Large language models can interpret enormous quantities of written information but cannot see what is happening in a field, forest, port or community.
In a presentation to the AI for Good Global Summit, Planet co-founder and Chief Executive Will Marshall described a possible answer: “planetary intelligence”.
The idea brings together frequently updated Earth observation and other real-world sensor data with artificial intelligence. Instead of asking only what normally happens during a flood, fire or crop disease outbreak, users could ask what is happening at a specific location, how that situation is changing and what action might be needed.
Marshall calls the models trained on this physical-world information Large Earth Models. Where a large language model learns from text, a Large Earth Model would learn from imagery and other observations of changing conditions on Earth.
From pixels to answers
For the Earth observation industry, the important change is not simply the addition of AI. It is the potential reduction in time between observation and action.
Marshall describes systems capable of combining current and historical imagery with weather, infrastructure and population data. Possible applications include:
- Identifying the extent and direction of a wildfire
- Detecting illegal forest clearance or fishing activity
- Monitoring crop health and agricultural compliance
- Assessing damage to individual buildings following a disaster
- Tracking construction, infrastructure and supply-chain activity
Planet says its emerging AI application can already use natural-language questions to search imagery, identify change and produce analytical reports. The company is presenting this as an early step towards a broader planetary intelligence capability. Planet’s supporting article describes the application as being in open beta.
A commercial opportunity for Earth observation
The concept points towards a market in which the value of Earth observation is increasingly measured by the quality, speed and reliability of the answer rather than the number of images delivered.
That creates opportunities throughout the commercial EO supply chain. Data providers, analytical specialists, software developers and organisations with deep sector knowledge will all be needed to turn observations into dependable decisions.
Lowering the technical barrier could also open Earth observation to organisations without specialist GIS teams. A local authority, insurer or environmental organisation might ask a question in ordinary language and receive relevant imagery, detected changes and supporting evidence without constructing a complex processing workflow.
Intelligence still requires trust
Making geospatial analysis easier to access does not remove the need for expertise. Automated answers must retain clear links to their source data, methods and uncertainty.
Validation, interoperability, transparent provenance and appropriate human oversight will become more important as AI-generated geospatial conclusions move into operational decisions.
Planetary intelligence remains an ambitious and partly speculative vision. However, the direction is already visible: Earth observation is moving beyond showing people what the world looks like and towards helping them understand what has changed, why it matters and what they should do next.
The presentation expands on Marshall’s Planetary Intelligence essay, published in May 2026.