Using NASA data and machine learning, scientists have found over 10,000 possible new planets in a single survey. In a new study, researchers applied a machine-learning approach to perform a sweeping search through observations from NASA’s Transiting Exoplanet Survey Satellite (TESS). By reprocessing this archive, they uncovered exactly 10,091 candidate planets that had never been seen before in prior catalogs, a haul that immediately dwarfs the number of worlds currently confirmed.
The work is detailed in the 2026 paper “The T16 Planet Hunt: 10,000 New Planet Candidates from TESS Cycle 1 and the Confirmation of a Hot Jupiter around TIC 183374187*.” The T16 project reports a uniformly detrended, systematics-corrected set of TESS Cycle 1 light curves and then mines those data with machine learning to flag transit-like signals. According to the study, this single pass through the mission’s early full-frame images is sufficient to surface 10,091 new candidates, indicating that a substantial population of potential planets remained hidden in the original processing.
These 10,091 objects are classified as exoplanet “candidates,” not confirmed planets. In exoplanet surveys, a candidate is a signal consistent with a planet but still lacking the follow-up evidence needed to rule out false positives. Some of the newly reported signals may ultimately be attributed to other astrophysical sources or to noise in the TESS data rather than bona fide planets. It would appear we now have 10,091 candidate exoplanets to go through and confirm, implying a long tail of follow-up work with radial-velocity measurements, additional photometry, or other validation techniques before any revised planet count can be established.
The scale of the candidate list is notable against the current census. To date, humanity has discovered over 6,200 confirmed exoplanets, according to NASA’s Exoplanet Archive. If even a modest fraction of the 10,091 candidates survive detailed scrutiny, the confirmed catalog could expand significantly. The T16 paper’s subtitle notes the “Confirmation of a Hot Jupiter around TIC 183374187*,” indicating that at least one object emerging from this pipeline has already been promoted from candidate to confirmed planet, but the broader sample remains unvetted.
The survey’s yield also addresses why so many potential planets were missed in earlier TESS analyses. TESS has been operational since 2018 and has continued on an extended mission since 2020, and its standard pipelines have focused on stars where transit signals are easiest to detect. TESS identifies exoplanets by measuring the dimming of a star as a planet transits across its disk, and planets orbiting brighter stars are easier to spot because their transit signatures stand out more clearly from instrumental and astrophysical noise. The new study specifically pulled data from fainter stars, where individual transits are harder to distinguish and where conventional search methods are more likely to overlook shallow or noisy events.
By combining a large, systematics-corrected light-curve set with machine-learning classifiers tuned to transit-like features, the T16 project effectively pushes TESS deeper into this fainter-star regime. That strategy trades a higher false-positive rate for completeness, which is why the authors emphasize the candidate status of the 10,091 signals. The result is a catalog that dramatically broadens the pool of potential exoplanets derived from TESS Cycle 1 alone and sets up a multi-year effort to determine which of these machine-flagged events correspond to real planets orbiting distant stars.
Original source: https://www.space.com/science/astrophysics/scientists-found-10-000-possible-exoplanets-hiding-in-nasa-data