AI helps botanists make sense of difficult plants

Botanists could spend less time sorting specimens and more time protecting plants, thanks to artificial intelligence that helps recognise difficult species.

Kew’s June 2026 report describes computer vision learning from labelled plant images, suggesting identities and flagging unusual specimens for experts. A possible new species still needs scientific confirmation.

One example is FireNetSEz, developed to distinguish North American peat mosses. Its published project code uses images just 64 pixels across. Kew’s report explains why this is useful: identifying these mosses normally involves examining tiny features under a microscope.

Another test in Kew’s report correctly identified 63% of specimens in one sedge genus. The correct species appeared among five suggestions around 90% of the time.

But does this AI run on renewable electricity, and how much water does it use? The project materials checked do not disclose those figures. Training and subsequent identification runs need separate accounting, including direct cooling water and water used to generate electricity.

Human checking remains essential. Used carefully, these tools can free specialists to investigate overlooked plants and guide conservation.

Sources

• Kew — State of the World’s Plants and Fungi 2026
• Kew — digitisation, access and human verification
• Little and colleagues — FireNetSEz project code and datasets