Scientists have uncovered new chemical clues that life existed on Earth more than 3.3 billion years ago—much earlier than many previous estimates.
Even more surprising, they found evidence that oxygen-producing photosynthesis may have begun nearly a billion years earlier than scientists once believed.
These discoveries were made possible through a new approach that combines advanced chemistry with artificial intelligence.
The international research team, led by the Carnegie Institution for Science, used machine learning to detect extremely faint chemical traces—what they call “whispers” of ancient life—embedded in rocks that have been heated, crushed, and altered over billions of years.
Even though the original biological molecules have long vanished, tiny chemical fragments still remain. The new AI system can recognize patterns in these fragments that point to past life.
One of the researchers involved in the study is Katie Maloney from Michigan State University. She studies early complex life and contributed rare, well-preserved fossils of one-billion-year-old seaweed found in Canada’s Yukon Territory.
These fossils are among the earliest seaweeds ever discovered and help show what ancient ecosystems may have looked like at the time.
The new findings were published in the Proceedings of the National Academy of Sciences.
The results offer an exciting new view of Earth’s earliest biosphere and also open the door to new methods for searching for life beyond our planet.
Rocks older than three billion years rarely preserve clear evidence of life. Early microorganisms left behind extremely fragile remains—such as microbial mats and tiny cells—that were later buried, compressed, heated, and fractured as Earth’s crust shifted.
Most biosignatures were believed to be erased during these transformations.
But this new research suggests that chemical fragments in ancient rocks still carry meaningful information, even if the original biological molecules no longer exist.
To uncover these hidden traces, scientists used high-resolution chemical tools to break down both organic and inorganic materials into their smallest molecular pieces. They then trained an AI model to analyze more than 400 samples, ranging from modern plants and animals to ancient fossils and even meteorites.
The AI learned to distinguish biological materials from non-biological ones with more than 90 percent accuracy. Even more impressively, it detected signs of photosynthesis in rocks at least 2.5 billion years old.
Previous techniques could only reliably identify chemical traces of life in rocks younger than about 1.7 billion years. This new method nearly doubles that window, giving scientists a much longer timeline to study early life on Earth.
Dr. Robert Hazen, one of the study’s leaders, explained that early life leaves chemical “echoes” inside rocks. With machine learning, researchers can now decode these faint signals for the first time. For Maloney, the study provides an exciting new way to understand how early photosynthetic organisms changed the planet and may also guide future missions searching for life on Mars or other worlds.
Overall, this work shows that Earth’s early biosphere left behind more evidence than scientists ever realized. Thanks to AI, these faint chemical imprints—once hidden deep inside ancient rocks—are finally coming to light.
