Skip to main content

The findings of our first generative AI experiment: GOV.UK Chat

A screenshot of the experimental GOV.UK Chat system. A user has typed in a question and the system is generating a response. The question reads “As a small business, do I need to charge VAT to people in Italy when I sell a service online?”

GOV.UK has run its first generative AI experiment to see how a Large Language Model (LLM) would work using GOV.UK content. Here are our early findings.

Updating GOV.UK when there is a machinery of government change

Three cogs of varying colours with individual icons in the middle of each cog. From left to right the icons are, a lightbulb, a power button and a briefcase.

You might’ve spotted a change in the number of government departments. Here’s how we supported this change, updated GOV.UK to reflect it and what we learned from doing so.

Responding to the changing user needs of the COVID-19 landing page

A woman sitting at a table with a laptop. On the laptop screen is the GOV.UK coronavirus page.

As the subject of over 100,000 Google searches a month, we know people are still looking to GOV.UK for information on coronavirus. So when we discovered the page was not meeting changing user needs, we made some improvements.