The work was led by David O’Ryan and Pablo Gómez and published in Astronomy & Astrophysics last year; a report on the study appeared in Space Daily on June 6, 2026.

"Previously undocumented" refers to objects not previously appearing in the literature, not to new categories; these as-yet-candidate catalog entries require spectroscopic follow-up for confirmation.

The core takeaway is that this method demonstrates a scalable workflow for spotting rare or unusual objects in massive astronomical archives, though confirmation and follow-up work remain essential to determine their nature and distances.

An AI tool called AnomalyMatch scanned nearly 100 million cropped images from the Hubble Space Telescope archive and surfaced a shortlist of unusual objects, identifying more than 800 that had not been described in the scientific literature.

The system ranked images by novelty against its training data, after which human astronomers visually evaluated the top candidates, confirming over 1,300 visually anomalous objects and cataloguing 1,255 unique objects across 18 classifications.

Most flagged objects are galaxies in merging or interacting states, including more than 400 such galaxies, plus 86 new gravitational lens candidates and examples like collisional ring galaxies, jellyfish galaxies, and edge-on planet-forming disks within our Milky Way.

The article notes the broader significance: automated anomaly detection enables scalable discovery as future surveys (Euclid, Rubin Observatory) produce datasets far larger than human teams can inspect unaided.

A subset of objects did not fit existing classification schemes, marking them as the most promising targets for follow-up studies.

The search covered roughly 35 years of Hubble data, making it the first systematic anomaly search of the Hubble Legacy Archive.