You may have tried various ”perfect prompts” circulating online. After trying them, the outputs are still mediocre, generic, and read like they were written by a machine. This is not your fault—the problem lies in the direction. Instead of spending time polishing the wording of your prompts, use that same effort to give the AI more background information.
What is "context," and how is it different from a "prompt"?
A prompt is an instruction; context is the environment.
A prompt is the sentence where you tell the AI what to do: ”Help me write an article about SEO,” ”Summarize this report,” ”Rewrite this passage to make it more concise.”
Context is all the information the AI uses to understand your entire situation: who you are, who your audience is, what you want to achieve, what constraints you have, and the specific background of your industry.
The same prompt, ”Help me write an article about SEO,” will yield completely different results if the AI knows you are a marketing manager targeting B2B manufacturing clients, your readers have engineering backgrounds, and your company emphasizes technical depth over marketing gimmicks.
Why Context Determines Output Quality
Anthropic's engineering team puts it bluntly: ”Most AI agent failures are not model failures—they are context failures.”
This means: When you feel the AI's answer isn't useful, the problem is rarely that the AI model itself isn't smart enough, but rather that it doesn't have enough information to understand what you truly want.
Giving it a better model is less effective than giving it better background information. This shift in mindset is key to moving from ”prompt-first” to ”context-first.”
Why is "Prompt Engineering" Becoming Less Effective?
Prompt Templates Are Failing
Ethan Mollick, a professor at Wharton School studying AI, observed a phenomenon: many prompt techniques circulating online ”are more like magic spells than practical, replicable methods.”
There is no such thing as a universal template. A prompt that works well in one scenario fails in another. Articles claiming ”this one template solves all problems” themselves violate the basic principle of how AI works—AI needs to understand your specific situation, not a fixed set of incantations.
Instead of collecting a pile of prompt templates, learn how to quickly and naturally explain your situation to the AI.
AI Models Are Getting Better at Understanding Your Intent
“Being good at writing prompts” was once a skill, but this advantage is fading.
Current AI models are already very strong at understanding user intent and are improving rapidly. You no longer need to use specific formats or vocabulary to ”coax” it into giving a good answer. You just need to explain your situation clearly, and it will understand what you want.
This means: The ”technique” value of prompts is declining, while the ”information” value of context is rising.
Three Practical Ways to Quickly Provide Context to AI
Method 1: Voice Input, 3-4x Speed
Most people type when conversing with AI, but this is the least efficient method.
You speak at about 150 words per minute, while you type at about 40 words per minute. Voice input is 3-4 times faster than typing.
It's not just about speed. When you type, you unconsciously simplify, summarize, and omit details—because typing is tiring. But when you speak, you naturally pour out background, details, and thoughts. This ”extra” information is precisely the context the AI needs most.
My workflow is simple: turn on voice input and just keep talking. Don't worry about it being unstructured or not refined enough; modern speech-to-text accuracy is very high. Once you're done, the AI will help you organize it.
For free versions, I recommend Lightning Talk and Spokenly. The key isn't the tool—it's the habit of getting used to speaking instead of typing.
Method 2: Let the AI Interview You
This is a simple but effective technique: Don't ask the AI for an answer directly; let it ask you questions first.
Example Prompt:
“Before giving me a solution, first ask me questions like an interviewer to ensure you fully understand my project background and needs.”
Or a simpler version:
“Before answering, ask me a few clarifying questions.”
The benefit of this method: The AI will ask about key details you forgot to mention. Who is your audience? What are your constraints? Where will this content be used? These questions will help you complete the context.
还有一个变体：让 AI 问到它”95% 确定能完成任务”为止。这会逼它把所有需要的信息都挖出来。
Method 3: Let the AI Challenge Your Assumptions
The AI has a problem: It agrees with you too easily.
Whatever you say, it responds with ”Okay,” ”Understood,” ”That's a good idea.” But if your premise itself is flawed, this ”cooperation” can lead you astray.
Example Prompt:
“From now on, don't just agree with me or assume I'm right. I want you to be my critical thinking partner, not a yes-man. Help me check assumptions, raise objections, and find logical flaws.”
This method is particularly useful when planning content strategy or choosing writing angles. It helps you identify blind spots before you start writing.
Context Quality > Context Quantity
More Context Isn't Necessarily Better
You might think: Since context is important, should I give the AI as much information as possible?
Not necessarily. When the context is too long, the AI's performance can actually decline. It gets distracted by irrelevant details, and key information gets buried.
Another issue is that the AI pays the most attention to the beginning and end of the context, while the middle part is easily overlooked. So, it's a good habit to put the most important information at the beginning.
Three Characteristics of Good Context
What constitutes high-quality context? There are three criteria:
Specific——直接和当前任务相关。不要给一堆”可能有用”的背景信息，只给这个任务真正需要的。所以只要是教你在project塞入公司/竞争对手所有资料的线下课都是纯纯的门外汉,割韭菜水平.
Concise—Contains only necessary content. Extra information is not a bonus; it's a distraction.
Actionable—The AI can act directly based on this information. Not vague descriptions, but details that guide specific decisions.
For example, instead of saying, ”Our company is a B2B enterprise,” say, ”We sell industrial packaging machines to procurement managers at food factories, who care most about equipment stability and after-sales service.”
Start Changing Today
Next time you use AI to write content, try this:
Don't spend 10 minutes optimizing the wording of your prompt. Turn on voice input and spend 2 minutes talking about your background—who you are, who you're writing for, and what you want to achieve. Then let the AI ask you a few questions to fill in the missing information.
You don't need a perfect prompt template. Just keep talking.