From Earth to Algorithms: Generative AI in Geoscience

My feature article published today in GEOSCIENTIST adds to the growing discussion on the role of Large Language Models (LLMs) in the geosciences.

AI is not an isolated technology; it is embedded in broader scientific, technical, and institutional systems. Developing AI literacy in geoscience therefore requires (i) solid foundational geological understanding, (ii) technological and data-science competence to grasp both possibilities and limitations, (iii) critical thinking, and (iv) ethical awareness.

As with any information technology, geoscientists using the same tools can arrive at very different outcomes. This is already true with search engines, and the effect is magnified by the generative and interpretive capabilities of LLMs.

AI literacy matters not just for researchers and developers of algorithms and models, but for the far larger group of geoscientists who will be consumers of these systems. Understanding when AI helps, when it hinders, and when it should not be used at all, is increasingly essential for responsible practice.

LLMs can support use cases which are far more than just question and answer “chatbot style” interactions. Many geoscientists are aware of this, but many are not. The article outlines a range of emerging use cases from the literature and from my own experience, across minerals and mining, carbon storage, oil and gas, geohazards, and hydrogeology.

To support local experimentation, I released open source code for an AI agent that orchestrates the retrieval and visualization of external open geoscience data from natural-language queries.

The article also highlights the recent International Union of Geological Sciences (IUGS) Geoethics Commission’s AI Ethics Recommendations for the Geosciences. This has already having a significant impact on policy, endorsed by geological institutions representing more than 300,000 geoscientists worldwide. This covers topics from responsible use, data privacy and transparency to environmental, scientific integrity and geopolitics.

We must continue to question how AI shapes geoscience knowledge, remain alert to its limitations, and choose not only how, but when, to use it. With a pragmatic approach, LLMs can help us deepen our understanding of Earth and better serve society, while we prepare the next generation of geoscientists for an AI-augmented future.

Link to the online article below, “From Earth to Algorithms: Generative AI in Geoscience” in the 2025 Winter Edition of Geoscientist, along with many other excellent articles such as Mars’ Watery Underworld.

From Earth to algorithms: Generative AI in geoscience