Perplexity for Lawyers: Sourced Answers, Enterprise Terms and the Order That Named Fake Parties

Partners do not log in on day one. At Gunderson Dettmer, when the firm switched on Perplexity Enterprise, “more than half of the partners logged in on the first day”; 80% of lawyers and partners are now active, the firm runs more than 35,000 queries a month, and the net promoter score is 68. One attorney put the appeal in a sentence: “I now use it for all my internet searches” (Perplexity customer story).

That is the pull of Perplexity for lawyers: a search engine that reads the results for you and shows the sources. The push the other way came from a federal courthouse in Mississippi, where a law clerk used Perplexity “strictly as a foundational drafting assistant” and the judge docketed a temporary restraining order naming parties who were not in the case. Both stories are true. The difference is what each user asked the tool to do.

What Perplexity does differently: sources first

A chatbot answers from what it remembers; Perplexity searches first, then writes an answer with numbered references to the pages it read. When ChatGPT names a regulation, you have to find out whether it exists; when Perplexity names one, the link is there, and the only question is whether the summary above it is faithful.

That makes Perplexity strongest on questions whose answer is on the open web (what a regulator published last month, what a counterparty’s standard terms say) and weakest on questions whose answer lives in a reporter series, because the web it searches is not Westlaw. Tara Waters’s caveat about ChatGPT applies equally: its “sourcing hasn’t yet been optimised for identifying proper legal sources first” (Artificial Lawyer). A blog post ranks as a source; so does a court’s own PDF; the tool does not know which you would rather cite unless you tell it.

Gunderson Dettmer’s rollout: 80% active, 35,000 queries a month

Gunderson is not a technophobe firm being dragged along: it built its own ChatGD assistant on Azure OpenAI in August 2023. Perplexity Enterprise went firm-wide in a May rollout (the customer page says only “May of this year”), and the numbers are the firm’s own as reported by the vendor. Treat them as a vendor case study, but note what they measure: daily use by lawyers and partners, not a pilot. The lesson is about the job, not the brand: a tool that replaces the browser search bar gets used because the habit already exists, while a tool that needs a new habit gets piloted, as the tools map shows across the market.

Computer for Counsel (June 2026): what it promises and what is verified

Perplexity announced a legal product at a “Computer for Counsel” event in New York in late June 2026, built on its agentic “Perplexity Computer” and aimed at mid-sized firms and in-house teams (Artificial Lawyer), after a patent search tool and a LegalZoom partnership.

Perplexity’s launch post describes the product as “Powered by 20+ frontier AI models”, with premium legal sources including Midpage, Deel and LegalZoom, connectors to Box, Carta, DeepJudge, Docusign and (coming soon) Clio, and the ability to run inside Microsoft 365. Those specifics come from the vendor and had not been independently tested as of September 2026. Midpage at least is a known quantity: it scored 79% on Vals AI’s October 2025 legal research benchmark, against a lawyer baseline of 71%, as reported in LawSites’ write-up of the results.

Legal research with Perplexity: where it helps and where it does not

The Divisional Court’s warning in Ayinde is worded for ChatGPT but is about the category: “Freely available generative artificial intelligence tools, trained on a large language model such as ChatGPT are not capable of conducting reliable legal research” (Ayinde v Haringey). Perplexity was among the tools MyPillow’s lawyers disclosed in Coomer v. Lindell before being fined $3,000 each for nearly thirty defective citations. A source link is a head start on verification, not a substitute; the guide to legal research without hallucinations has the full protocol.

Confidentiality: Free, Pro and Max train by default; Enterprise does not

Perplexity’s help centre, last modified 3 September 2026, is refreshingly direct (Perplexity Help Center).

For most lawyers the simpler rule holds: Perplexity is for public information, and public information needs no client data in the prompt.

The Wingate order: a clerk, Perplexity and non-existent parties

On 20 July 2025 Judge Henry Wingate of the Southern District of Mississippi issued a temporary restraining order in Jackson Federation of Teachers v. Fitch that named parties who were not in the case and misquoted Mississippi law. It was withdrawn and reissued; his 1 August 2025 ruling called the problems “clerical errors”. In October 2025, answering Senator Grassley, the judge explained that a law clerk had used Perplexity “strictly as a foundational drafting assistant” (FedScoop). His chambers now require a second clerk’s review, and every cited case is printed from Westlaw and attached to the final draft.

The clerk’s phrase, “foundational drafting assistant”, describes precisely the job Perplexity is bad at: generating the skeleton of a legal document. Wingate’s remedy, a second reader plus a printed copy of every authority, is a protocol any firm can copy at no cost; the citation verification guide turns it into a checklist.

Perplexity vs ChatGPT vs Gemini for research

The Vals figures come via LawSites’ write-up (LawNext). The ChatGPT guide and the Gemini guide go deeper. The short version: Perplexity for finding, ChatGPT or Claude for making, a legal platform for anything that ends in a citation.

A safe research prompt pattern

The pattern that keeps Perplexity honest has three parts: name the sources you will accept, demand the URL and date next to every claim, and ask for what it did not find. Longer versions with verification steps are in the prompt library.

List the rules of court, practice directions and statutory provisions that govern [service of a claim form on a company outside the jurisdiction] in [England and Wales]. Use these primary sources only: [legislation.gov.uk], [justice.gov.uk/courts/procedure-rules]. For each provision: exact citation, the operative text quoted verbatim, any time limit, the consequence of non-compliance, and the URL with the page's stated date. Present as a checklist in the order a practitioner would apply it. Flag anything amended since [January 2026] with CHECK CURRENCY. If the named sources do not cover a point, write NOT FOUND IN NAMED SOURCES rather than using another site.

Here is my existing analysis dated [date] on [topic]: [paste, anonymised]. Using only [named regulator site], [official legislation site] and client alerts from [three named law firms], identify anything published since [date] that affects it: legislation, amendments, appellate decisions, regulator guidance or withdrawn authority. For each change: what changed, the date, the URL, and which paragraph of my memo it affects. Where nothing affects a paragraph, write NO CHANGE FOUND. Do not rewrite the memo. End with the sources you searched and found nothing in.

Produce a briefing for a lawyer new to [the EU AI Act's obligations for deployers] as at [today's date]: governing instruments with official links; who is regulated and who enforces; the compliance timeline with exact dates and any deferrals; the three most-cited practitioner summaries; the open questions. Then a source ledger: for every source, its URL, publisher type (official, law firm, news, other), date, and the single claim it supports. Prefer official and law-firm sources over news. Maximum 1,200 words plus the ledger. Cite no case unless the linked source contains it.

Run the last prompt on a topic you already know before one you do not; a tool that does not flag the EU AI Act’s Digital Omnibus deferral of the high-risk deadline to 2 December 2027 has told you something about its currency.

Pro se litigants winning with Perplexity

The same tool cuts the other way in court. Lynn White, a Long Beach tenant, overturned an eviction and avoided roughly $55,000 in penalties using ChatGPT and Perplexity Pro. “It was like having God up there responding to my questions,” she told NBC News; “I never, ever, ever, ever could have won this appeal without AI” (NBC News).

Two consequences for practising lawyers. The opponent across the aisle may arrive with a sourced argument that is better than you expect and worse than it looks: Damien Charlotin’s database counts 1,173 pro se litigants among its 2,039 decisions involving hallucinated material, against 811 lawyers, so cite-checking the other side’s brief is now part of the job. And the client on the phone has often already run the question through Perplexity; the advice that survives is the advice that can point to a better source.

Where to go next: the tools cluster has the companion guides, and the GEO guide shows how Perplexity decides which lawyers to recommend when a prospect asks who to hire. Perplexity is one of the tools used live in AI Lab for Lawyers, and the exercise is the one above: run a real question, open the sources, and see which citations hold.