds many more footprints than actual fossils of the creatures that left them. Indianexpress
