Guide to recognising political prompts

Spotting a politician who overuses ChatGPT has become a minor exercise in literary criticism. There’s no magic formula for identifying it. A good politician can write like a chatbot, and a good chatbot can imitate a politician. But there are a few clues: suddenly, everyone seems to have attended the same course in rhetoric.

The number one tell-tale sign is the phrase that aims to sound profound without taking any risks. “It is not a question of choosing between growth and sustainability, but of building sustainable growth.” “The challenge is not to stop change, but to manage it.” They work, they sound good, they mean almost nothing. AI loves these structures because they are orderly, symmetrical and difficult to challenge.

Second clue: the triads. Innovation, inclusion, competitiveness. Security, growth, cohesion. Courage, responsibility, vision. Three words fit almost anywhere, especially when you don’t know what to put next. The problem arises when every paragraph seems designed to become a slide. Then there’s the overblown language. “In a rapidly evolving global context”. “Faced with the complex challenges of our time”. “A pragmatic and forward-looking approach is needed”. It’s the perfect fog: no one can say it’s false, yet no one can remember it five minutes later.

Another tell-tale sign is the sudden disappearance of human error. A genuine politician has mannerisms, favourite phrases, obsessions, awkward syntax and questionable metaphors. Artificial text, on the other hand, tends to smooth out the rough edges. Everything is correct, everything flows smoothly, everything is presentable. After three minutes, you can’t remember a thing. It’s like a room in an international hotel: clean, efficient – it could just as easily be in Singapore as in Frosinone.

Then there is compulsive caution. Every statement is immediately qualified. Every argument already contains its own counter-argument. Every judgement is couched in phrases such as ‘whilst acknowledging’, ‘without underestimating’ and ‘at the same time’. But when this becomes automatic, it produces a discourse in which no one ever risks being wrong because no one is actually saying anything.

The real wake-up call comes when you ask for a concrete example. The text generated by the AI can speak eloquently about school, work, young people and the future. But then all you have to do is ask: which school? which worker? which law? what unpopular decision would you make tomorrow morning? And suddenly the fog either lifts or thickens. The AI is formidable when it comes to the big picture, but less so when it comes to assigning specific responsibility within that picture.

That is why the problem is not that a politician uses ChatGPT. It would be rather ridiculous to expect them, in 2026, to continue writing every press release themselves, as if autocorrect were a betrayal of democracy. They can use it to summarise, identify potential objections, polish a text, prepare alternatives or scrutinise a draft. The key is knowing who is using whom.

The best rule is this: if the AI makes the politician’s message clearer, that’s good. If it makes them indistinguishable from others, that’s bad. If it enables them to express their thoughts more effectively, it is a useful tool. If it allows them to avoid thinking about what they are saying, it has become a problem.

A politician who overuses ChatGPT eventually becomes unrecognisable because they say ‘fundamental’, ‘strategic’ or ‘resilient’ five times over. You recognise them when, whilst listening, a simpler suspicion arises: everything here is very well written, but there’s no one there.