I just started using Perplexity AI and I’m not really sure how to get the best results from it. I tried asking a few questions, but the answers weren’t always what I expected, and I feel like I’m missing some basic tips or features. I need help հասկanding how to use Perplexity AI better, including how to write better prompts and use it more effectively.
Treat Perplexity like search plus follow-up, not like magic.
A few things help a lot:
-
Ask one task at a time.
Bad: “Tell me about EVs and which one to buy and tax credits too.”
Better: “Compare Model 3, Ioniq 6, and Polestar 2 for highway driving in the US.” -
Add constraints.
Say your budget, country, skill level, deadline, format.
Example: “Explain Roth IRA vs 401(k) for a 28-year-old in the US, under 200 words, plain english.” -
Ask for sources every time.
Use prompts like:
“Answer with 5 sources.”
“Quote the source for each claim.”
“Only use papers from 2022 or newer.” -
Verify facts.
Perplexity is solid for finding info fast. It still misses stuff. Open the sources. Check dates. Check if it cited a blog when you needed a study. This part matters a lot. -
Use follow-ups instead of restarting.
“Make this shorter.”
“Turn this into bullet points.”
“Which claim is weakest?”
“Compare source 1 and source 3.” -
Tell it what you do not want.
“No Reddit.”
“No affiliate sites.”
“No opinion pieces.”
This helps more than ppl think. -
Use it for first drafts, not final truth.
Great for summaries, research plans, shopping comparisons, code ideas. Weak spot is confidence. Sometimes it sounds right and is wrong, lol.
My go-to format:
“Answer in bullet points. Keep it under 150 words. Use recent US sources. If unsure, say so.”
If your answers felt off, your prompt was prob too broad or missing constraints. Narrow it down and it gets way betetr.
Big thing nobody tells you: Perplexity gets way better when you use it in the right mode.
If you just want facts, ask for a straight answer. If you want ideas, say that. If you want a comparison, make it compare. A lot of weird replies happen because the tool is guessing the job.
A few practical tips:
- Start with a rough question, then refine based on what it got wrong
- If the answer feels too polished, ask: “what did you leave out?”
- Ask it to separate facts vs interpretation
- For shopping or product stuff, I actually don’t fully agree with using it as the main source. It can lean on recycled review content. I’d use it to narrow choices, not pick the winner
- Ask for a timeline if the topic changed over time
- If you’re learning something, tell it to quiz you after explaining
Also, use it to find terminology. Half the battle is learning the right words so your next search is better.
@stellacadente is right about constraints, but I’d add this too: sometimes people overpack prompts and make them kinda messy. Short, clear, then follow up usually works betetr.