Why the AI Gave You a Generic Answer, and the Follow-Up Questions That Fix It

You ask ChatGPT for advice on switching careers, and it hands you five bullet points that could apply to literally anyone on earth. Update your resume. Network. Consider your transferable skills. Thanks for nothing. It's easy to read that and decide the AI is shallow. But the blandness is a mirror. A generic question gets a generic answer, and the fix is almost never a cleverer opening line. It's the back-and-forth after.

This piece is about the conversation, not the prompt. Anyone can write a good first message. The people who get genuinely useful work out of these tools are the ones who treat the reply as a starting point and keep steering.

Why the first answer is almost always mediocre

An assistant knows nothing about you except what's in front of it. When you ask a vague question, it has no way to guess whether you're a 22-year-old barista or a 50-year-old accountant, so it aims for the middle and gives you something safe. Safe is bland. That's not a flaw so much as the only honest thing it can do with too little information.

The moment you add specifics, the answer changes character. Compare "how do I get better at public speaking" with "I have to present quarterly numbers to 40 people in two weeks, I freeze when I lose my place, and I have about three hours to prep. What should I actually do?" The second gets you a rehearsal plan, not a listicle.

The single most useful habit: dump context

Before you refine your question, tell the assistant who you are and what you're actually dealing with. You don't need to be elegant about it. Overshare. It's a machine; it won't judge the mess.

A weak request: "Help me write an email declining a meeting."

A strong one: "I need to decline a recurring 8am meeting my manager set up. I don't want to seem uncommitted, but the time genuinely wrecks my mornings with my kids. My manager is direct and hates long emails. Keep it short, warm, and propose an alternative time. Don't be groveling."

Same task. The second produces something you could send. The details that felt like too much information are exactly what made the answer usable: the relationship, the real reason, the tone your manager responds to, the constraint on length.

Treat the first reply as a draft to argue with

Here's where most people stop and where the value actually lives. When the answer isn't right, say so specifically. Vague dissatisfaction gets you vague improvements. Precise corrections get you precise ones.

Useful follow-ups sound like:

• "This is too formal. Rewrite it like I'm talking to a colleague I like."
• "The second option is closest. Drop the others and give me three variations of that one."
• "You're assuming I have a budget. I don't. Redo this for zero dollars."
• "That's the generic advice. What's the part people usually get wrong?"
• "Cut it in half and lead with the most important point."

That last category matters. When an answer feels like the obvious stuff you already knew, ask directly for the non-obvious part: "What would someone experienced tell me that a beginner wouldn't think of?" It forces the assistant past the safe middle.

Make it ask you questions

One of the most underused moves: flip the direction. Instead of front-loading everything, tell the assistant to interview you.

"Before you answer, ask me the three or four questions you most need answered to give me genuinely tailored advice."

Now it might ask about your timeline, your budget, your experience level, what you've already tried. You answer, and the response it builds is shaped around your situation instead of a hypothetical one. This works especially well for anything with a lot of variables: planning a trip, choosing between options, troubleshooting something that could have ten causes.

Give examples of what good looks like

If you want output in a particular style, show it rather than describe it. Paste an email you liked and say "match this voice." Paste a paragraph you wrote and say "more like this, less stiff." Assistants are far better at imitating a sample than interpreting an adjective. "Professional but friendly" means something different to everyone, including the machine. A concrete example removes the guesswork.

Keep the context in one conversation

A running chat remembers what you've told it. If you spend a few messages explaining your job, your constraints, and your taste, don't start a fresh chat for the next related question. Stay in the thread and it carries all of that forward. Open a new one and you're back to explaining yourself from scratch.

For things you repeat often, set it once and forget it. ChatGPT lets you save details about yourself in settings so it stops asking. Claude's Projects let you keep a standing note of context that applies to every chat inside them. Gemini can draw on your Google account if you allow it. Whichever you use, a little permanent context means shorter conversations that still land.

Know when more context won't save you

Follow-ups fix bland answers. They don't fix wrong ones. If the assistant states a fact confidently and it matters (a date, a figure, a legal or medical claim, a citation), no amount of steering makes it reliable. It can sound equally sure whether it's right or inventing. For anything where being wrong has a cost, use the AI to draft and organize your thinking, then verify the specifics yourself. Asking "are you sure?" sometimes gets a correction and sometimes gets a confident doubling-down. It's not a fact-checker.

Context also can't rescue a question you haven't thought through yourself. If you don't know what a good outcome looks like, the assistant can't read your mind. Sometimes the most useful follow-up is one you ask yourself: what would actually solve my problem here? Then you can tell it.

The short version

The quality of what you get out is downstream of what you put in and how hard you steer afterward. Dump more context than feels necessary. Correct specifically instead of vaguely. Make the assistant interview you when the situation is complicated. Show examples instead of describing them. Stay in one conversation. And keep your skepticism switched on for the facts that matter. Do that and the tool stops feeling like a search engine that talks, and starts feeling like something that's actually thinking alongside you.