Anyone can now join the search for new worlds, thanks to a powerful AI tool NASA scientists have opened to the public.
For years, NASA’s planet-hunting missions have been steadily expanding our catalog of exoplanets—planets that orbit stars beyond our solar system. Together, the retired Kepler Space Telescope and the still-operating Transiting Exoplanet Survey Satellite (TESS) have played a major role in confirming more than 3,000 exoplanets. Kepler focused on closely monitoring a relatively small patch of the sky for long periods, while TESS scans nearly the entire sky, looking for planetary candidates around many more stars.
The challenge is that space telescopes don’t “see” planets directly in most cases. Instead, they watch stars for tiny, repeating dips in brightness. That dip can happen when a planet crosses in front of its star from our viewpoint—an event known as a transit. But not every dip is a planet. Stellar activity and other astronomical phenomena can mimic the same kind of signal, which makes sorting real exoplanets from false alarms a time-consuming process.
That’s where NASA’s ExoMiner project comes in. In 2021, researchers introduced an AI-driven tool called ExoMiner that successfully helped confirm 370 new exoplanets using Kepler data. Now the team has taken the next step with an upgraded model named ExoMiner++. Trained using data from both Kepler and TESS, ExoMiner++ is built to evaluate TESS observations and predict which dimming events are most likely caused by an exoplanet transit rather than something else.
On its very first run, this improved AI model flagged around 7,000 possible exoplanets from TESS data—an enormous pool of candidates that can give researchers (and citizen scientists) a head start in identifying promising targets for follow-up study.
In a move that could speed up discoveries even further, ExoMiner++ has been made freely available to the public through GitHub. With open access to the tool, more people can participate in exoplanet research, explore real space telescope data, and help narrow down which candidates deserve more detailed analysis.
NASA scientists are already looking beyond the current release. A future version of ExoMiner++ is expected to go a step further by working directly with raw telescope data, rather than relying on pre-identified transit signals. That kind of capability could make the system even more effective for upcoming space observatories and large-scale surveys, including future missions such as the Nancy Grace Roman Space Telescope.
The ExoMiner++ work and results were detailed in a study published in The Astronomical Journal, highlighting how artificial intelligence is becoming a key driver in modern exoplanet discovery—and how the hunt for new worlds is increasingly something anyone can take part in.
