Artificial intelligence tools are helping citizen scientists discover undiscovered exoplanets (artist illustration). Credit: Hypersphere/SPL
In September, artificial-intelligence expert Pavel Rabtsevich asked an AI model to search telescope data for previously undetected planets. About two weeks later, the system identified signals suggesting that two planets could be orbiting a star 116 light-years from Earth. Professional astronomers are now examining the findings, which have not yet been confirmed as planets.
Rabtsevich, who works in Spain and has no formal astronomy training, is among the most successful of a growing group of amateur astronomers using AI to investigate worlds beyond the Solar System. Citizen scientists have been discovering exoplanets for decades with tools released by NASA and other researchers. AI models can now locate those tools online and apply them to large datasets, lowering the technical barrier for non-experts who want to search for planets.
However, the growing availability of automated data-analysis tools also creates risks, says Michelle Kunimoto, an exoplanet researcher at the University of British Columbia in Vancouver, Canada. Some planet hunters do not fully understand the algorithms used by AI systems and therefore cannot properly validate their results. Kunimoto says there has been an overwhelming surge in AI-driven papers and warns that “the systems we use to disseminate science are being overwhelmed.”
AI searches TESS data
Rabtsevich used Claude Code, an AI coding tool developed by Anthropic, for his search. He asked it to examine observations of more than 120,000 stars collected by NASA’s Transiting Exoplanet Survey Satellite (TESS) in late 2025 and early 2026.
He instructed Claude Code to apply a widely used planet-detection algorithm that searches for repeated dips in a star’s brightness. Such regular dimming can occur when a planet passes between its star and the observer, blocking a small amount of starlight during each transit.
The model focused on a star called TIC 4206066. Its light showed two distinct dimming cycles: one occurring approximately every three days and another about every 11 days. The pattern suggests that the star may be orbited by two planets, with the planet completing a transit every three days orbiting closer to the star than the one with the longer cycle.
Rabtsevich then asked Claude Code to search archived TESS observations for the same repeating patterns. The 2018 and 2020 data contained the same dimming signals from TIC 4206066. At Rabtsevich’s direction, Claude Code located a planet-detection algorithm created and published by Kunimoto, which Rabtsevich ran to verify the signals.
Rabtsevich published the findings as a preprint on Zenodo on September 25.1 The work has not been peer-reviewed, but TESS astronomers independently analyzed the signals and added them to the official list of planet candidates.
Source: www.nature.com












