A cold war first
The Soviets beat the U.S. to the Moon in 1966, when Soviet engineers landed Luna 9 on Feb. 3. The spacecraft became the first human-built machine to achieve a soft landing on another world.
Engineers designed the mission under the leadership of Chief Designer Sergei Korolev within the Soviet space program. The lander transmitted panoramic images and soil data for three days during the height of the space race between the Soviet Union and the United States, photographs that marked the first views ever sent from the lunar surface.
Tracking data from 1966 wasn't very accurate, relatively speaking. Mission planners calculated a landing region, yet the margin of error stretched for tens of kilometers. Luna 9 was only a few meters across, making orbital confirmation difficult even after NASA launched the Lunar Reconnaissance Orbiter in 2009 under Administrator Charles Bolden.
Published coordinates didn't fully match the flat horizon visible in the original panoramas.
The mystery lingered for nearly six decades.
Scientists turn to artificial intelligence
Researchers working in England and Japan developed a lightweight machine learning system named YOLO-ETA, short for You Only Look Once Extraterrestrial Artefact.
The team trained the model using high-resolution imagery of Apollo landing sites captured by the Lunar Reconnaissance Orbiter Camera. The algorithm learned to recognize shapes, shadows, and surface disturbances that indicate human-made hardware. It identified known spacecraft, such as Luna 16, with high confidence despite variations in lighting and terrain. Engineers kept the model compact so it could work on limited onboard computing power in future missions.
Strong signals on the surface
Researchers directed YOLO-ETA to scan a 5×5-kilometer region around the historical Luna 9 coordinates, and the system flagged several high-confidence detections within it.
There is one primary area containing smaller objects within 200 meters, with dark patches nearby aligned with possible impact sites from discarded landing components.
Topographic analysis from the Lunar Orbiter Laser Altimeter shows the candidate site rests on a gentle rise about 25 meters above the surrounding plains. That elevation creates a flat horizon stretching nine to ten kilometers, matching the terrain profile captured in Luna 9 images from 1966. Three-dimensional reconstructions built from orbital imagery also correspond to crater patterns seen in the original panoramas.
A competing search
Vitaly Egorov, a space historian and founder of the Zelenyikot blog, conducted an independent search using publicly available lunar data. He enlisted volunteers to analyze NASA QuickMap imagery and flagged an alternative location roughly 25 kilometers away from the AI-identified site. Some surface features at that spot also looked like details from the Luna 9 landing sequence.
Anatoly Zak, a space journalist and founder of RussianSpaceWeb, noted that both teams can't be right, emphasizing that higher-resolution imaging will determine which location matches the historic lander.
Researchers describe the AI detection as a promising lead rather than confirmed proof.
Awaiting confirmation
The Chandrayaan 2 orbiter, operated by the Indian Space Research Organization, carries cameras capable of imaging at resolutions of 1 meter per pixel or better.
Planned passes over both candidate zones in March 2026 may provide decisive evidence.
High-resolution imagery could confirm the shape, size, and layout consistent with Luna 9 hardware.
A verified ID would mark the close of a 60-year chapter in lunar exploration. Luna 9 proved that a soft landing worked. Modern AI now attempts to pinpoint the achievement with mathematical precision.
If confirmation arrives, history and machine learning will meet at a set of exact lunar coordinates, tying a Cold War milestone to a new era of orbital analysis and automated discovery.
