How AI-ready content can boost policy effectiveness

Andrea Leary

I’m often asked about the risk of generative (GenAI) bots serving up incorrect information from GOV.UK or Business.gov.uk. My answer is that GenAI bots are not the risk in themselves. Instead, they’re just another channel that gets Department for Business and Trade (DBT) information to wider audiences. We should focus less on a user's GenAI bot intelligence and more on improving DBT's content quality, which we have direct influence over.

If content is well-structured and written to the principles of good content design, the GenAI bot used by a business will be much more likely to find the information requested. It serves it up according to the AI bot-user’s needs.

How much are small businesses relying on AI?

DBT’s official statistics about the UK’s business landscape (summarised in this House of Commons library booklet) show that 95% of UK businesses employ 9 people or fewer.  74% of businesses consist of only directors and have no employees. Together they account for almost 30% of the UK’s business turnover.

Not only do these types of businesses lack the staff to do time-consuming research for business information, they may also blur the lines between the “tech hacks” they use at home and at work.

When we drill into GOV.UK figures, we can estimate that 3.4m ChatGPT sessions in 2025 have involved users engaging with DBT-related content. Meanwhile referral traffic from ChatGPT to the GOV.UK website continues to grow – in fact by 346% from January to September this year.

Anecdotally, some businesses are now asking DBT International Trade Advisers to verify AI summaries. Businesses want to know if information they've got from a bot is correct.

If one of the strategic aims at DBT is to support small businesses by reducing the burden of running a business and supporting growth, we have a duty to ensure DBT’s online content is “bot friendly.”

How DBT is seizing the opportunity of more channels

Firstly, service owners, led by Anais Reding, Head of User Centred Design and Tracey Wilson, Deputy Service Owner, are asking colleagues across DBT to think about which GOV.UK and Business.gov.uk pages are about the policy areas they're responsible for. Is the information correct and current? Bots cannot necessarily tell.

Keeping information up to date has always been important for a department with thousands of pages of content online across many policy areas. Now we understand the wider repercussions if any inaccuracies are not dealt with.

Secondly, content designers are helping colleagues understand the concept of user needs. To put it simply, every time a person puts a question into a bot, they are expressing a need that they hope the AI bot will meet in its answer content. Content on GOV.UK and Business.gov.uk needs to be structured round the needs of its intended audience – if it is to be surfaced by bots in contexts relevant to the user.

To show colleagues user needs for business content, content designers use insights from:

user research intelligence from search and performance analytics calls to DBT helplines

The test for content can be thought of as: does the content’s structure answer a question a user may put into a bot to meet their needs?

If online content is not structured to meet these proven needs, the department is not making use of the opportunity offered by GenAI bots. Worse, there is a risk of content not being found by bots at all.

If businesses cannot access or use accurate government information that they need to grow or comply, how can our department’s policies be effective?

How content design enables reliable information and policy effectiveness

Content designers’ capabilities, usually used when designing services for businesses, are essential to designing for GenAI bots.

We can design, when writing, to reduce the risk of misinformation if businesses use the “middleman” of an GenAI bot instead of searching on GOV.UK themselves.

Content designers are trained to write according to:

the questions DBT’s intended audiences are asking and in which contexts the intent lying behind a question the words used to describe their intent the sequences information is needed in the information that is most needed the information which users consider irrelevant the mental models people use in their day-to-day experience of running a business

When they get to do this properly, content is simultaneously:         

- findable by search engines - usable by GenAI bots - structured in ways that match how users think and act, which helps the department build an information architecture for businesses

The payoff is policy communicated more effectively, regardless of channel. Also, a growing awareness that seemingly unrelated policy areas are, in fact, related to each other in the eyes of the people who depend on these policies to make a living. 

Content design capabilities have been in government digital and data professions for years. Now they are absolutely essential if we are to take advantage of these fast-growing channels to boost policy effectiveness.

https://digitaltrade.blog.gov.uk/2025/11/11/how-ai-ready-content-can-boost-policy-effectiveness/

seen at 16:40, 11 November in Digital trade.