How do I give AI better context?

Use the 3-layer context framework: Layer 1 is identity (who you are, how you think, what you value). Layer 2 is situation (what you're working on right now, your constraints, your goals). Layer 3 is task (what specifically you need help with). Most people only provide Layer 3, which is why AI gives generic answers.

Why does AI give generic answers?

AI gives generic answers because it has no context about who is asking. Without knowing your role, expertise level, constraints, and goals, AI defaults to the safest, most broadly applicable response. Adding even basic context — your job, your goal, your constraint — dramatically improves response quality.

What should I put in AI custom instructions?

Your custom instructions should include: your role and expertise level, your communication style preferences, your common use cases, and any constraints AI should respect. For a deeper approach, use a soul archive — a file that captures your full identity and context from your own writing.

How to Give AI Context So It Actually Helps

April 8, 2026 · 5 min read · By Nbidea

You ask AI a question. The answer is technically correct and completely useless. It reads like a Wikipedia article written for nobody in particular.

This happens because AI doesn't know who it's talking to. Fix that, and everything changes.

The 3-Layer Context Framework

Every interaction with AI has three layers of context. Most people only provide one.

Layer 1: Identity — who you are

Your role. Your expertise level. How you think. What you value. Your communication style.

This layer almost never changes. A senior engineer and a first-year student asking the same question should get completely different answers. Without Layer 1, they get the same generic one.

Layer 2: Situation — what you're dealing with

Your current project. Your constraints. Your deadline. Your team. The decision you're facing.

This layer changes weekly or monthly. "I'm launching a product next Tuesday with a $0 marketing budget" produces radically different advice than the same question without that constraint.

Layer 3: Task — what you need right now

The specific question. The specific request. "Write this email." "Review this code." "Help me decide."

This is the only layer most people provide. And it's the least important one.

A question without context gets a textbook answer. A question with identity and situation gets an answer that's actually useful.

What This Looks Like

Without context:

"How should I price my product?"

AI response: A 500-word essay about pricing strategies covering cost-plus, value-based, competitive, and freemium models. Technically correct. Practically useless. You knew all of this.

With Layer 3 only:

"How should I price my SaaS product?"

Slightly better. AI narrows to SaaS pricing. Still generic.

With all three layers:

"I'm a solo founder (Layer 1) launching a niche developer tool next month with 200 beta users and no funding (Layer 2). How should I price the launch? (Layer 3)"

Now AI gives you specific, actionable advice: start with a lifetime deal to convert beta users, price based on what solo developers actually pay for similar tools, consider a free tier to build word-of-mouth since you can't afford paid acquisition.

Same question. Three different answers. The only difference is context.

How to Build Each Layer

Layer 1 (set once, use forever)

The fastest way: generate a soul archive from your own writing. It captures your identity — values, communication style, expertise, thinking patterns — in a file you can paste into any AI.

The manual way: write 3-5 sentences about who you are, what you do, and how you like information delivered. Even this much context transforms AI responses.

Layer 2 (update when your situation changes)

At the start of a major project or decision, give AI a brief: what you're working on, what your constraints are, what success looks like, what you've already tried. Think of it as a project briefing for a new team member.

Layer 3 (every conversation)

Be specific about what you need. Not "help me with this email" but "make this email shorter and more direct — the recipient is busy and hates fluff." Constraints produce better results than open-ended requests.

The Compound Effect

Context compounds. The more AI knows about you, the less you need to explain each time. The less you explain, the faster you get to useful output. The faster you get to useful output, the more you actually use AI for real work.

This is why people who set up their context once get dramatically more value from AI than people who start from zero every conversation. It's not that they're better at prompting. They're better at context.

Common Mistakes

Too much context. A 2000-word briefing is worse than 200 words. AI doesn't need your life story. It needs the relevant parts.
Wrong context. Telling AI your personality type doesn't help it write your email. Telling it your communication style does.
Stale context. Context from last month's project doesn't help with this month's problem. Update Layer 2 when your situation changes.
No constraints. "Write something good" is not a constraint. "Write something in 3 sentences that a busy CEO will actually read" is.

Layer 1, handled. Permanently.

Generate your identity file from your own writing. Paste it once. Every AI conversation starts with context.

Create Your Soul Archive