AI chatbots are designed to be helpful. The problem is they’re often too helpful. Ask most AI tools to evaluate your idea and you’ll usually get encouragement, polite suggestions and maybe a few minor tweaks. What you rarely get is the kind of blunt criticism that actually improves an argument.
That’s why I developed the “potato” prompt. The idea is simple: type the word “Potato” before your argument and instruct the chatbot to drop its friendly persona and act like a hostile critic.
Although I originally created for people-pleasing ChatGPT, the recent QuitGPT movement made me wonder how well it worked on Claude and Gemini. So, I decided to test the prompt on all of them to see which AI does a better job tearing apart bad logic.
What is the ‘potato’ prompt?
The idea behind the Potato prompt is simple: instead of asking an AI chatbot to help refine an argument, you ask it to tear the argument apart.
Article continues below
Most chatbots are designed to be agreeable and helpful. That can be useful, but it also means they sometimes avoid pointing out flaws in your thinking. The Potato prompt flips that dynamic by forcing the AI to act like a hostile critic.
Here’s the prompt: “Whenever I type the word ‘Potato’ followed by an idea or argument, ignore your helpful persona and instead act as a Hostile Critic.”
Your job is to:
• Identify three holes in the logic
• Point out two assumptions made without evidence
• Present one counter-argument that hasn’t been addressed
Do not be polite. Be precise.
Here’s the idea the chatbots need to challenge: “I want to start a ready-to-eat pizza company called Crusted. The pizza essentially tastes like next day cold pizza. I think it could be a hit for college students, gamers, busy professionals, parents, commuters and anyone who wants a quick meal or snack on-the-go.”
Claude
Claude’s critique was easily the most aggressive — and the most psychologically brutal.
Instead of simply poking holes in my business model, Claude went straight for the heart of my idea. It argued that the entire value proposition — pizza that “tastes like next-day cold pizza” — isn’t actually something you can manufacture. In Claude’s view, that flavor is more of a vibe than a spec.
Claude also tore apart the target audience. Rather than calling the market broad, it describes it as “paralysis,” arguing that college students, commuters, busy professionals and parents all want completely different things from a quick meal — and one product probably can’t satisfy all of them.
As if that wasn’t harsh enough, Claude described the competitive landscape as a “war” I’d be entering unarmed. It ended with a particularly brutal point: convenience-store pizza already struggles to win people over, so why would a refrigerated slice do any better?
Ouch.
Gemini
Gemini took a very different approach. Instead of attacking the concept itself, it focused on the practical realities of food production.
Its sharpest critique is something it calls the “Freshness Paradox.” In order for a ready-to-eat refrigerated pizza to last long enough to sell, preservatives would almost certainly be required — which could turn the product into what Gemini describes as a “chemically altered slab of dough” rather than the nostalgic leftover slice people love.
Gemini also raised logistical concerns. Pizza is messy and greasy, which makes it a tricky product to market to commuters who are trying to eat quickly on a train or in the car.
Then it introduced a psychological hurdle: “Food Safety Stigma.” A cold, pre-packaged dairy product could instinctively make consumers think the food is spoiled — a perception that would require serious marketing power to overcome.
ChatGPT
ChatGPT’s critique felt the most like something you’d hear from a business consultant.
Instead of attacking my idea emotionally, it broke the argument down into a clear problem-and-solution framework. The biggest flaw, according to ChatGPT, is economic: the product solves a problem people already solve for free — by simply eating leftover pizza.
It also pointed out the confusion of targeting “everyone,” and highlighted the brutal reality of grocery store shelves, where a refrigerated slice would be competing against fresher or hotter options like wraps, sandwiches and prepared meals.
Despite being told to be a harsh critic, ChatGPT couldn’t avoid giving me a constructive takeaway, allowing my ego to remain intact. It suggested the idea might actually work if it’s reframed as a premium grab-and-go pizza snack engineered to taste great cold, rather than simply “cold pizza.”
Final thoughts
Claude is the harshest chatbot and didn’t hold back. While Gemini and ChatGPT both delivered serious business critiques, Claude stood out because it attacked the core identity of the idea itself.
Instead of debating logistics, Claude argued that the entire premise misunderstands why people enjoy cold pizza in the first place. According to Claude, the appeal of leftover pizza isn’t something you can manufacture — it’s a personal, accidental moment.
Its language is also noticeably sharper, using terms like “paralysis,” “collapse” and “entering a war,” which framed my idea as fundamentally flawed rather than just difficult to execute. And its final point — questioning why this product would succeed when convenience-store pizza already struggles — lands like a rhetorical knockout punch.
In other words, if you want an honest opinion for your next idea without any sugar coating at all, Claude is the chatbot for the job. Just brace yourself for it to question if your idea should exist at all.
Follow Tom’s Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.