Thameslink is testing ChatGPT to guide passengers when trains are delayed

What Thameslink is testing

One of Britain's biggest train companies, Thameslink, is trialling a way to let ChatGPT give passengers travel advice when services are disrupted. As Yahoo reported, the advice is delivered through WhatsApp, the messaging app. Thameslink was nationalised in May.

The idea is simple. When a train is delayed or cancelled, a passenger asks what to do and gets an answer for their own journey, instead of hunting through several websites and apps to piece it together.

The plan came out at the Rail Industry Association's annual conference last week. Chief executive John Whitehurst was speaking in a discussion about keeping passengers informed of alternative routes. He joked that IT staff keep telling him he cannot talk about the project.

How the chat would work

Passengers would type questions in everyday language, as they would to a friend. They could then ask follow-up questions in the same conversation, such as what the alternative is if the first suggestion does not work.

Thameslink's parent operator, Greater Thameslink Railway (GTR), says the service is built to draw on GTR's own and other trusted rail information. It is not meant to rely on whatever ChatGPT already 'knows' from its general training. That distinction matters because a chatbot working from general knowledge can sound confident about things it has no way of checking, such as today's live timetable.

The company also says the tool is designed to say clearly when it cannot give a reliable answer, rather than guess. That is a design goal, not a proven result, and the trial exists to test it.

Still a staff-only trial

Nobody on a platform can use this yet. Thameslink confirmed it is testing the service, but only with an invitation-only group of staff. No decision has been made about a public rollout.

A spokesman said the plan is to complement rather than replace existing sources such as official social media accounts, apps and websites. So the chatbot would be one more way in, not the only one.

Mr Whitehurst was candid about the stage the project is at. He said the company is not there yet, but that bringing the technology together offers a big opportunity. He also pointed out that there is always a cyber security threat when a company adds technology like this.

The accuracy problem

This move comes after years of warnings that AI systems can produce false, misleading and sometimes harmful information. Rail makes the risk easy to see. Delay advice can change minute by minute, and an out-of-date answer might send someone to the wrong platform or make them miss a connection.

The spokesman said accuracy and reliability are a key part of what the company is assessing before any decision. Two things will probably decide whether the trial succeeds. One is how fresh the data feeding the chatbot is. The other is how often the tool admits it does not know instead of inventing an answer.

There is also a convenience question. The company says it is assessing whether putting information into one conversation gives customers a simpler, more tailored experience than the sources they have today. If the answer is no, the tool has little reason to exist.

What to take from it

For most readers, the lesson goes beyond trains. AI is arriving in ordinary services, often through apps people already have on their phones. It is most tempting to trust at the moment you are most rushed.

A sensible habit is to treat a chatbot's travel answer as a first suggestion. Ask it for the next service and who to contact, then compare it with the official station board or the operator's own app. Regular commuters have an advantage here, because they can tell quickly whether the advice matches what actually happens on their route.