ChatGPT, Google Gemini, and similar artificial intelligence tools can answer questions, write code, compose creative text, discuss ideas, and help automate tasks. But the quality of the result often depends on how clearly the user frames the request.
That is why prompt engineering has become a practical skill. A few extra words can change a broad, generic answer into something shorter, more useful, more critical, or better matched to the person asking.
Make the chatbot challenge your thinking
One common weakness of AI chatbots is that they can be too agreeable. If a user presents a vacation plan, a side hustle idea, or another early concept, the answer may sound supportive without pushing hard enough on the risks.
A simple way around that is to assign the chatbot a sharper role. The source suggests asking ChatGPT to act like a curious, inquisitive 10-year-old who wants to help and has many questions. That changes the task from approval to interrogation.
This kind of prompt can expose missing assumptions. It may ask about details the user skipped, point out practical problems, or surface weak parts of an idea before they become harder to fix.
Another route is to ask ChatGPT to respond from an established perspective. Prompts such as “What would Steve Jobs do here?” or “What would Churchill's perspective be on this?” can help a user get unstuck by forcing a different frame. The important caveat is that the chatbot does not truly know what those people would think; it is making educated guesses based on what it knows about widely discussed figures.
Use context to control the shape of the answer
ChatGPT can return too much information when the question is broad. For learning, one way to make the answer more efficient is to ask for “the 80-20 on …” a subject.
The source connects this to the Pareto principle, which it describes as the idea that 20 percent of the work results in 80 percent of the outcome. In a ChatGPT prompt, the aim is to get the most important part of a topic without being buried in every detail.
That format can be used for historical periods, bands, scientific concepts, movies, and other subjects. It works best when the goal is orientation: enough understanding to move forward, not a full survey.
Users can also tell ChatGPT how they want information delivered. Adding “I'm a very lazy person” to the end of a prompt can push the chatbot toward shorter, simpler answers. The same idea can be adapted by saying the user is very busy or very fastidious, depending on the desired response style.
For even more tailored output, ChatGPT can be given personal context through Personalization. On mobile, that setting is reached by tapping the menu button, then the avatar, then Personalization. On desktop, it is available by clicking the avatar and then Personalization.
There, a user can provide a nickname, job, movie preferences, age, DIY experience, or other relevant details. ChatGPT can then take that information into account in new conversations.
Bring images and media into the prompt
Prompting is not limited to typed text. In the ChatGPT phone app, users can include photos from the device camera as part of a prompt. The source gives examples such as asking “How tall is this landmark,” “What type of insect is this,” or “What does this sign say in English?”
To do that, the user taps the + button in the prompt box and then selects Camera. The photo becomes part of the question, letting the chatbot respond to what the user is looking at.
Images can also be used as starting material for creative work. ChatGPT can remix existing pictures, not just generate new ones from scratch. On desktop, users can tap or click the + button and choose Add photos & files; on mobile, they can choose Files.
That opens up prompts built around sketches, family portraits, pet photos, and backgrounds. A rough doodle can be turned into something more finished, a portrait can be reworked in another visual direction, or separate images can be combined. Follow-up prompts can then adjust the result, such as asking to make a scene sunnier.
Ask ChatGPT to slow down before answering
Sometimes the problem is not the chatbot's answer, but the prompt itself. A user may not know whether they have included enough information or whether the request is too vague.
In those cases, the source recommends adding a line such as “Ask me any questions you need clarity on before providing an answer.” That instruction tells ChatGPT not to rush straight into a response.
This can make the exchange take longer at first, but it may reduce wasted follow-up. Instead of guessing what the user means, the chatbot can request the missing context before producing the final answer.
There are also times when previous context is the problem. ChatGPT can remember earlier conversations and personalize responses, which can be useful, but not every task should be influenced by old chats.
A user can ask ChatGPT to forget what came before, or use temporary chat. The source describes temporary chat as an incognito mode where the conversation is forgotten after it ends. On desktop or mobile, it can be started by selecting the dotted speech bubble in the top right corner when opening a new chat.
Turn specific goals into specific outputs
Better prompts work because they give the system a job with boundaries. Asking for a playlist is one example. Instead of requesting music in general, the source suggests a prompt like “Give me a 2-hour chill-out playlist for a rainy Sunday night featuring classics from the 1970s.”
That prompt has duration, activity, setting, and era. Each part can be changed, which makes the output easier to steer.
ChatGPT also includes plug-ins for platforms like Apple Music and Spotify, according to the source. Users can connect accounts and move playlists across by going to the ChatGPT Apps page on the web and adding those plug-ins.
The broader lesson is simple: ChatGPT prompts improve when they include role, context, format, constraints, and permission to ask questions. Whether the task is learning a topic, critiquing an idea, editing an image, interpreting a photo, or creating a playlist, the prompt should make the desired result hard to misunderstand.