How Artificial Intelligence Is Rewriting Planet Hunting

For more than three decades, astronomers have searched for planets orbiting stars beyond our solar system. Since the first confirmed exoplanet detection in the 1990s, scientists have identified more than 5,000 worlds using space telescopes and ground observatories. Yet the search generates enormous volumes of data, far more than human researchers can analyse manually. Artificial intelligence is now transforming how astronomers sift through that data, accelerating discoveries and uncovering planets that traditional methods missed.

The Challenge of Finding Exoplanets

Most exoplanets are discovered using the transit method, which measures tiny dips in a star’s brightness when a planet crosses in front of it. NASA’s Kepler Space Telescope and the Transiting Exoplanet Survey Satellite (TESS) have monitored hundreds of thousands of stars, producing light curves that record changes in brightness over time.

Each light curve contains noise caused by stellar activity, instrumental fluctuations, and background interference. Identifying genuine planetary signals requires distinguishing subtle, repeating patterns from random variability. A single planet transit may reduce a star’s brightness by less than one per cent, which makes detection statistically demanding. According to NASA’s Exoplanet Archive, Kepler alone collected data on more than 150,000 stars during its primary mission. Manual inspection of every potential signal would require enormous human effort and time.

Enter Machine Learning

Machine learning algorithms excel at pattern recognition in large datasets. In 2017, researchers from Google AI and the University of Texas at Austin published a study in The Astronomical Journal demonstrating that a neural network could identify exoplanet candidates within Kepler data. The model was trained using confirmed planet signals and false positives, enabling it to classify new signals based on learned patterns.

The team applied the neural network to previously analysed Kepler data and discovered two additional exoplanets in the Kepler 90 system. One of them, Kepler 90i, made the system comparable in planet count to our own solar system. Andrew Vanderburg, a co-author of the study, explained that the neural network helped identify weak signals that had been overlooked by earlier analysis pipelines. This work showed that artificial intelligence could not only speed up classification but also recover real planets that traditional filters missed.

Expanding the Search With TESS

TESS continues to generate vast streams of stellar brightness data. Researchers have developed increasingly sophisticated machine learning frameworks to process this information. A 2021 study published in Monthly Notices of the Royal Astronomical Society described the use of deep learning algorithms that automatically distinguish between planetary transits and astrophysical false positives such as eclipsing binary stars.

Dr. Chelsea Huang of the Massachusetts Institute of Technology, who has worked on machine learning applications for TESS, has noted that automated classification systems allow astronomers to prioritize the most promising candidates for follow-up observations. This reduces time spent on spurious detections and improves telescope scheduling efficiency. AI systems can analyze thousands of light curves in minutes, a task that would otherwise take months of human review.

Finding Planets in Archived Data

One of the most significant advantages of artificial intelligence is its ability to reexamine archival data with fresh methods. In 2020, a team led by researchers at the University of Bern used machine learning to identify previously undetected exoplanet signals hidden within Kepler datasets. Their results, published in Astronomy and Astrophysics, showed that algorithms trained on simulated transit signals could recover small planets that had escaped earlier detection thresholds.

These findings suggest that existing archives still contain undiscovered worlds. As computational models improve, astronomers can revisit old observations and extract new insights without launching new spacecraft.

Beyond the Transit Method

Artificial intelligence is also being applied to other detection techniques, including radial velocity measurements, which track subtle shifts in a star’s spectrum caused by gravitational tugs from orbiting planets. Machine learning models can filter out stellar noise and identify periodic signals that indicate planetary companions.

In addition, AI tools assist in atmospheric characterisation. When telescopes such as the James Webb Space Telescope analyse exoplanet atmospheres, they produce complex spectra that must be interpreted carefully. Neural networks trained on atmospheric models can accelerate the identification of molecules such as water vapor, methane, and carbon dioxide.

Limits and Human Oversight

Despite its advantages, artificial intelligence does not replace astronomers. Machine learning models depend on the quality of their training data and can inherit biases from existing classifications. False positives remain a concern, especially when algorithms are applied to faint or borderline signals.

David Hogg, an astrophysicist known for applying data science to astronomy, has emphasised that machine learning should complement physical modelling rather than substitute for it. AI can highlight patterns, but interpretation still requires theoretical understanding and observational confirmation.

A New Era of Discovery

Artificial intelligence is reshaping planet hunting by accelerating data analysis, recovering missed signals, and expanding the search into new detection domains. As telescope datasets grow larger with missions like TESS and upcoming observatories, automation becomes increasingly essential.

The integration of AI into astronomy reflects a broader shift toward computational science, where algorithms and human expertise work together. Thousands of exoplanets have already been discovered, yet many more likely remain hidden in existing data. With artificial intelligence refining the search, astronomers are entering an era in which new worlds can be uncovered not only by building bigger telescopes, but also by teaching computers to see patterns in the stars.