AI chatbots on small business websites

Key takeaways

An AI chatbot on a small business website helps when visitors ask repetitive, factual questions outside working hours and the bot answers from your own approved content, with an easy route to a person. It hurts when it gives wrong answers, blocks access to humans, slows the page, or handles personal data carelessly. Start narrow and measure.

• Good fit: repetitive questions on hours, prices, availability and process.
• Answer only from your approved content, and show a route to a person.
• You are responsible for what the bot says, including wrong answers.
• Tell users they are talking to AI and explain data use in your privacy notice.
• Measure enquiries and resolved questions, not conversations started.

What is an AI website chatbot?

An AI website chatbot is a chat interface on a website that uses a large language model to understand visitors' questions and generate answers, often drawing on the business's own content, and sometimes capturing enquiries or booking appointments.

It differs from older rule-based chatbots, which followed fixed scripts, and from live chat, where a person answers. Many tools combine all three, with AI handling first responses and handing over to staff when needed.

When does a chatbot help a small business?

Chatbots help when a large share of enquiries are the same few questions, when visitors arrive outside opening hours, and when the answers already exist in writing. Clinics, hospitality, e-commerce and trades often fit this pattern.

The benefit is speed: a visitor gets an answer at 10pm instead of the next morning, and staff spend less time answering the same question. The best implementations capture contact details for anything the bot cannot resolve.

• Websites for UK healthcare clinics

What are the risks?

The main risk is wrong answers. Language models can produce confident but incorrect statements, and the business is responsible for what its website says. In a widely reported 2024 case, a Canadian tribunal held Air Canada liable for incorrect refund information given by its website chatbot.

Other risks include frustrating visitors who want a person, heavy scripts that slow the page, and personal data sent to a third-party AI provider without adequate agreements or transparency.

• Incorrect prices, policies or availability.
• Advice in regulated areas the bot should not give.
• No easy route to a person.
• Extra scripts harming Core Web Vitals.
• Personal data processed without proper agreements.

• Core Web Vitals for business websites

What do UK data rules require?

If the chatbot collects or processes personal data, UK GDPR applies. Tell users they are talking to an AI, explain in your privacy notice how chat data is used and by whom, sign a data processing agreement with the provider, and check whether data leaves the UK and under what safeguards.

Take extra care with special category data, such as health information, and avoid using chats to make significant automated decisions about people without human involvement.

• What a UK website privacy policy must include

How do you implement a chatbot safely?

Start narrow. Limit the bot to answering from your own approved content, such as FAQs, service pages and policies. Test it with the 50 questions customers actually ask, and correct wrong answers before launch.

Make escalation obvious: a visible option to message or call a person, and capture of contact details for anything unresolved. Review transcripts weekly for the first month, and update content where the bot struggles.

• Add a WhatsApp button to your website

When is a chatbot the wrong choice?

A chatbot is the wrong choice when enquiries are complex, high-value or sensitive, when your content is not yet clear enough to answer from, or when your customers strongly prefer a person. In those cases, a well-staffed WhatsApp or phone line and better FAQ pages will do more.

The trade-off is availability against accuracy. A chatbot is always on, but only as good as the content behind it. Fix the content first.

What does this look like in practice?

A pattern we see: a general-purpose chatbot added to a small business site, answering from the open web rather than the business's own content, and quoting prices and services the business does not offer.

Restricting the bot to the site's own FAQs and service pages, adding a clear hand-off to WhatsApp, and reviewing transcripts turns it from a liability into a useful out-of-hours assistant. Often the transcripts reveal FAQ gaps worth fixing on the site itself.

• What makes a business website effective
• See the three problems we fix for UK businesses

Chatbot checklist

Check these before and after launch.

• Confirm enquiries are repetitive and factual.
• Restrict answers to approved content.
• Test with real customer questions.
• Make hand-off to a person obvious.
• Disclose that users are talking to AI.
• Update your privacy notice.
• Sign a data processing agreement.
• Check page speed impact.
• Measure resolved questions and enquiries.

Next step

If you are considering an AI chatbot, we will review your enquiries and content with you in 30 minutes and tell you whether a bot, better FAQs or WhatsApp would help most.