More than 10,000 possible planets have turned up in a single sweep of NASA data, after researchers used machine learning to search for signals that had not been flagged before.

In the new study, the team examined data from NASA’s Transiting Exoplanet Survey Satellite, or TESS, and identified exactly 10,091 new planet candidates.

These are still candidates, not confirmed planets. As the study notes, objects beyond our solar system are treated as candidates when they are first spotted, until there is enough evidence to confirm them. Some may turn out to be other objects or noise in the data.

According to NASA’s Exoplanet Archive, more than 6,200 exoplanets have been confirmed so far.

The researchers focused on fainter stars than TESS usually targets. The survey looked at stars 16 times fainter than those typically observed by the mission.

Using machine learning, the team surveyed more than 83 million stars observed during TESS’s first year. Of those, 10,091 appeared to host previously unseen, transiting planet-like objects.

TESS searches for planets by watching for transits, when a planet passes in front of its star from the spacecraft’s point of view. That passage makes the star appear dimmer, and TESS measures that change in brightness.

The team has already confirmed one of the candidates, a planet called TIC 183374187 b.

The study says the planet appears to be a hot Jupiter, a gas giant orbiting very close to its host star, with a mass similar to Jupiter’s.

Lead author Joshua Roth, a graduate researcher at Princeton University, told IFLScience the team plans a follow-up study using TESS’s second year of data.

The spike in candidate detections comes about 30 years after the first confirmed exoplanet, 51 Pegasi b, was found in 1995.

The study was published on April 28 in The Astrophysical Journal.

Read more from Space.