The analogy

Walk into a shop and ask "what do you recommend?". The clerk, who knows nothing about you, will point you to the best sellers: an average answer, safe, good for the typical customer. If instead you say "I'm looking for a waterproof jacket for walking in the mountains in autumn, around a hundred euros, size M," they bring you three precise models. They didn't get better: you gave them the footholds to be so.

The AI works this way. To the vague question it answers with the "best-selling merchandise," the generic talk that pleases anyone. The more you narrow its field, the more you force it out of the generic and into your specific case.

How it really works

Faced with a request poor in details, the AI has no way of knowing what level of depth you need, nor which situation to apply itself to, so it chooses the most "central" answer, the one statistically suited to the greatest number of cases. It's a cautious behavior: better a general talk that might be useful than a specific detail guessed at that risks being beside the point. The generic, after all, is its safest bet when you don't give it enough to personalize.

What you can do in practice

  • Give it the concrete context: who you are, what you need it for, the constraints, the real numbers. "A training plan" becomes "a running plan for a beginner, three outings a week, goal 5 km in two months."
  • Ask for depth explicitly: "don't give me the generalities, go into the operational detail step by step."
  • Demand an example played all the way through: "give me a concrete example with real numbers, not an abstract case."
  • If the first answer stays on the surface, don't start over: come back with "too generic, anchor it to my case and add the details that are missing."

A common misconception

People think generic answers are a limit of the AI, a sign that "it can't go deep." Almost always it's the opposite: it could go deep just fine, but you asked it something that only a generic answer can honestly respond to. The generic isn't incapacity, it's the mirror of a question with no details. Change the question and the depth arrives, from the same exact model.

Frequently asked questions

Which details are worth giving first?

The goal (what you need the answer for), the recipient (for whom), and the concrete constraints (time, budget, level, numbers). They're the footholds that shift the answer most from generic to specific. Even just one of these changes the result a lot.

Can I directly ask it to be more detailed?

Yes, and it's effective, especially if you combine the request for depth with an example: "go deeper with a concrete case and numbers." Asking "be more detailed" on its own helps little if you don't also give it the material to be detailed about.

Is it better to ask many small questions or one big one?

For detail, it's better to start focused and then dig into one point at a time. A huge question gets a generic overview; a series of focused questions digs deep where you really need it.

If the AI stays generic, does it mean it doesn't know more?

No, and it's the misunderstanding that makes people give up too soon. In the vast majority of cases the generic is the answer to a generic question, not the ceiling of its knowledge. The proof is simple: add three details of your case and ask again. The same AI, on the same thing, will give you an answer that before seemed not to be there. The limit was in the question.