Sebaraya Tech – If an AI chatbot gives you a bland or off-target answer, the fix is usually in the prompt. Write to it the way you’d brief a smart new colleague who knows nothing about your situation: say what you want, why, for whom, and in what shape.
Both Google and Anthropic publish prompting advice for their chatbots, and the core ideas overlap. Here are the habits that make the biggest difference, with before-and-after examples.
Ask Like You Would a Person
Google suggests using natural language, as you would with a coworker, rather than typing keywords.
| Weak prompt | Better prompt |
|---|---|
| Training plan | Write a training plan for the sales team for the launch of a new product |
| Marketing talking points | Give me 12 thoughtful questions to ask a Chief Marketing Officer about their strategy |
Those rewrites come straight from Google‘s own examples. The better versions name the audience, the goal, and how much output is wanted.
Pass the Colleague Test
Anthropic recommends a simple check: show your prompt to someone with little background on the task. If they’d be confused, the AI will be too.
The chatbot can’t see your inbox, your boss’s preferences, or last week’s meeting. Anything it needs to know has to be in the prompt.
Give Context and the Reason Why
Context helps the model understand what a useful answer looks like. Google‘s example adds the job title, required skills, and a company summary when asking for a job description.
Explaining why a rule matters works even better. Anthropic contrasts “never use ellipses” with telling the model the text will be read aloud by speech software that can’t pronounce them; the reason helps it apply the rule sensibly.
Be Exact About What You Want Back
Say whether you want edits or ideas. Anthropic notes that “Can you suggest some changes” often gets suggestions only, while “Change this…” gets the changes made.
Add specifics such as subject, timeframe, length, and format. Google recommends specific keywords, like stating both the project and its deadline in a planning request.
Show an Example of the Result
If you care about format or tone, show it. Anthropic calls examples one of the most reliable ways to steer output, and suggests three to five that are relevant and varied.
For instance, paste two product descriptions you like before asking for a third. Examples like these usually pin down length and style more clearly than describing them as “punchy but professional.”
Split Big Jobs Into Steps
Long, multi-part requests are easier to get right in pieces. Google advises breaking complex tasks into separate prompts.
A useful pattern from Anthropic: ask for a draft, then ask the model to review it against your criteria, then ask it to revise. You can also end a prompt by asking it to check its answer against a requirement before finishing.
What Good Prompts Can’t Fix
Even a perfect prompt doesn’t guarantee correct facts. Google notes that Gemini can hallucinate and present inaccurate information as factual, so check names, numbers, and sources yourself.
Keep private data out of prompts too: passwords, ID numbers, and confidential files don’t belong in a chatbot unless you know how the service handles them.
Take the last prompt you typed and add one line of context and one line describing the output you want. Then compare the two answers.
Komentar Ditutup! Anda tidak dapat mengirimkan komentar pada artikel ini.