Perplexity is an answer engine, not a chatbot. You ask a question, it searches the live web, and it returns an answer with a numbered citation attached to every claim. Click the number, see the source. That one feature changes what the tool is for: instead of “the AI said so,” you get “here are the six sources, check them yourself.” Across my 10 firms, that makes Perplexity the default for factual research questions, and completely wrong for anything headed into a filing without verification.
Why citations matter more to lawyers than anyone else
Most professionals can live with a plausible answer. We can’t. Our entire job runs on “says who?”
When ChatGPT tells you something, the claim comes from training data that might be two years stale, and checking it means a separate research session. When Perplexity tells you something, the source sits one click away. You read the actual statute, the actual news article, the actual study, and you decide in thirty seconds whether the answer holds up.
That workflow difference matters more than model quality. Claude is still my pick for drafting and for anything involving a long document. But when the question is “what is true and who says so,” Perplexity gets there faster because verification is built into the output instead of bolted on afterward.
What I actually use it for
Preliminary orientation in an unfamiliar area of law. A referral comes in outside your practice area and you need to know, before lunch, whether the case is worth a conversation. Twenty minutes with Perplexity gets you the governing statute, the deadline that matters, and the two or three issues that decide these cases. That is orientation, not research, and orientation is exactly what it’s good at.
Opposing party and company background. Before my lemon law firm engages a manufacturer, someone runs the background: open recalls on that model, NHTSA complaint volume, recent settlements, news coverage of the defect. Perplexity compresses that from an hour of tabs into four minutes with sources.
Medical and technical literature. Injury firms live on this. “What does the published literature say about long-term outcomes for this injury type” returns actual studies you can pull and hand to your expert, not a chatbot’s memory of studies that may not exist.
Jurisdiction quick checks. Filing fees, e-filing requirements, local notice rules, service quirks in a county you’ve never touched. Confirm on the court’s own site before you rely on it, but Perplexity finds that site and the answer in one step.
Cases touching the news. When a matter intersects with a live news story, a company collapse, a recall, a criminal investigation, chatbots are blind past their training cutoff. Perplexity reads today’s coverage.
What Perplexity is not
It is not a citator. It will surface real cases, and it will occasionally surface real cases that say something different from what it claims they say. A source link lowers the odds of a fabricated citation. It does nothing about mischaracterization.
So the rule stays absolute: no case citation goes into a filing until you have pulled the actual opinion on Westlaw, Lexis, or the reporter and read it yourself. Lawyers have been sanctioned for skipping that step, starting with the Avianca case in 2023 and repeated dozens of times since by lawyers who apparently don’t read sanctions orders either. For tools actually built for case law, see the AI legal research tools guide.
Spaces: one thread per matter
Perplexity’s Spaces are folders for research threads. Create one per matter, and every search, source, and follow-up stays together instead of scattering across your history. Each Space takes custom instructions, so mine say things like “answer as if briefing a consumer protection attorney, prefer primary sources, flag anything older than 2023.”
When the research pile gets big, export the sources into NotebookLM and question the pile itself. The two tools chain together well: Perplexity finds, NotebookLM digests.
How it compares
| Perplexity | Claude | ChatGPT | ||
|---|---|---|---|---|
| Citations on every claim | Yes, built in | Only when searching | Only when searching | You do the reading |
| Current information | Live web, always | On request | On request | Live |
| Long document handling | Pro file upload, limited | Best available | Good | None |
| Drafting quality | Weak | Strongest for legal | Strong | None |
| Cost | Free or $20/mo | $20/mo | $20/mo | Free |
The pattern: Perplexity for finding facts, Claude for working with documents and drafting, Google when you already know which site you need.
Free vs Pro
The free plan gives you unlimited standard searches and a few advanced searches a day. Pro at $20 a month gets you 300+ advanced searches, a choice of stronger underlying models (including Claude), and file upload so you can ask questions about your own PDFs.
Start free. Upgrade the week you catch yourself saving searches for later. Rationing a $20 tool is a bad trade for anyone billing in six-minute increments.
Getting started
- Create a free account at perplexity.ai. No card required.
- Ask one real question from a live matter, something you would otherwise spend twenty minutes googling.
- Click through the citations. Judge the sources like you’d judge a junior associate’s memo.
- Create a Space for one active matter and add custom instructions for your practice area.
- Run a background sweep on the opposing party in that matter.
- After a week of daily use, decide on Pro based on whether you hit the search limits.
Step 5 is the one that sells lawyers on the tool. Pick your most active matter and run the opposing party sweep today. It takes four minutes, and you will find at least one thing you didn’t know.