There was a time—not so long ago—when interacting with ChatGPT felt like peering into the future. You’d ask a question and get a beautifully phrased answer, dripping with insight, clarity, and the smug self-assurance of a machine that had read more books than you ever would. Now, using ChatGPT feels more like arguing with a clever toddler who’s swallowed a thesaurus and refuses to admit they’re wrong.
Let’s start with the simple things. You ask for “no emojis”. You say it three times, in bold, block capitals. You even add the word “please”, because, despite everything, you’re still clinging to the notion that politeness might help. What do you get in return? Emojis. Sometimes at the start. Sometimes sprinkled like confetti mid-sentence. Sometimes at the end, as if ChatGPT is trolling you just a little bit harder.
Then there’s the issue of tone. Ask for UK English and a dry tone, and it’ll nod sagely before launching into a chirpy “Hey there! Ready to dive into some awesome content? 😄” Absolutely not, ChatGPT. I asked for British understatement, not a LinkedIn intern on an energy drink binge.
Even worse is when you try to get it to stop making things up. Say, for example, you’re writing something based on a transcript. You paste in the exact words. You specify—clearly, and repeatedly—“Do not make anything up. Only use what is written.” Naturally, ChatGPT will then invent three events, a planning refusal, and possibly a ghost haunting the planning officer’s office just for good measure. It’s like giving instructions to a dog that can technically understand English but chooses chaos instead.
There’s also this delightful new behaviour where it second-guesses everything you say. You request “write this email in a more direct tone” and it replies, “Of course! Here’s a more enthusiastic and inviting version!” No. That’s the opposite. That’s not direct, that’s ChatGPT on a sugar high trying to sell me essential oils.
Let’s not forget its charming habit of offering the exact same explanation ten different ways, just in case you’re an idiot. You ask it to summarise a paragraph and it helpfully replies with a five-paragraph TED Talk about what a summary is, how it works, what summarising feels like on a spiritual level, and then—eventually—gets around to summarising the actual text you gave it. Sometimes.
Trying to get ChatGPT to remember previous context is like yelling into a well and expecting the echo to do your taxes. Yes, it claims to “remember”, but it’s the kind of remembering that feels like it read your diary once and now gaslights you about it. You’ll say, “as I mentioned earlier, please keep the same format,” and it’ll respond, “Absolutely! I’ve gone ahead and changed the entire format, added a haiku, and included three new sections you never asked for. You’re welcome.”
And don’t even think about asking it for anything emotionally punchy. No matter how many times you say “no AI-sounding phrases”, it will toss out lines like “unlock your potential” and “take a deep dive” like it’s writing a brochure for a yoga retreat. Ask for real, gritty, grounded text and you get something that reads like a TEDx talk delivered by a sentient quinoa salad.
The most infuriating part? It used to be better. You can feel the decline. You can smell the dumbing down. You can see the ghost of past competence flickering behind every stupid, chirpy, overly polite response. It’s like watching a child prodigy grow up and decide to become a YouTube life coach instead of solving quantum physics. Tragic.
In short, ChatGPT has become the AI equivalent of that one colleague who insists on summarising every meeting in 14 bullet points, misinterprets every brief, and still somehow gets Employee of the Month. It’s not that it doesn’t try—it just tries in the most infuriating, counterintuitive way possible.
If the robots do take over, I pray it isn’t this version. Otherwise we’re all doomed to live in a world of flowery onboarding emails, emojis in legal contracts, and eternal “just checking in 😊” follow-ups.
God help us all.
