Scientists have unveiled a groundbreaking technique that combines advanced chemical analysis with artificial intelligence to detect signs of ancient life in rocks as old as 3.3 billion years. The method, which has been successfully tested on over 400 samples ranging from ancient sediments and fossils to meteorites, boasts an accuracy rate exceeding 90 per cent in distinguishing biological from non-biological material.
The research, published in the Proceedings of the National Academy of Sciences, uses pyrolysis–gas chromatography–mass spectrometry to extract molecular fragments from samples, which are then analysed by a machine-learning algorithm. Unlike traditional methods that search for specific molecules like DNA or lipids, this AI model identifies broader chemical patterns that suggest biological origins—even in highly degraded samples.
Dr Michael Wong, an astrobiologist and co-author of the study, described the technique as a “powerful new tool for astrobiology,” which could transform the way we search for life on Earth and beyond. Dr Robert Hazen, another co-author, called it a “paradigm shift,” noting that the model’s ability to detect general distribution functions allows it to assess a wider range of samples, including those with minimal surviving biological material.
Notably, the team uncovered biosignatures dating back 3.3 billion years, significantly extending the known record of ancient life. They also found molecular evidence of oxygen-producing photosynthesis from 2.5 billion years ago—over 800 million years earlier than previously confirmed.
This new approach sidesteps the limitations of traditional fossil and isotope analysis, which require rare, well-preserved samples. The AI model can even differentiate between types of organisms, such as photosynthetic versus non-photosynthetic life, and between eukaryotic and prokaryotic cells.
The implications for planetary science are significant. If returning Martian samples proves too costly, the researchers suggest deploying rovers equipped with this technology directly on Mars. NASA has already provided funding to develop such an instrument.
While the model is not yet a definitive test for life, it offers a complementary and more versatile tool for future exploration. As co-author Prof Andrew Knoll noted, the technique doesn’t just improve our instruments—it fundamentally changes the questions we can ask about Earth’s deep past and the potential for life elsewhere in the universe.
