She explained: “It’s so new, it’s exploded onto the scene and a lot of people don’t really know how it works. There’s this race to get it embedded and be an early adopter, but it also feels like you can’t ignore it and you can’t be left behind.”

Timperley said that AI “is not thinking for itself”, so it couldn’t write a piece of content in the same way that a human can.

“So, if you want to write a piece of content, you might have your concepts floating around in your head and then you try and put words to that and hang those words on the idea.

“AI has a data training set of 300 billion words, and to put that into context, the oldest estimate of the known age of the universe is about 27 billion years, so we’re talking about a number that is 11 times bigger than that,” she noted.

Timperley said that AI was taking all of its training data, then the prompt, and predicting word-by-word “what is the most probable word to come next, so it’s flying blind”.

“It’s working on probabilities. When you do your prompt, you have to bear in mind that you are fighting against the weight of all the biases baked into that training data,” she added.

As an example, she said that if you put a mortgage-related question into ChatGPT, but do not qualify in the UK, the answers will tend to be more US-centric, which may not be relevant.