We hope that headline looks obvious. You can’t be trained by AI, can you? And yet that is one of the objections prospective clients have raised when their PR team has proposed media training with Clapperton Media.
There are a few things AI can do extremely well, of course. Tell Claude, Perplexity or any of the others that your client is going into an interview with the Times, Guardian or whatever paper you want and ask it what to prepare for and the chances are you’ll get something realistic out of it even if it’s not completely comprehensive. Ask those LLMs for a rough breakdown of the readership and it will find something that’s been reported so it can give you a feasible-sounding response. Often, and increasingly so, it will be correct, although it will sometimes repeat old information or blatant nonsense if someone else has published it, it has no way of telling.
It can even role play an interview and you can talk to it by voice so it will feel more real. There really is a lot it can do as long as you’re positive that your interview will be polite, uncritical and with pretty much no pushback. AI is designed to reassure and that’s where it starts to fall down. It can also offer pretty generic advice and that’s where you end up with a generic set of quotes from a spokesperson who ends up forgettable as a result.
Why experienced journalists matter in media training
It has other faults too. AI is a great distiller of everyone else’s experience but by its very nature it has no experience itself. A good media trainer will have experience as a journalist and this means they will have faced loads of variables. Our founder Guy Clapperton alone has come across:
- The spokesperson who didn’t think they needed to answer the journalist’s question (of course there is no legal obligation but why, then, are you doing the interview..?)
- The chief executive who said they didn’t want to comment on something but demanded that this should not go in as a no-comment (er….)
- The sales director who said after an interview that they’d need to check everything the journalist wrote and said they had entered the interview on that understanding (Guy had no option but to explain that there had been no agreement or understanding prior to the interview and the answers offered had been perfect – but he had a deadline and an editor who would criticise heavily if they felt the answers had been vetted)
- The director who told him exactly what was going on and then tried to “amend” this with the “official version” – in other words, decided he’d rather the press knowingly published a lie.
AI may anticipate these issues and will certainly advise against the underlying assumptions but it won’t interrupt as an actual journalist would. A good trainer would stop the interview and explain the situation; AI will wait its turn, which is not how an actual interview would go.
Why AI’s rapid correction can backfire
The other thing we’ve noticed about AI, which is admirable in many ways, is its willingness to admit to errors almost immediately. Thing is, we only know they’re errors because of experience. Literally on the day of writing we asked one of the LLMs for a steer on a particular journalistic practice and it said one feature of a normal interview was that people got to check what they’d said afterwards. We asked “what makes you think that?” and it replied immediately that we were right, it shouldn’t have said it, which is great as long as you have a human to check. To someone who didn’t know the correct version, the incorrect one might have seemed quite reasonable. You could also have some fun here: if you’re checking an AI’s answer, push back on something it got right and see whether it holds its ground or concedes immediately.
AI is an excellent support
There could be a number of reasons prospective clients like the one in the opening tell their PR team they could just “use AI”. Some of it may simply be time or budget, which is fair enough. Some, though, may genuinely believe AI is a substitute for an experienced human or two and that is a job it isn’t suited to. AI can be a terrific support in all sorts of places but if someone starts to use it instead of human beings then first they run the risk of sounding like everyone else who’s using the same prompts, and second they could well end up acting on completely unchecked nonsense. Use it for background and as an extended colleague for batting ideas around by all means – we do – but always use an actual trainer for practice and sanity checks.