Skip to content
ADVERTISEMENT

AI fails to spot Japanese knotweed in 7% of cases, research finds

Environet said that while AI can help as a guide, relying on it alone is risky for property transactions where accuracy is needed.

AI fails to spot Japanese knotweed in 7% of cases, research finds
ADVERTISEMENT

Analysis from Environet found artificial intelligence (AI) tools misidentified Japanese knotweed in 7% of cases, leading to warnings for homeowners, buyers and sellers. 

The analysis by Environet used 100 images of Japanese knotweed in different conditions and at various times of year, uploading them to ChatGPT. 

The tool got 93 cases right, but in seven it confused knotweed with cherry laurel, spiraea, rose, dock, Virginia creeper and peony.

Homeowners and sellers have a legal duty to declare Japanese knotweed on the Law Society’s TA6 form when selling property. 

ADVERTISEMENT

Environet said that while AI can help as a guide, relying on it alone is risky for property transactions where accuracy is needed.

ADVERTISEMENT

Emily Grant, director at Environet, said: “AI tools can be a useful starting point, but they are far from 100% reliable when it comes to identifying invasive plants like Japanese knotweed. 

“Our analysis showed AI is competent at recognising knotweed when it presents typically, but it becomes confused when it presents atypically, which is common after it’s been chemically treated or repeatedly cut back.

“It’s always best to have a plant checked by a qualified specialist who can confirm what it is and advise on the appropriate next steps.”

ADVERTISEMENT

Grant added: “Early and accurate identification is critical to keeping treatment costs down and avoiding legal complications further down the line.”

ADVERTISEMENT