## TL;DR
AI is transforming archaeology by enabling automated detection of buried structures from satellite imagery, classification of millions of pottery sherds, decipherment of ancient scripts, and reconstruction of fragmented artifacts. From discovering hidden Maya cities in LiDAR point clouds to identifying new Nazca geoglyphs from satellite data, deep learning is giving archaeologists a powerful lens through which to see the past.

## Core Explanation
The archaeology-AI pipeline: (1) Prospection -- satellite imagery (visible, multispectral, SAR) and airborne LiDAR are analyzed by CNNs and vision transformers to detect subtle surface signatures of buried archaeological features. Transfer learning from pre-trained models enables detection even with limited labeled training data; (2) Artifact classification -- deep learning models classify pottery fragments by style, period, and origin based on visual features, achieving expert-level accuracy while processing millions of items; (3) Text and inscription analysis -- LLMs and specialized OCR (Transkribus, PyLaia) transcribe and translate cuneiform tablets, hieroglyphics, and medieval manuscripts; (4) 3D reconstruction -- photogrammetry combined with neural rendering creates high-fidelity 3D models of sites and artifacts for virtual museums and remote research.

## Detailed Analysis
The Springer 2025 review identifies four main AI application domains: computer vision (object detection, semantic segmentation for site/artifact identification), deep learning (CNN, Transformer for image classification), geospatial analysis (LiDAR point cloud processing, multispectral satellite data fusion), and natural language processing (historical document transcription, ancient language translation). Key breakthroughs: (1) Maya lowlands LiDAR survey -- deep learning detected ~60,000 previously unknown structures across 2,100 km² in Guatemala, revealing a civilization far more populous than previously estimated; (2) Nazca Lines discovery -- a ResNet-50 classifier identified 303 new figurative geoglyphs in 2024, nearly doubling the known catalog; (3) Vesuvius Challenge -- AI-enabled virtual unwrapping and ink detection on carbonized Herculaneum papyri, revealing previously unreadable philosophical texts. Gaps remain in multi-modal fusion, standardizing archaeological data formats for AI, and addressing ethical implications of automated heritage interpretation.

## Further Reading
- Journal of Computer Applications in Archaeology (CAA)
- Vesuvius Challenge: scrollprize.org
- Transkribus AI for Historical Documents: readcoop.eu