AI Just Found 118 Hidden Worlds in NASA’s Data

EARTH, Laniakea Supercluster––What exactly lies out there in space has been a source of debate for as long as humans have looked up. Don’t even get us started on the one about what constitutes a “planet.”

Flash back: 2006, Pluto, Neil deGrasse Tyson. Spoiler, our solar system is now one giant world short of a cosmic baseball batting order.

But don’t worry, this news might not sting as much as Pluto's demotion to a dwarf planet. University of Warwick astronomers discovered more than 100 planets using a powerful artificial intelligence (AI) system, as reported May 2026 in Monthly Notices of the Royal Astronomical Society.

The breakthrough came after researchers developed an AI pipeline called RAVEN (RAnking and Validation of ExoplaNets), designed to sift through the enormous volumes of data collected by NASA's Transiting Exoplanet Survey Satellite (TESS). By analyzing observations from more than 2.2 million stars, the system successfully confirmed 118 exoplanets (planets outside of Earth’s solar system), as well as identified some 2,000 additional candidates.

Related

Humanity Just Found the Milky Way’s Edge And Earth’s Closer Than We Thought

By tracking ancient giant stars, astronomers have traced a fuzzy, long‑hidden boundary where the Milky Way’s star‑forming disc ends that is only tens of thousands of light‑years beyond our own Sun.

“These candidates and planets took several months of work from all our team, so we felt very rewarded when we started to get preliminary results showing a significant number of candidates and planets,” first author Dr. Marina Lafarga Magro, postdoctoral researcher at the University of Warwick, told Milky Way News in an Email.

“Still, it took a lot of verification and testing of our code to make sure that what we were sharing with the community were indeed planets and candidates that we are convinced have a high probability of being true planets.”

Among the newly confirmed worlds are some of the most unusual planets ever found. Several belong to a rare class known as ultra-short-period planets, which orbit their host stars in less than a single Earth day. Others occupy the enigmatic “Neptunian Desert”: a region of space where planets of Neptune-like size are unexpectedly scarce.

These discoveries are significant not only for the planets themselves, but because they challenge existing theories about how planetary systems form and evolve. Finding planets in places where astronomers expected few or none to exist provides valuable clues about the forces that shape planetary migration and survival.

“Now we have been able to quantify how ‘empty’ this desert is using our sample of candidates,” said Lafarga. “The fact that this desert exists has strong implications for how planets form and evolve, so future theoretical work will be able to use our sample to better constrain models of planet formation and evolution.”

The key to RAVEN's success lies in its ability to distinguish genuine planetary signals from false alarms. Space telescopes detect planets by observing tiny dips in a star’s brightness when a planet passes in front of it. However, many other astronomical phenomena, including binary star systems and instrumental noise, can mimic these signals.

To overcome this challenge, researchers trained RAVEN on hundreds of thousands of realistic simulations of both planets and false positives. The AI learned to recognize subtle patterns that indicate whether a signal is likely to be a true planet. Lafarga added that the automatic coding without human input avoids mistakes that may be introduced by human biases.

Beyond discovering new worlds, the system is also helping scientists answer larger questions about the galaxy’s planetary population. Using the validated dataset, researchers found that roughly one in 10 Sun-like stars hosts a planet orbiting close to its star. The analysis also provided the first precise measurement of the rarity of Neptunian Desert planets, revealing that they occur around only 0.08 percent of Sun-like stars.

The findings demonstrate how AI is transforming modern astronomy. As telescopes generate increasingly massive datasets, systems like RAVEN can analyze information at a scale that would be impossible for humans alone. By automating planet detection while maintaining scientific rigor, these tools are giving astronomers more precision than ever before.

Related

27 Tatooines: Astronomers Spot New Worlds With Double Suns in Our Own Galaxy

Luke Skywalker’s home planet may be fictional, but Earth's scientists recently identified several planets in the Milky Way galaxy with Tatooine-like sunsets.

“Astronomy is moving towards large surveys that are generating huge amounts of data, which need to be somehow analysed automatically,” said Lafarga. “For this, we will rely more and more on technological improvements, so technology is going to play a key role in the near future.”

So artificial intelligence may in fact become one of astronomy's most important instruments, sifting through data no human team could ever hope to read alone. Which means the ongoing boxing match over what counts as a planet just got 118 rounds longer.