AI Discovers 118 Hidden Planets in NASA Data

Scientists just unlocked a treasure trove of hidden worlds using an AI detective that can spot planets humans might miss.

An international team of astronomers developed RAVEN, an artificial intelligence system that scoured four years of data from NASA's Transiting Exoplanet Survey Satellite. The AI analyzed observations of 2.2 million stars and successfully validated 118 new planets orbiting distant suns.

Dr. Marina Lafarga Magro, who led the research at the University of Warwick in the United Kingdom, says this represents one of the best-characterized samples of planets near Earth. The discovery includes 31 completely new planets and over 2,000 additional high-quality candidates waiting for confirmation.

Some of these worlds challenge what scientists thought was possible. The team found ultra-fast planets that complete entire orbits in less than 24 hours, racing around their stars faster than Earth rotates once. They also discovered planets in the "Neptunian desert," a region where theoretical models predict planets shouldn't exist.

The real magic lies in how RAVEN works. Traditional methods struggle to tell whether a dimming star means a planet passed in front of it or if something else caused the shadow, like two stars eclipsing each other. Dr. Andreas Hadjigeorghiou, who developed the AI, trained RAVEN on a catalog of simulated planets and lookalike phenomena until it learned to spot the difference.

Professor David Armstrong, a co-author of the study, explains that RAVEN can analyze massive volumes of astronomical data systematically and objectively. The technology reduces uncertainty in calculations up to ten times compared to previous studies, making planet hunting far more reliable.

The Ripple Effect

The team didn't keep their tools to themselves. They released public catalogs and interactive software so astronomers worldwide can select targets for deeper study using ground-based telescopes and upcoming space missions like the European Space Agency's PLATO satellite.

The statistical findings paint a clearer picture of our galactic neighborhood. Between 9 and 10 percent of sun-like stars host planets in short orbits, confirming what NASA's earlier Kepler mission suggested. Meanwhile, only 0.08 percent of solar-type stars host the rare Neptunian desert planets.

This AI-powered approach transforms planet hunting from painstaking manual review into rapid, accurate discovery, accelerating our understanding of how common planetary systems are throughout the galaxy.