AI-Guided Cave Survey Suggests Human Presence in Ireland 5,000 Years Earlier Than Known

A computer model trained on the caves of southwest Wales has led archaeologists to eight sites in Ireland where they found finger marks pressed into clay, engravings cut into stone, and paintings made with ochre and charcoal. The marks, specialists believe, are at least 15,000 years old—which would mean humans reached Ireland roughly 5,000 years before the previously accepted earliest evidence of settlement on the island.

The discovery emerged from a straightforward logical chain. During the last Ice Age, when sea levels dropped dramatically as water locked into ice sheets, a broad plateau emerged between Pembrokeshire in Wales and the coast near Rosslare in Ireland—a walkable span of about 46 miles connecting two landmasses that are now separated by open water. The plateau was cold, with summer temperatures estimated between minus 10 and minus 20 degrees Celsius, but it supported large herds of wild horses, mammoth, bison, and reindeer. For hunter-gatherer groups, herds meant survival. Where the animals went, people followed.

South Wales already held evidence of this pattern. The region shows signs of human occupation stretching back 35,000 to 40,000 years, with caves preserving a long sequence of marks made on stone across many millennia. Dr. George Nash of Liverpool University, a specialist in Welsh rock art, reasoned that if people had been living and marking caves on one side of the exposed plateau for tens of thousands of years, the far side was not an unlikely destination—it was the obvious one. The reason no one had found much evidence there may have been simply that no one had looked systematically. Ireland's previously accepted oldest evidence of human presence dated to around 10,500 years ago, and that discovery had come more or less by chance rather than through deliberate search. A baseline established by accident is a statement about where archaeologists happened to dig, not where people once lived.

Instead of walking the coastline hoping for luck, the team built a filter. Researchers trained a computer model on the measurable characteristics of known cave-dwelling and rock-art sites in south Wales—the geology, orientation, and setting that repeatedly coincided with human occupation and marking. When the model was applied to southwest Ireland, it returned 28 caves worth investigating. The team, which included researchers from University College Dublin, the University of Adelaide, and institutions in Wales, Portugal, and Dubai, along with caving specialist Shane Diffily and digital mapping consultant Darren Lee, then went to look.

Eight of the 28 caves held art. The marks took multiple forms. Some were engravings cut into rock. Others were paintings made with ochre or charcoal. A third category consisted of fingertip line sketches drawn into clay and earth—the softest possible medium for recording the oldest possible gesture. There were also flutings, parallel grooves left where fingers had been dragged through soft layers on a cave wall, a mark type known from Ice Age sites across Europe and one that records, quite literally, the width of a hand. The project has withheld the exact locations of the caves to reduce the risk of vandalism and looting, a standard precaution for fragile prehistoric art that carries an unavoidable cost: the public receives the count and the categories, not the coordinates.

It is important to be precise about what this finding does and does not establish. No radiocarbon dates from a laboratory have been published for the art. The estimate of at least 15,000 years ago, possibly much earlier, is a contextual and stylistic assessment offered by specialists who work on comparable material—a considered judgment, not a laboratory measurement. Ochre, charcoal, clay, and engraved stone can in principle be dated through several methods, but until results exist and are examined by other researchers, the age remains an argument rather than a fixed point on a timeline. The land-bridge migration is likewise a hypothesis the project is advancing, not a demonstrated population history. There is no genetic evidence of who walked from Pembrokeshire to Rosslare, and no claim that there was a single migration event.

Nash expects the eight caves are not the end of the list. His reasoning is geographic: people moving inland from a southeastern landfall would not have skipped the counties of Wexford, Waterford, and Cork on the way to southwest Ireland, which suggests further sites are waiting in between. That is a prediction, and a testable one—the same model that produced 28 candidates can be pointed at other coastlines. For now, the eight caves establish something narrower and still remarkable: in a country whose human story was thought to begin around 10,500 years ago, there are walls with fingermarks on them that specialists believe are considerably older, found not by accident but because a machine was asked where to look.