Articles

Why does AI give me such generic answers?

Short answer

Because you asked a question that would suit anyone, so you got an answer that suits anyone. A text AI completes whatever you hand it: if you do not say who you are, who you work for and what result you need, it fills those gaps with the average of everything it has read — and the average always reads like a brochure.

The mechanism, in one sentence

A text AI is not looking up a stored answer. It builds the sentence word by word, and at each step it only decides what best fits what is already written. Your request is the start of that sentence. If the start is vague, every likely continuation is vague too.

That is why "write me a post about productivity" and "write me a post about productivity for self-employed people who lose their mornings to email" are not the same request with more words. They are two different starting points.

The three signals that fix it

You do not need to fill in a form. Three signals change the result:

  1. Who you are and who it is for. "I am a physiotherapist writing to patients who have just had knee surgery."
  2. What result you need. Not "tell me about", but "a five-line email that gets them to confirm the appointment".
  3. How it should arrive. Length, structure, a tone described in checkable terms, and how it should end.

The third is the one almost nobody gives and the one that saves the most time: if you have to reorder the answer before you can use it, format was missing from the request.

A real before and after

Before
"Write me something announcing a price rise."
After
"I am a carpenter putting prices up 5% in January because material costs have risen. Write the notice for my regular customers: friendly, five lines maximum, no apologising, and end by reminding them that quotes already sent still stand."

The second is not longer for the sake of it. Every part of it removes a decision the AI would otherwise have made on your behalf, and almost always badly.

The mistake that keeps the problem alive

Hunting for the magic prompt. There is no secret phrase that rescues a request with no context in it. What separates someone who gets value out of AI from someone who gives up is not a formula: it is understanding what information the model needs in order to get it right, and handing that over.

And when an answer lands half-right, correct it instead of starting again. Saying what is missing — "be more specific in the second paragraph", "drop the exclamation marks", "this is not my tone, here is an example of mine" — nearly always gets there faster than rewriting the whole request.

If an answer reads like a brochure, the tool is not the problem: the question would have suited anyone.

The lessons that practise this

This article sums up what you practise in these lessons of the course.

Keep reading

Reading it is not the same as doing it

This takes three minutes to read; making it automatic takes practice. The course is 20 short lessons, and the first 3 are free.

Start with the first lesson

See all the articles