Context is the prompt

Most people think a good prompt is a detailed prompt. It isn't. A detailed prompt and a contextual prompt are not the same thing, and confusing them is the most common reason people feel like AI "doesn't understand what I'm going for."

Here's the difference.

A detailed prompt tells the AI what to do:

Make me a landing page with a headline, a subheadline, a call-to-action button, and a footer with three links.

A contextual prompt tells the AI who it's for, what success looks like, and what constraints matter:

I'm building a landing page for a one-person bookkeeping service targeting small restaurant owners. The tone should feel personal and trustworthy, not corporate. The goal is to get people to book a 20-minute intro call. The owner's name is Maria.

Same task. Completely different output.

Why context changes everything

When you give an AI enough context, you're not instructing it — you're briefing it. You're handing over the things that would otherwise require fifteen rounds of back-and-forth to establish. The AI can then make judgment calls on your behalf: what font feels "trustworthy," what microcopy sounds like Maria, whether to lead with the problem or the offer.

Without that context, it defaults to the middle of the distribution. You get a decent result that doesn't feel like yours.

Three questions that unlock context

Before I write any prompt, I ask myself three questions. I don't always answer them explicitly in the prompt — sometimes the answers shape how I frame the request — but I always have them in mind.

Who is this for? Not just the demographic. Their anxiety. What they already believe. What would make them trust this thing.

What does done look like? Not the features. The feeling. "I want someone to look at this and immediately understand what it does" is more useful than "I want a hero section."

What would I hate to see? Constraints are free information. "Don't make it look like every other SaaS startup" is a powerful instruction that costs you nothing to give.

The compounding effect

Here's the part nobody tells you: context compounds. When you give Claude Code a rich brief at the start of a session, every subsequent request inherits that context. You don't have to re-explain the project, the tone, or the constraints. You just say "now add a testimonials section" and it already knows what "on-brand" means for this project.

This is why people who get the best results from AI tools tend to front-load effort into the setup prompt. Not because it magically solves everything — it doesn't — but because it shifts the cognitive load to the beginning, where it costs less.

A prompt worth saving

If you do nothing else with this post, save a version of this prompt somewhere and reach for it when you start a new project:

I'm building [thing] for [who]. Their main concern is [anxiety or goal]. 
The tone should feel [adjective]. The measure of success is [concrete outcome]. 
Avoid [things you don't want]. My constraints are [budget / time / tech].

Fill it in before you write a single feature request. Then paste it at the top of every new session.

You'll be surprised how much work that one paragraph does.

---

Questions, pushback, or examples of context prompts that worked for you — send them to hello@plainbuilt.co. I read everything.