A machine-learning system built to hunt for overlooked planets has pulled 118 new worlds out of NASA’s existing telescope data, including a handful that complete a full orbit around their star in less than 24 hours. The headline figure of 100 confirmed exoplanets circulating in early coverage reflects a round-number summary; the actual validated count in the underlying study is 118. The discovery, led by astronomer Michelle Kunimoto at MIT and detailed in a preprint posted in March 2026, represents one of the largest single batches of confirmed planets ever extracted from the Transiting Exoplanet Survey Satellite (TESS). It also suggests that traditional planet-hunting methods have been leaving a significant number of worlds on the table.
What RAVEN found
The pipeline behind the haul is called RAVEN, short for RAnking and Validation of ExoplaNets. Kunimoto’s team pointed it at full-frame images from TESS sectors 1 through 55, covering roughly 2.2 million Sun-like stars. RAVEN works in two stages: first, it uses a Box Least Squares algorithm to flag periodic dips in starlight that could signal a planet crossing in front of its host star. Then it runs each candidate through a Bayesian validation step, weighing the probability of a genuine planet against false-positive explanations like eclipsing binary stars or camera glitches.
That two-stage design allowed the team to move from raw light curves to statistically validated planets without manually reviewing every signal. Of the 118 confirmed worlds, 31 were newly validated planets that had not been flagged as candidates or statistically confirmed in any prior TESS or Kepler planet search; they may have produced detectable signals in the raw data, but no earlier pipeline had promoted them to candidate or validated status. The pipeline also vetted more than 2,000 existing candidates and flagged roughly 1,000 new ones for future follow-up. Every confirmed planet orbits its star with a period between about 0.5 and 16 days, placing them all far closer to their stars than Mercury is to the Sun.
To put the number in perspective: TESS has confirmed roughly 500 exoplanets since its 2018 launch, according to NASA’s Exoplanet Archive. Adding 118 in a single study is a substantial jump, and it hints at how much undiscovered material still sits in the satellite’s growing archive of full-frame images.
One in ten Sun-like stars hosts a close-in planet
A companion study, published in Monthly Notices of the Royal Astronomical Society, used the RAVEN-processed sample to answer a broader question: how common are short-period planets around stars like our Sun? Drawing on the same detections and applying corrections for planets TESS would have missed due to orbital geometry or faint signals, the authors arrived at an occurrence rate of roughly 9 to 10 percent for planets with radii between about 1 and 10 times Earth’s and orbital periods under 16 days, as detailed in the published analysis.
In plain terms, about one in every ten Sun-like stars hosts at least one planet hugging it in a tight orbit. The study also mapped how occurrence rates shift across different combinations of planet size and orbital period, producing a structured grid that planetary-formation theorists can test their models against. A public preprint of the demographics paper lays out the completeness corrections in detail.
What remains uncertain
Several gaps are worth flagging. The preprint reports that some validated planets have orbital periods near 0.5 days, but it does not break out exactly how many of the 118 fall into the ultra-short-period category. Secondary coverage has highlighted these extreme cases, yet without a dedicated table listing individual parameters, it is hard to assess how they compare to previously known ultra-short-period planets from Kepler or ground-based surveys.
RAVEN’s validation is statistical, not observational. It estimates how likely a signal is to be a real planet rather than an impostor. That approach is widely accepted, but radial-velocity measurements or high-resolution imaging can still overturn a small fraction of statistically validated planets. The preprint does not provide a comprehensive list of independent follow-up observations for the 118 worlds, so some may be reclassified as further data come in.
NASA has not announced whether the RAVEN-validated planets will be folded into the agency’s official TESS Objects of Interest catalog. The agency’s own AI tool, ExoMiner, has also been used to validate large batches of planets from Kepler data and has been adapted for TESS candidate vetting. How the two pipelines compare in precision on TESS data specifically has not been published in a head-to-head analysis, so cross-comparisons remain speculative for now.
There are also questions about how representative the surveyed stars are. TESS observes bright, nearby stars spanning a range of ages and chemical compositions. The demographics study restricts its sample to main-sequence FGK-type stars, but selection effects tied to stellar brightness, sky position, and crowding may bias the results. Until similar analyses are repeated on complementary datasets or extended to fainter stars, it is uncertain whether the 9 to 10 percent occurrence rate holds across different stellar environments in the Milky Way.
Scorching worlds, not Earth twins
None of these planets are candidates for habitability. With orbital periods maxing out at 16 days and some completing a lap in roughly half a day, these are scorching worlds blasted by stellar radiation and likely tidally locked, with one face permanently aimed at their star. They are not Earth analogs, and the researchers do not frame them as such.
The real payoff is statistical. By establishing that roughly one in ten Sun-like stars hosts a short-period planet, the RAVEN results give theorists a sharper benchmark for modeling how planets migrate inward after forming in the cooler outer reaches of a protoplanetary disk, and how stellar radiation and tidal forces sculpt the final layout of a planetary system.
Why automated pipelines are reshaping the exoplanet census
RAVEN fits into a broader shift toward automated discovery that is reshaping exoplanet science. Rather than relying on individual research groups to comb through TESS data with slightly different thresholds and vetting rules, pipelines like RAVEN apply uniform detection and validation criteria to millions of stars at once. That consistency matters: it reduces the patchwork quality of earlier catalogs and makes population-level statistics more reliable.
Each validated planet also implies several more that current surveys cannot yet detect, whether because the planet’s orbit never lines up for a transit as seen from Earth or because its signal is too faint to rise above the noise. As TESS continues to accumulate data and as machine-learning pipelines grow more sensitive, the census of nearby planetary systems will keep expanding. The emerging picture is that close-in planets, many of them smaller than Neptune and on orbits far tighter than anything in our own Solar System, are remarkably common. The question is no longer whether they are out there, but how many we have yet to find.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
