Technique Sorts 50 New Planets From ‘Fakes’

Satellite, star and planets
Illustration of NASA’s Transiting Exoplanet Survey Satellite — TESS — observing an M dwarf star with orbiting planets.

While it might seem like astronomers discover new planets by spending hours with their eyes against telescopic viewfinders, such discoveries are actually made by analyzing data. Now, researchers have used artificial intelligence to comb this data to sort real planets from other celestial objects. Fifty new worlds made the cut.


NASA is always looking into deep space to find new planets. The TESS mission (Transiting Exoplanet Survey Satellite), for example, has just completed a survey of about 75% of the sky visible from Earth’s orbit. In 2026, the European Space Agency plans to launch its PLAnetary Transits and Oscillations of stars (PLATO) mission to add to the hunt for new planets. Missions such as these create a huge amount of data, which can take an equally huge amount of time to analyze.

What’s more, some “planets” aren’t planets at all. The issue arises due to how we spot planets here on Earth. Because we can’t see them directly, we have to rely on their transits—the points at which they pass in front of their suns. When we see the starlight dim, we know something has gotten between the sun and our view of it. While that can often be a planet, it can also be one sun passing in front of another in a binary star system or camera error.

Math to the Rescue!

To help with the massive task of analyzing data from telescopes to tease out real planets from imposters, scientists at England’s University of Warwick created an algorithm to do the heavy work. They fed both confirmed planetary data and false-positive data from the Kepler mission (which concluded in 2013) into a machine-learning algorithm to “teach” it what to look for. They then set the computer loose on unanalyzed data from Kepler, and it hit on 50 actual planets that had not been previously identified.

“In terms of planet validation, no-one has used a machine learning technique before,” said Dr. David Armstrong, from the University of Warwick Department of Physics. “Machine learning has been used for ranking planetary candidates but never in a probabilistic framework, which is what you need to validate a planet. Rather than saying which candidates are more likely to be planets, we can now say what the precise statistical likelihood is. Where there is less than a 1% chance of a candidate being a false positive, it is considered a validated planet.”

A Year as Long as a Day

Of the 50 new planets identified, some are as big as Neptune while others clock in smaller than Earth. Their orbits also vary widely—from just one day to 200 days.

As TESS continues to provide interstellar data, the researchers feel confident that their planetary validation method will save time and offer more standardization.

“We still have to spend time training the algorithm, but once that is done, it becomes much easier to apply it to future candidates,” said Armstrong. “You can also incorporate new discoveries to progressively improve it. A survey like TESS is predicted to have tens of thousands of planetary candidates, and it is ideal to be able to analyze them all consistently. Fast, automated systems like this that can take us to validated planets in fewer steps let us do that efficiently.”

The work has been accepted for publication in the peer-reviewed journal Monthly Notices of the Royal Astronomical Society. The video below provides an update about NASA’s TESS mission.