Perplexity 101: Getting Started (AI Search That Shows Its Sources)

When I need to actually research something (not brainstorm, not draft: research), I don't use Claude or ChatGPT. I use a dedicated research tool, and Perplexity is the one I reach for most. The reason fits in one sentence: it shows me where every claim came from, so I never have to wonder what was guessed.

Here's the full starter: what it is, the setup, the exact kinds of questions I run through it, and the habit that makes the whole thing trustworthy.

What Perplexity is

Perplexity is AI search. You ask a question in normal language, it searches the live web, reads what it finds, and gives you a synthesized answer with citations: numbered links showing exactly which source backs each claim.

That's a structurally different thing from ChatGPT or Claude answering from memory. A chat model can guess, and it guesses confidently. Perplexity's answers are anchored to documents it just read, and you can click every anchor. For anything where being wrong costs money, sources aren't a nice-to-have. They're the product.

Setup (10 minutes)

1. Go to perplexity.ai and create a free account.
2. Ask it something you'd normally Google, phrased like a real question, not keywords: "what are typical retainer rates for social media management agencies in 2026?" Notice the shape of what comes back: an answer, not ten blue links.
3. Click two citations. Do this on your very first question. The habit of spot-checking sources is what makes this tool safe to rely on, and it takes seconds.
4. Run one deep research query (the more thorough mode; it takes minutes instead of seconds, reads far more sources, and returns something closer to a briefing than an answer). Give it something meaty: "the current state of the creator economy for business educators: market size, major platforms, and what changed in the last 18 months."
5. Set it as a search habit, not an app you remember monthly. The switch that matters: question-shaped searches go to Perplexity, navigation-shaped searches ("that one article on...") stay on Google.

The questions I actually run through it

Steal these shapes:

Pricing and benchmarks:

What are typical rates for [service] in 2026? Break it down by
experience level and market. Cite sources for every number, and tell
me which numbers you're least confident in.

Market landscape (the pre-Manus scout):

I'm considering [decision]. Research the current landscape: who the
main players are, how they price, and what's changed in the last
year. Cite your sources, and flag anywhere your sources disagree
with each other.

That last clause is my favorite trick anywhere in my stack. Where sources disagree is usually where the real answer lives, and where a single confident summary would have misled you.

Fact-check before publishing:

I'm about to publish this claim: "[claim]". Is it accurate as written?
What's the strongest source for it, and what would a critic say?

I run this on statistics before they go in a newsletter. Two minutes, and it has saved me from confidently repeating numbers that turned out to be five years stale.

Vendor/tool comparison:

Compare [tool A] and [tool B] for [my use case] as of right now:
core differences, pricing approach, and known complaints. Cite
sources, and note anything that changed recently.

The "as of right now" matters. Tool landscapes shift monthly, and live search is exactly what chat models are worst at.

Understand a space fast:

Give me a working understanding of [topic] in 10 minutes of reading:
key concepts, main players, current debates, and the 3 best sources
to go deeper.

When I use Perplexity vs. everything else

• Perplexity: anything where the answer lives on the live web and being wrong costs money. Benchmarks, landscapes, fact-checks, comparisons, "current state of X."
• Manus: when it stops being a question and becomes a project: multi-step research with a deliverable at the end. If briefing it properly takes a paragraph, it's Manus. (The Manus 101 covers the handoff.)
• Claude: everything involving MY data and MY voice: writing, strategy, operations, meeting notes. Perplexity knows the internet. Claude knows me.

The one-liner: Perplexity for facts, Manus for projects, Claude for you.

Mistakes to avoid

• Never clicking the citations. The citations are the feature. If you don't spot-check, you've just got a confident summarizer with better branding.
• Using it for writing. It's a researcher, not a ghostwriter. Pull the findings out and do the writing where your voice and rules live.
• Keyword-speak. "social media pricing agency 2026" gets worse results than a full, specific question with your context in it.
• One-shotting big questions. Follow up. Narrow. Ask what it would ask next. Research is a conversation, and Perplexity keeps the thread.
• Trusting synthesis over sources on high-stakes calls. The summary is a map. Before you spend real money, read the two or three underlying sources themselves.

FAQ

What's the difference between Perplexity and ChatGPT? ChatGPT is a general assistant that mostly answers from training. Perplexity searches the live web for each question and cites where every claim came from. For research, that difference is everything.

Is Perplexity free? There's a solid free tier. Paid buys more deep research runs and stronger models; upgrade when research becomes a weekly habit instead of an occasional one.

Can I trust Perplexity's answers? Trust them like a smart intern's memo: genuinely useful, verify anything expensive. The design makes verifying take seconds, which is exactly why it's my research default.

Perplexity or Manus? Perplexity for questions and briefings, Manus for multi-step research projects with a deliverable. Start with Perplexity; you'll feel it when a question outgrows the question shape.

Does Perplexity replace Google? For question-shaped searches, mostly yes. For navigation (finding a specific site or doc you already know exists), Google's still fine.