Is It Legal for Lawyers to Use Perplexity?

Is it ethical for lawyers to use Perplexity? Yes, but its citation model poses a risk of misattribution and inaccuracy.

Perplexity links to sources in its answers, which makes outputs feel reliable. However, they may not be. 

This article covers the compliance risks of using Perplexity AI, the four Model Rules that present the highest exposure, and the workflows that help lawyers remain compliant with their professional and ethical obligations.

Key Takeaways

• Use of Perplexity AI is permissible for attorneys, but its citation model presents risks for misattribution and contextual inaccuracy.
• Every Perplexity citation must be independently confirmed through Westlaw, LexisNexis, or an equivalent legal research database before it enters client work or court filings.
• Standard Perplexity tiers are not designed to protect confidential client data. Enterprise tiers offer stronger data governance, but attorneys must independently assess whether those controls satisfy obligations under the Model Rules.
• ABA Formal Opinion 512 governs Perplexity use under Model Rule 1.1 (Competence), Model Rule 1.6 (Confidentiality), Model Rule 3.3 (Candor Toward Tribunal), and Model Rule 5.3 (Nonlawyer Supervision). The lawyer—not the AI—retains the non-delegable duty of accuracy.

[cta-1]

The Core Risk Lawyers Need to Understand Before Using Perplexity

Perplexity’s risk profile is categorically different from that of other AI tools. Unlike ChatGPT (OpenAI) and Claude (Anthropic), Perplexity AI runs on large language models with a retrieval layer that pulls from the live web and attaches links to sources. That makes results feel verified when they are not.

Perplexity might attribute a dissent’s reasoning to the majority or cite a vacated opinion because it appeared in a recent news summary. A source may misstate a holding, cite a secondary article with no precedential authority, or surface a blog post as though it were binding. While Perplexity can still hallucinate, the risk shifts primarily from fabrication to misattribution and contextual inaccuracy.

Attorneys evaluating Perplexity must also account for the platform's unresolved data sourcing litigation. The New York Times filed a copyright suit against Perplexity in December 2025, alleging unauthorized crawling and reproduction of its reporting. Reddit filed a lawsuit in October 2025, accusing Perplexity and three data-scraping intermediaries of bypassing access controls on an industrial scale.

These cases raise direct questions about the provenance of the data that generates the citations attorneys are warned to verify. 

What Jurisdictions Say About Perplexity in Legal Practice

No jurisdiction currently prohibits attorneys from using Perplexity AI. The table below summarizes guidance from the ABA, key state bars, and a landmark federal ruling.

‍

‍

Can lawyers use Perplexity responsibly under these standards? Yes, but only when they maintain competence, protect confidentiality, and treat every result as unverified until independently confirmed.

Where Perplexity's Compliance Risks are Highest

ABA Formal Opinion 512 governs the four compliance risks below, ordered from most to least severe.

Citation Liability: The Risk Unique to Perplexity (Model Rule 3.3)

Perplexity shows citations that link to web pages, but those pages may be blogs, news articles, or secondary summaries that courts reject as authoritative. 

We’ve seen what happens when attorneys rely on AI-generated citations without verifying them. In Mata v. Avianca, 678 F. Supp. 3d 443 (S.D.N.Y. 2023), attorneys submitted a brief containing AI-generated citations to nonexistent cases. Judge P. Kevin Castel sanctioned both attorneys and their law firm a total of $5,000 for acting in subjective bad faith. 

Mata involved fabricated citations from ChatGPT (OpenAI) rather than misattributed ones from Perplexity, but the Model Rule 3.3 exposure is identical, and the court made clear that reliance on the AI tool is not a defense.

Every citation must be cross-referenced in Westlaw, LexisNexis, or an official database before any court filing.

Confidentiality Exposure in Standard Tiers (Model Rule 1.6) 

On February 10, 2026, Judge Jed S. Rakoff ruled in United States v. Heppner (S.D.N.Y.) that 31 documents a defendant generated through consumer Claude (Anthropic) were not protected by attorney-client privilege or the work-product doctrine. The court based its decision on three grounds:

• Claude is not an attorney. A privileged relationship cannot exist with an AI platform.
• Anthropic's privacy policy permits the collection and disclosure of data to third parties, including regulators. That destroyed any reasonable expectation of confidentiality.
• The defendant used Claude on his own initiative without counsel's direction. The work-product doctrine did not apply.

Judge Rakoff noted that enterprise tools with strict confidentiality guarantees "may present a materially different analysis."

Perplexity's standard tiers carry the same risk profile established in Heppner. Before any use involving client data, attorneys must audit the Perplexity API or the specific tier's agreements to determine whether they provide adequate protection.

For teams that need enterprise-grade confidentiality controls, Spellbook enforces Zero Data Retention policies with its AI partners. 

Try Spellbook for free.

Competence and Coverage Gaps (Model Rule 1.1) 

The competence obligation under Model Rule 1.1 goes beyond understanding how Perplexity works. Attorneys must understand that Perplexity sources information from the live web, not verified legal databases such as Westlaw, LexisNexis, or Fastcase. Unpublished opinions, subscription-gated case law, and jurisdiction-specific statutory updates are all outside its indexing reach or paywalled.

While Perplexity can find a case, it cannot perform a citator check. It cannot reliably tell an attorney if a case has been overruled, vacated, or superseded—a core requirement of Rule 1.1 competence.

Evolution of Comment 8 to Model Rule 1.1 positions AI search competence as a mandatory professional skill, and Continuing Legal Education (CLE) programs increasingly reflect this expectation.

Supervisory Liability When Using Perplexity (Model Rules 5.1 and 5.3) 

Partners and supervising attorneys are ethically responsible for ensuring that the conduct of associates and nonlawyer staff—including their use of Perplexity—aligns with the lawyer’s professional obligations.

If a subordinate’s unverified Perplexity output leads to a filing error or a confidentiality breach, the supervising attorney bears the disciplinary risk. Effective AI governance must go beyond a written policy; it requires active training and audit workflows to ensure every AI-generated citation is cross-referenced against a primary legal database.

[cta-2]

The Compliant Perplexity Workflow for Attorneys

Attorneys who use Perplexity can maintain ethical and professional compliance by following a three-step workflow: research, verification, and handoff. 

Use Perplexity for early-stage research and determine the authority level of each source it returns. Then, verify every citation against an official legal database before it enters client work. Once the research is confirmed, move the contract work into a purpose-built legal tool

Step 1: Use Perplexity for Early-Stage Research

Under Model Rule 1.1, an attorney must understand that Perplexity retrieves public web content rather than authoritative legal research databases. 

Use Perplexity for early-stage research and background information. Before acting on any result, determine whether it is a primary authority, a secondary summary, or a blog post with no precedential value.

Step 2: Verify Every Source Independently 

Perplexity attaches citations to its outputs, which helps attorneys locate sources faster. Lawyers must verify all citations and legal assertions generated by Perplexity against official legal databases before use. 

Step 3: Hand Off to a Purpose-Built Legal Tool for Contract Work

After research through Perplexity is complete and verified, contract work requires a different tool.

Spellbook runs in Microsoft Word, enabling attorneys to draft, review, redline, and benchmark without switching platforms. Its Thomson Reuters Practical Law integration grounds suggestions in vetted legal content, and the Library feature powers AI with a firm's own knowledge and precedents, meaning every suggestion reflects how the firm actually writes.

Try Spellbook free today.

Perplexity Uses Under the Model Rules

Does Perplexity generate reliable legal work product? That depends on what the attorney does with the output, not on the platform's capabilities.

Permissible (attorney treats outputs as unverified starting points):

• Initial issue spotting and research scoping.
• Surfacing regulatory developments on standard tiers.

High-Risk Non-Compliant Actions:

• Citing any Perplexity source in a court filing without independent verification. Relying on unverified citations creates malpractice exposure.
• Entering client names, matter details, or privileged facts into standard tiers without enterprise controls and informed client consent.
• Sending Perplexity output to a client without attorney review.

Spellbook vs. Perplexity: Closing the Compliance Gaps

These gaps represent compliance violations for any contract attorney, in-house counsel, or lawyer relying on Perplexity's standard configuration for contract work. Spellbook closes each one inside the contract workflow.

‍

‍

If your team reviews contracts, Spellbook is built to handle the compliance obligations that Perplexity cannot meet. Try Spellbook.

Using Perplexity is Ethical, But Citations are Always Your Responsibility

Can lawyers use Perplexity? Yes, within the four-rule framework. No jurisdiction prohibits it, and existing Model Rules apply to Perplexity use without requiring new ethical standards.

The practical question is whether the verification burden outweighs the research utility for contract-heavy work. For drafting, review, redlining, and benchmarking, Perplexity creates friction at every stage. 

Perplexity orients. Attorneys move into Spellbook to execute. It runs in Microsoft Word, handling contract drafting, review, and redlining grounded in firm precedent.

Start a free trial with Spellbook.

Frequently Asked Questions

Does Using Perplexity Create Disclosure Obligations for Lawyers?

Generally, no. There is no blanket "AI disclosure" rule in the Model Rules. However, disclosure obligations are triggered in three specific scenarios:

1. Confidentiality (Rule 1.6): If you input protected client data into Perplexity’s standard (non-enterprise) tiers, you must obtain informed client consent first, as these tiers may not meet the reasonable precautions standard to prevent privilege waiver.
2. Significant Use (Rule 1.4): If Perplexity is used for a significant portion of the research or drafting that deviates from the client’s expectations, you have a duty to consult the client.
3. Court Rules: Check for Local Standing Orders. As of 2026, an increasing number of federal and state judges require a specific "AI Certification" or disclosure in every filing.

Can a Perplexity Citation Link be Submitted in a Court Filing? 

No. A Perplexity citation link is not a recognized legal authority. Perplexity indexes the live web, meaning it may surface a summary from a blog post, a retracted news article, or a vacated opinion as if it were binding law.

Under Model Rule 3.3 (Candor) and Rule 11 (Sanctions), attorneys have a non-delegable duty to ensure the accuracy of every citation. Most Standing Orders (e.g., District of Kansas Standing Order 26-01) now explicitly require attorneys to certify that every citation has been checked by a human against an official legal reporter or a sanctioned database such as Westlaw or LexisNexis before filing.

Is Perplexity Safe to Use with Client Information?

Not on standard or Pro tiers. While Perplexity Pro offers an improved user experience, its standard terms do not provide the "reasonable precautions" required by Model Rule 1.6 to prevent the waiver of attorney-client privilege.

Following the February 2026 ruling in U.S. v. Heppner, federal courts established that using a consumer-grade AI tool with client data constitutes a disclosure to a third party, potentially waiving privilege because the platform’s terms allow for data collection and model training.

How Should Law Firms Build a Perplexity Use Policy?

A firm-level policy should address mandatory training, acceptable use, citation verification responsibilities, and data tiering (Consumer vs. Enterprise). Under Model Rule 5.1 and 5.3, supervising attorneys are responsible for AI-related errors or data breaches by their staff.

To meet the "reasonable efforts" standard established in ABA Formal Opinion 512, a 2026-compliant policy must include:

• A "Traffic Light" System: Clear labels on which tasks are Green (administrative), Yellow (research requiring human-in-the-loop verification), and Red (inputting PII into consumer-grade tiers).
• Mandatory Verification Logs: A requirement that every AI-generated citation is cross-referenced in a legal reporter (e.g., Westlaw) before being filed.
• Data Segregation: Strict rules prohibiting the use of "Standard/Free" tiers for any work involving confidential client data.

Purpose-built legal AI tools like Spellbook reduce this supervisory burden by incorporating these efforts in the software, ensuring compliance is automated rather than manual.

Should Transactional Lawyers Use Perplexity or Spellbook?

For contract work, use Spellbook. The tools are not interchangeable. Perplexity is useful for drafting outlines and brainstorming clause language. A lawyer who uses both correctly uses Perplexity to orient to a legal question, then moves into Spellbook to draft, review, and benchmark a contract.

[cta-3]

‍

Ask AI About this Topic

ChatGPT | Claude | Perplexity | Grok | Google AI Mode