An artificial intelligence system trained on millions of earthquake recordings has uncovered six previously unknown structures deep inside the Earth, sitting near the boundary between the planet’s molten outer core and the rocky mantle above it. Researchers fed the AI more than two million seismic waveforms gathered from thousands of powerful earthquakes, allowing it to detect subtle patterns in how seismic waves bend and scatter that had gone unnoticed in decades of manual analysis. The discovery adds a striking new layer of detail to scientists’ picture of the deep Earth, a region no instrument has ever directly observed.
How the AI Found What Humans Missed
According to researchers behind the study, the system analyzed more than two million individual seismic waveforms drawn from roughly 5,000 earthquakes of magnitude 6 or greater, a dataset far larger than any team of human seismologists could realistically comb through by hand. Seismic waves generated by earthquakes travel through the entire planet, and subtle changes in their speed, timing, and shape as they pass through different layers can reveal structures that never show up on the surface.
Machine learning models are particularly well suited to this kind of pattern recognition because they can flag faint, repeated anomalies across enormous datasets that a human analyst would likely dismiss as noise or simply never have the time to review in such volume. The technique builds on a long tradition in seismology of using distant earthquakes as natural probes of the planet’s interior, since waves recorded thousands of miles from their source carry information about every layer of rock and metal they passed through along the way. By training the system on known seismic signatures and then letting it search for statistically consistent deviations, the researchers were able to isolate signals corresponding to genuinely new structures rather than random variation.
Six Zones Near the Core-Mantle Boundary
The analysis identified six distinct zones, labeled B1 through B6, clustered near a depth of roughly 2,900 kilometers, which is where the Earth’s rocky mantle gives way to the molten metal of the outer core. This boundary is one of the most dramatic transitions inside the planet, marking the point where solid, slowly flowing rock meets a liquid iron-nickel layer responsible for generating Earth’s magnetic field.
Scientists have long known the core-mantle boundary is not uniform, with prior seismic studies revealing large-scale features such as continent-sized zones of unusually dense material. The newly identified B1 through B6 structures appear smaller and more localized than those previously known features, suggesting the boundary region is considerably more complex and varied than earlier lower-resolution surveys were able to capture. Some of the earlier known features near the core-mantle boundary span areas comparable in size to entire continents, making the newly identified, more compact B1 through B6 zones a genuinely different category of structure rather than a finer-grained view of something already mapped.
What the Structures Might Be
Researchers have proposed several possible explanations for the newly detected zones, including pockets of partially molten rock, chemically distinct material left over from the earliest stages of Earth’s formation, or the remnants of ancient tectonic plates that sank deep into the mantle over hundreds of millions of years before settling near the core boundary. Each explanation would carry different implications for how heat and material move between the core and mantle over geological time.
Distinguishing between these possibilities will likely require additional seismic data, since the current findings identify the presence and rough location of the structures without fully resolving their composition or origin. Follow-up studies using different earthquake sources and denser networks of seismic stations could help narrow down which explanation, or combination of explanations, best accounts for each of the six zones.
Why Deep-Earth Structure Matters
The core-mantle boundary plays a central role in how heat escapes from Earth’s interior, a process that drives mantle convection, powers plate tectonics at the surface, and sustains the churning motion in the outer core that generates the planet’s protective magnetic field. Understanding the fine structure of this boundary helps researchers refine models of how these deep processes have shaped the planet’s surface and its magnetic field over billions of years.
Because no instrument can physically reach depths anywhere close to 2,900 kilometers, seismology remains essentially the only tool available for studying this part of the planet, making advances in how seismic data is analyzed directly responsible for what scientists are able to learn about Earth’s interior.
A Preview of AI’s Role in Earth Science
The discovery also illustrates a broader shift underway in geoscience, where machine learning tools are increasingly applied to enormous archives of seismic, satellite, and sensor data that have accumulated for decades but were never fully mined for subtle patterns. Similar approaches are being tested in fields ranging from volcanic monitoring to earthquake early warning, where the ability to detect faint signals buried in noisy data can translate directly into better forecasts and a clearer picture of processes still not fully understood.
This article was produced with the assistance of AI and reviewed by Morning Overview editors.
More from Morning Overview
