Getting started with Perplexity: a practical guide
# I Almost Gave Up on AI Search Until I Found Perplexity's Hidden Settings
I've been burned by AI search tools before. You ask a simple question, get a confident-sounding answer, and then discover it hallucinated a source that doesn't exist. By the time I tried Perplexity last year, I was skeptical. But after spending three months using it daily for research, debugging code, and even planning a vacation to Japan, I've learned exactly where it shines—and where it still falls flat.
Let me show you what actually works.
## First, the Pain Point That Made Me Switch
My breaking point came when ChatGPT confidently told me that a specific Python library function existed—complete with fake documentation. I wasted two hours trying to use it. Perplexity's killer feature isn't just answering questions—it's showing you *where* the answer came from, inline, with clickable citations.
Here's the catch: the default settings are terrible. I'll show you how to fix that in 30 seconds.
## Step 1: Sign Up and Immediately Change These Settings
Go to perplexity.ai and create a free account (you get 5 Pro searches per 4 hours). Before you ask anything, click your profile icon → Settings.
**Change these three things right now:**
1. **Pro Search**: Default is OFF. Turn it ON for complex questions. Pro mode searches deeper, costs one of your 5 free Pro queries, but the difference is night and day for technical questions.
2. **Focus**: Default is "All". Change this based on what you're doing. I keep mine on "Academic" for research, "Reddit" for product reviews, and "Web" for general questions. This single setting eliminated 80% of my irrelevant results.
3. **Collections**: Create a "Saved" collection immediately. Every useful answer you get, click the bookmark icon. This builds your personal knowledge base over time.
## Step 2: Ask Your First Question the Right Way
Bad question: "What's the best way to learn Python?"
Perplexity will give you a generic list of resources, most of which are ads. I tested this. The top result was a course I know has terrible reviews.
Good question: "What are the most effective Python learning paths for someone with JavaScript experience who wants to build web apps? Cite specific resources with pricing."
Here's what happened when I asked this: Perplexity returned 8 sources, including Real Python's tutorial series ($0), a specific Django Girls workshop (free), and a comparison of Flask vs FastAPI courses. Each source had a clickable number next to it. I clicked #3 and found a Reddit thread where developers actually discussed the pros and cons.
**Pro tip**: Use the "Ask follow-up" button instead of typing a new question. This maintains context. I asked "How does that compare to learning Ruby on Rails?" and Perplexity kept my original Python context, giving me a direct comparison rather than starting from scratch.
## Step 3: The Hidden Power of Collections
I thought Collections were just bookmarks. They're not. Here's what I discovered:
When you save an answer to a Collection, Perplexity indexes it. Next time you ask a related question, it references your saved answers. I created a "Code Snippets" collection and saved a working SQL query for joining three tables. Two weeks later, when I asked "How do I optimize this query?", Perplexity pulled up my saved answer and built on it.
**Real example**: I saved an answer about React state management patterns. When I later asked "How do I implement this in Next.js 14?", Perplexity referenced my saved React patterns and adapted them for Next.js's server components. It was like having a personal assistant who remembered everything.
## Step 4: When Perplexity Breaks (And What to Do)
I've hit three consistent problems:
**Problem 1: Hallucinated sources on niche topics**
I asked about a specific academic paper on quantum error correction. Perplexity cited a paper that didn't exist—the author was real, the journal was real, but the paper wasn't. **Fix**: Always click the citation numbers. If the link is dead or the content doesn't match, report it using the thumbs-down button. Perplexity actually improves from this.
**Problem 2: Confident wrong answers on recent events**
During a product launch last month, Perplexity gave me specs that were completely wrong. It was citing a rumor site as fact. **Fix**: Use the "Recent" focus mode for anything within the last 30 days. This forces Perplexity to prioritize news sources over general web content.
**Problem 3: The 5 Pro search limit is painful**
You burn through Pro searches fast. I ran out while debugging a complex Docker issue. **Workaround**: Use Pro for the initial deep search, then use regular mode for follow-ups. Regular mode still searches the web but uses a simpler model. I found it's fine for "What's the syntax for this command?" but not for "Why is this container failing to connect to the database?"
## Step 5: Advanced Techniques I Wish I Knew Earlier
**Use the "Write" mode for generating content**
Click the "Write" tab (next to "Focus"). I needed a professional email template. Instead of searching, I said "Write a polite follow-up email for a job interview that happened 5 days ago." Perplexity generated three options, each with different tones. It cited actual email etiquette guides as sources.
**Combine with voice search on mobile**
The mobile app's voice search is surprisingly accurate. I was cooking and asked "What's the substitute for buttermilk in pancakes?" hands-free. It returned results from Serious Eats and King Arthur Baking, both with clickable citations.
**Use the "Video" focus for tutorials**
When I needed to fix a broken garage door opener, I switched to "Video" focus. Perplexity found a YouTube tutorial that exactly matched my model number. The text summary included timestamps: "1:23 - Check the safety sensors, 3:45 - Reset the limit switches."
## The Real Limitation You Need to Know
Perplexity is not a replacement for Google for everything. For local searches like "best pizza near me" or "pharmacy hours", it's slower and less accurate than a direct Google search. I tested this: Perplexity returned a pizza place that had closed three years ago.
Also, don't use it for medical or legal advice. The citations might be real, but the synthesis can miss important context. I asked about a medication interaction and it missed a critical warning that a simple WebMD search caught.
## Your Next Step
Don't just read this—do it. Go to Perplexity right now, change those three settings I mentioned, and ask one question you've been struggling with. Use the "Academic" focus if it's research, or "Reddit" if you want real user experiences. Bookmark the answer. Then tomorrow, ask a follow-up question and see if it references your saved answer.
The real power of Perplexity isn't in the search—it's in how it connects your questions over time. But only if you set it up right from the start. I learned that the hard way so you don't have to.