Amateur Astronomer Uses AI to Find Exoplanet Candidate
Amateur astronomers are using artificial intelligence tools to find previously unknown exoplanet candidates, resulting in mixed validation outcomes according to Nature.
-
The Tech TL;DR:
- Amateur scientists are deploying large language models and public telescope repositories to detect planetary candidates without formal academic training.
- The surge of automated submissions has overwhelmed academic peer-review channels and verification systems.
- Professional research teams must manually vet algorithmic detections to rule out false positives like binary star systems.
Pavel Rabtsevich Scours TESS Data
In September, Pavel Rabtsevich sat down at his computer and asked an AI model to discover a planet, Nature reported. Rabtsevich works in Spain as an artificial-intelligence specialist and has no formal astronomy education, or a telescope, according to szinas.com.
The bot was given a specific set of telescope data by Rabtsevich alongside a broad strategy to search for signs of hidden planets. As Nature noted, he utilized Claude Code—an Anthropic-developed AI programming assistant—to parse through information regarding upwards of 120,000 stars monitored by NASA’s Transiting Exoplanet Survey Satellite (TESS) across late 2025 and early 2026.
TIC 4206066 Signal Detection
According to Nature, Claude focused intently on TIC 4206066, a star exhibiting dual dimming cycles occurring roughly every 3 days and every 11 days. The search homed in on TIC 4206066, a star slightly smaller than the Sun about 116 light years from Earth, szinas.com noted.
szinas.com reported that if confirmed as a planet, its size would measure roughly 1.4 times that of Earth, with Rabtsevich calculating its surface temperature to be near 540C. Rabtsevich instructed Claude to look for those same repetitive patterns in archival TESS data, and data from 2018 and 2020 recorded the same pattern of dimming, according to Nature.
Algorithm Validates Signals Through Extensive Testing
At Rabtsevich’s direction, Claude then found a planet-detection algorithm written and made publicly available by Michelle Kunimoto, an exoplanet researcher at the University of British Columbia in Vancouver, Canada, and ran it to validate the signals, Nature reported. szinas.com said that rather than accepting the initial outputs from the models, Rabtsevich cross-examined them against alternative models and conducted 74 separate tests using Claude Code to actively challenge his own premise.
Rabtsevich published his findings to the platform Zenodo as a preprint on September 25, Nature stated. Although his findings have bypassed formal peer review, TESS researchers independently reviewed the signals and incorporated them into their official roster of planetary candidates, as reported by Nature.
Concerns Over Data Overload
The newfound accessibility of data-analysis tools has meant that not all amateurs are doing research responsibly, Kunimoto told Nature. Kunimoto pointed out that certain planet hunters lack a complete comprehension of the algorithms driving these AI systems, preventing them from properly verifying the conclusions generated by the models.
“There’s been a huge, overwhelming surge of AI-driven papers, which tend to have [planet] candidates that are not the best quality.”
Michelle Kunimoto told Nature.
szinas.com reported that Kevin Hardegree-Ullman, a research scientist with the NASA Exoplanet Science Institute, informed Gizmodo that while artificial intelligence might not directly verify planet discoveries, it will likely remain a go-to resource for spotting candidates and performing preliminary evaluations.
Ongoing Satellite Observations
Rabtsevich’s request for follow-up was approved, and TESS is measuring the star’s brightness every two minutes from October 31 to November 26, szinas.com noted. Rabtsevich published in advance the times when the star should dim.
“If the star dims on time in November, this gets a lot more serious. If it doesn’t, I’ll post that too,” Rabtsevich wrote on X, as reported by szinas.com.