Skip to content
POLARISAI ACADEMY
Basics · Prompting · May 2026

Stop using AI as Google

You ask questions and expect answers. Give the AI a task instead of a question and you will more often get what you want.

BasicsPromptingBeginnersContext
Two stone benches facing each other in a quiet moss-stained concrete courtyard.

The most common mistake

Most people write search queries to AI. "Best way to market on Instagram", "pasta carbonara recipe", "what is inflation". The questions are fine. The trouble is that AI does not work like a search engine.

This difference is the reason why most people who use ChatGPT, Claude, or Gemini every day are still never really satisfied with the results. They get answers, but not help with the actual task.


Why it is the wrong mindset

Google indexes the internet and ranks pages. A search engine takes your query, finds the best existing answer, and gives you a link. It's a retrieval system. It is built to give you what already exists.

AI generates answers based on patterns in training data. That has a practical consequence. AI cannot fetch the right page for you — but it can think with you, write for you, analyze your material, argue against your own assumptions, and produce something that didn't exist before you asked. So it is better at producing new material than at finding pages that already exist.

A search engine gives you a link to an answer. AI can be your assistant — but only if you treat it as one.

A short question gets a generic answer. Context about you and the task gets an answer you can use.


What AI actually is — a thinking partner

Think of it as an extremely knowledgeable colleague who has never met you, never knows what you work on, and never understands your context — unless you tell them. Every time you open a new conversation, you start from zero. There is no background, no notes from the last time, no context taken for granted.

That is how the technology works. Once you know it, you stop typing short search queries and start giving it context.


Three concrete examples — bad vs. good prompt

Same principle, three different situations. Read through and see if you recognize yourself in the column on the left.

Individual — plan a trip

Bad prompt
Tips for traveling to Japan
Good prompt
My partner and I are going to Japan in October for 12 days, budget 40,000 SEK excl. flights. We love food, architecture, and hiking but hate tourist traps. Help us plan a route including Tokyo and at least one rural region.

Entrepreneur — write a quote

Bad prompt
Help me write a quote
Good prompt
I run a small IT consulting firm with 3 employees. I need to provide a quote to a manufacturing client (50 employees) who wants help automating their inventory reporting. The project takes about 6 weeks. Write a quote letter in Swedish — formal but not stiff, max 400 words, emphasize our experience with industrial customers.

Creator — write headline

Bad prompt
Give me a good headline
Good prompt
I have written an article about how most people don't know how to prompt AI. The target audience is Swedish small business owners, aged 35–55, who have started using ChatGPT but don't quite understand why the results are mediocre. Give me 5 headlines — direct, slightly provocative, no clickbait fluff.

Do you notice the pattern? The good version tells who you are (in that specific context), what you want (concrete output), and what constraints apply. That's it. AI does the rest.


One rule: give context

Give context, not just a question. Three things are enough:

  1. 01
    Who you are — in context

    Not who you are as a person. What your role is in this specific task. Business owner with 3 employees. Private individual planning a trip. Creator writing for a specific target group.

  2. 02
    What you want — specifically

    Not "help" or "tips." A quote in Swedish, max 400 words. An itinerary including Tokyo and a rural region. Five headlines, direct, without clickbait. Specificity yields specificity in return.

  3. 03
    What constraints apply

    Time, tone, format, target audience, budget. Each constraint tells the AI more about the answer you want. An AI without constraints produces average answers. With constraints, it produces your answer.

Most people skip this step. They continue to write search queries and wonder why the results are mediocre.

AI without context is like asking a newly hired consultant to solve a problem without telling them what the company does.


If you want to go deeper into exactly this — how you build prompts that actually work, not just for simple tasks but for real workflows — the Prompt Course is available here on the site. 5 free lessons: lessons 1–3 open right away, lessons 4–5 by email.