Not blindly. When GPT-5 came out with the promise of “PhD-level” intelligence and far fewer hallucinations, my first prompt asked how many b’s are in “blueberry”. It said three. Twice. With total confidence.
AI is really, really good at predicting the next word – and at selling you the illusion that it’s intelligent. It’s useful for lots of tasks, but it won’t consistently give you high-quality output on its own.
Use it to AUGMENT your creative process, not replace it – and get really good at telling it what to do (and what NOT to do).
When OpenAI unveiled its latest, greatest AI model…
GPT-5.
“A total game-changer”, according to Sam Altman, OpenAI’s CEO.
Naturally, most of the world was pretty pumped about this announcement…
And I also wanted to check it out as soon as possible.
In my first-ever GPT-5 prompt, I gave it a simple task:
“How many b-s are in the word blueberry?”
You know, something that a PhD-level expert would easily be able to answer.
(unless they’re piss drunk, spinning on their head… but chances are, even then!)
GPT-5’s answer?
Huuuh, interesting 🤔
Free book + AI assistant
Turn the Leads Already Joining Your List Into Buyers – Faster
The Rapid Lead Activator reveals the psychology-based framework I use to write welcome sequences that sell – without hype. Normally $9.99 on Amazon, yours free. Plus a bonus AI assistant that writes your first 5 welcome emails for you – in minutes, not days.
Send Me the Rapid Lead Activator →What happened when I asked GPT-5 again?
Sam Altman specifically highlighted – multiple times – how this new model is supposedly FAR less likely to make mistakes, lie, or hallucinate.
And yet…
According to GPT-5, there are 3 b-s in blueberry.
“Ok, maybe it’s a one-time error”, I thought.
“Maybe it made a mistake because the model is so new and there are still bugs to be ironed out”.
So I asked the following:
Hmm….
Lo and behold, 3 b-s, again.
It seems like 100% right when you say “total b s”, GPT-5 😂
Because that’s total BS right there!
Why does a confident AI make you doubt yourself?
At this point, for a brief moment, I was like:
“Am I in the wrong here?
Surely the latest shiny whiz-bang AI model – the BEST one of the world’s most valuable companies can offer – can’t be this stupid??!!”
Especially since this PhD-level expert sounds soooo confident.
So, for a moment, I seriously started doubting myself…
Re-reading the word “blueberry” 3 times to make absolutely sure there aren’t 3 b-s there after all.
I mean…
Famous psychology experiments (Solomon Asch’s conformity studies) showed that people – even smart ones – can go along with clearly incorrect answers when everyone around them sounds sure.
But no matter how many times I looked at it…
Blueberry only has TWO (2) b-s!
Not 3.
Two.
And yet GPT-5 was super-duper confident in its abilities.
What is AI actually good for in marketing?
Now look…
OBVIOUSLY, overall this new model is probably smarter, better, faster, and more consistent than the previous one…
But – as you can clearly see – it’s far, faaaaaar from perfect.
It’s still not even close to “real” intelligence…
It’s just really, really good at predicting the next word… and selling you the illusion that it’s actually intelligent.
That doesn’t mean it’s not useful for lots of tasks…
Or it won’t revolutionize more and more industries…
But don’t make the mistake of thinking you can simply tell it to do stuff for you, and it will consistently give you high-quality output.
It most likely won’t.
That is…
Unless you’re really, REALLY good at telling it what to do.
(and perhaps more importantly, what NOT to do!)
That’s exactly what I’ve been experimenting with:
Finding almost dangerously effective ways of using AI to – not replace – but AUGMENT my creative process and workflow.
And this journey has given me countless game-changing insights…
“AHA moments”…
And “HOLY SHIT WOW”-style revelations.
No, I still don’t use it to write my emails…
But I HAVE developed several systems that can help me with tasks like:
– Strategizing whole email campaigns
– Coming up with new, different ideas I would have never thought of
– Critiquing my writing and suggesting ways of improving it (I especially LOVE this part)
– Helping me find gaps in messaging, campaigns, and offers
And many other things like that.
But THE most powerful use-case I’ve discovered?
Using the systems I built to help my private clients strategize, create, and write their emails in 1/50th the time it used to take them (at an 80% “good enough” level)…
…with me coming in as the trusted advisor to iron out the remaining 20%
(especially useful for people who are not pro copywriters!)
How to use AI without getting fooled
1. Never take a confident answer as proof. Check anything that matters – facts, numbers, claims.
2. Don’t ask it to “just do it”. Tell it exactly what to do – and what NOT to do.
3. Use it to augment, not replace. Strategy, new ideas, critique of your writing, gaps in your messaging.
4. Keep a human on the last 20%. AI can get you to “good enough” fast. The final polish is yours.
(Why AI copy falls flat without that foundation: the “paper house” problem.)
If you want your AI to “think” with better inputs, the Conversion Operating System (COS) gives it tested conversion strategies and marketing frameworks to work from.
Free book + AI assistant
Turn the Leads Already Joining Your List Into Buyers – Faster
The Rapid Lead Activator reveals the psychology-based framework I use to write welcome sequences that sell – without hype. Normally $9.99 on Amazon, yours free. Plus a bonus AI assistant that writes your first 5 welcome emails for you – in minutes, not days.
Send Me the Rapid Lead Activator →-Csaba
