New AI tool finds hidden planets in NASA data

Automated search reveals cosmic secrets

In the quiet expanses of our galaxy, countless worlds remain tucked away, obscured by the very light of the stars they orbit. For years, the data collected by NASA's Transiting Exoplanet Survey Satellite (TESS) has sat in digital archives - a silent library of potential discoveries waiting for a keen enough eye to read them. In March 2026, astronomers at the University of Warwick announced that a sophisticated machine learning pipeline named RAVEN had successfully validated 118 previously overlooked planets - along with over 2,000 high-quality candidates - that had escaped detection in earlier analyses. This breakthrough suggests that our cosmic neighborhood is even more crowded than we once imagined.

Celestron NexStar 5SE Computerized Telescope

A popular go-to telescope for amateur astronomers. Motorized mount, 5-inch aperture, and smartphone compatibility - ideal for readers inspired to explore the night sky.

Buy now

What is the RAVEN pipeline and how does it work?

RAVEN, which stands for RAnking and Validation of ExoplaNets, acts as a high-precision filter for astronomical noise. When a planet passes in front of its star, it creates a minute dip in brightness called a transit. However, stars are often restless, flickering with spots and flares, while background eclipsing binary stars can also mimic the signature of a planet. Human researchers have spent years manually vetting these signals - a slow and meticulous process that can miss the most subtle variations.

RAVEN changes this dynamic by processing the light curves of over 2.2 million stars simultaneously, applying a Bayesian framework that combines a Gradient Boosted Decision Tree and a Gaussian Process classifier to recognize the mathematical fingerprint of a true world. The pipeline handles the entire workflow in a single pass, from initial signal detection through machine learning vetting to final statistical validation - a level of automation that simply was not possible before.

A diversity of distant worlds

The census of new discoveries includes a wide array of planetary types, ranging from rocky terrestrials to gas giants that defy conventional expectations. Among the most notable finds are several Hot Jupiters - massive planets that orbit their host stars so closely that they complete a full revolution in under 16 days. These extreme environments serve as natural laboratories for understanding atmospheric physics under intense radiation.

The pipeline also flagged several multi-planetary systems, where gravitational dances between siblings create complex orbital patterns that require high computational power to untangle. Each of these systems tells a unique story about how planets form and evolve around different stellar types.

Exoplanets by Michael Summers & James Trefil

An accessible, beautifully illustrated guide to the science of exoplanet discovery - directly relevant to the article's topic and ideal for general science readers.

Buy now

What RAVEN revealed about planetary populations

A companion study published simultaneously in the Monthly Notices of the Royal Astronomical Society shed further light on the broader population of close-in planets. Key statistical findings include:

• Around 9-10% of Sun-like stars host at least one short-period planet, consistent with earlier estimates from NASA's Kepler mission
• RAVEN allowed researchers to put, for the first time, a precise number on the "Neptune desert" - a striking gap in the population of Neptune-sized planets at very short orbital periods
• Only 0.08% of Sun-like stars appear to host a planet in this region, confirming just how barren this zone truly is

These population-level statistics represent a significant leap in our ability to characterize not just individual planets, but the overall architecture of planetary systems across the galaxy.

Why this discovery matters for the search for life

Finding these worlds is a reminder of the vastness that surrounds our small blue marble. Each dip in a light curve represents a physical place - a horizon where different suns rise and different chemistries brew. By automating the search, researchers can now focus their efforts on characterization, using ground-based telescopes to peer into these atmospheres and search for the chemical signatures of life.

The efficiency of the RAVEN pipeline allows the scientific community to move beyond the simple question of whether planets exist, focusing instead on the intricate stories each system has to tell. In the context of astrobiology and the search for biosignatures, this shift in bandwidth from detection to characterization is enormously significant.

The future of AI-powered exoplanet discovery

The integration of machine learning into astrophysics represents a fundamental shift in how we approach the unknown. As our sensors become more sensitive, the volume of data generated by missions like TESS or the upcoming Nancy Grace Roman Space Telescope will exceed human capacity for manual review. Tools like RAVEN ensure that no signal is left unexamined.

This latest batch of over 100 validated planets is likely just the beginning of a larger surge in planetary census data. For the astronomers involved, the success of this tool provides a sense of quiet optimism - it suggests that the answers to some of our most profound questions about the universe are already in our possession, hidden in the archives we have already built.

As we refine these digital eyes, the darkness between the stars feels a little less empty, populated by the many worlds that RAVEN is finally bringing into the light. The discovery underscores a fundamental truth of modern science: sometimes, to see further into the distance, we must first find better ways to look at what is right in front of us.