Why AI Gives You Generic Answers (and the Role Fix)
- Tomasz Dylik
- Jul 19
- 3 min read
Updated: Jul 31
Ask a good question with no direction and you get the most average answer in the room. That's the default setting of every AI: safe, middle-of-the-road, technically fine, and forgettable.
There's a fast way out of average, and it takes about four words. You tell the model who to be. In the CROFTC framework this is Role, the second input, and it's the one that turns a generic assistant into a specific kind of expert.
What Role actually is
Left alone, a model answers as a blend of everything it has ever read. That's why the default voice feels like a committee: it's the average of a million sources with no point of view. When you assign a role, you narrow that blend to one perspective. "You are a senior email copywriter" and "you are a helpful assistant" pull from very different parts of what the model knows, and they produce very different work.
Role isn't a costume. It's a filter. It tells the model which standards to hold the answer to, which vocabulary to reach for, and what "good" looks like in this specific job.
Why skipping it flattens everything
Without a role, the model doesn't know how expert to be, so it defaults to broadly helpful and lightly informed. You get advice that would satisfy nobody in particular. A financial answer that skips the caveats a real advisor would insist on. Copy that reads like every other piece of copy. The information might be correct, but the judgment behind it is missing, because you never told the model whose judgment to borrow.
The fix costs nothing and changes the whole reply, which is why it's the highest return on the fewest words in the entire framework.
How to assign it well
A good role is specific and loaded with implied standards. "A writer" is weak. "A direct-response copywriter who's launched dozens of products" is strong, because it drags in a whole way of thinking. The more precisely you name the expert, the more of their instincts you get for free.
Here's a prompt with no role:
Review this landing page copy and suggest improvements.
And the same request with a role:
Role: You are a conversion copywriter who has audited hundreds of landing pages and cares about one thing: turning readers into buyers. Review this landing page copy and suggest improvements.
The first gives you tidy grammar notes. The second questions the headline, the offer, and whether the call to action earns the click, because that's what that person would actually care about. You didn't add information. You added a point of view.
Want prompts with the right expert already built in? My free bonus pack includes 500+ ready-to-use prompts plus the one-page CROFTC cheat-sheet, each one written to pull real expertise out of the model. Get the free prompt pack.
Role works with the rest
Role decides who answers; the other five inputs decide what that answer contains. Context gives the expert a situation to work in, the Objective points their expertise at a goal, the Format shapes how they hand it over, the Tone tunes how they sound, and Conditions keep them honest. On its own, a role sharpens the reply. Inside CROFTC, it sharpens a reply that's already aimed right.
If you change one habit, make it this: never ask a real question of a nobody. Spend four words naming the expert you want, and let the model answer as them.
CROFTC is the structure behind every one of the 1,350+ prompts in my series, and Role is what gives each one a point of view. If you want a running start, grab my free bonus pack: 500+ ready-to-use prompts plus the one-page CROFTC cheat-sheet, so you can stop settling for average and start getting expert answers.
Role changes the answer more than any other input for the least typing. The other five work the same way, and I put all six in one place, with the prompts already written, in AI Prompt Engineering Bible (7 Books in 1). Seven books in one volume, each input taken apart and rebuilt, with prompts you can copy and adapt to your own work.
The full CROFTC series
This is one input in a six-part framework. Here's the whole series:
Comments