Legal Services Marketplaces and Unauthorized Practice Risks

The U.S. legal services market is $387.6 billion in 2026 and is projected to reach $591.07 billion by 2035, growing at a 4.80% compound annual rate. That growth ceiling is set less by demand than by law: unauthorized practice of law rules, which vary state by state and are now being rewritten in response to AI, dictate what any platform can actually offer, to whom, and at what scale.

• A national legal services market is large enough to draw serious platform investment, projected to grow from $387.6 billion in 2026 to $591.07 billion by 2035 legal services market is large enough to draw serious platform investment, projected to grow from $387.6 billion in 2026 to $591.07 billion by 2035
• UPL prohibitions function as a structural ceiling on that growth, capping how far any single platform can expand across jurisdictions
• Legal demand is high, but the rules governing who may deliver legal services, and how, predate the internet entirely, let alone generative AI
• An estimated 92% of low-income people currently receive no legal assistance or insufficient legal assistance, a gap that UPL constraints directly shape by limiting who is allowed to step in

What unauthorized practice of law prohibits

UPL is not a vague gesture toward "acting like a lawyer." It has a working definition, even if that definition shifts depending on which state you're standing in.

• UPL generally covers applying legal principles and judgment to a specific person or entity's circumstances, preparing legal documents that affect legal rights, representing someone in a proceeding, and giving advice about specific legal rights and obligations
• No single national definition exists. Each state builds its own version. A platform operating coast to coast is really navigating 50-plus overlapping frameworks at once. South Carolina's 2025-2026 Bill 4819 shows where states are heading as they update these definitions for a software-driven world: It explicitly covers drafting, completing, or preparing legal documents, including pleadings, settlement agreements, parenting plans, and court filings. It covers advising on legal rights, remedies, or obligations. It covers representing others in negotiation, mediation, arbitration, or formal proceedings. It covers advertising or implying the ability to perform any of the above, a provision aimed squarely at platforms that market legal capability without licensed delivery behind it. "Drafting, completing, or preparing" is defined broadly enough to include editing, reviewing, selecting, recommending, or customizing language. This pulls template and form-fill products directly into scope. Each act of providing or advertising unauthorized services is treated as a distinct violation under the bill. Attorneys who know or reasonably should know about UPL conduct must report it in writing to the Office of Disciplinary Counsel within 15 days, and that office then has 10 days to refer the matter to the Attorney General. California treats UPL as a criminal matter. The State Bar's guidance on immigration consultants is instructive: consultants may translate a person's answers onto a form, but they cannot advise which form to use or represent the client, a narrow and precise example of where the line actually sits for document-adjacent services

Where platform business models run into the information-versus-advice line

The controlling distinction, in theory, is simple: general legal information is fine, legal analysis applied to someone's specific situation is not. In practice, that line runs straight through the middle of most platform business models.

• Non-lawyer consultants who drift into giving legal advice
• Non-lawyer employees negotiating contract terms or offering legal guidance to clients
• Multi-state operations navigating jurisdictions where they may lack licensed legal coverage
• Reliance on unlicensed online legal services for matters that are more complex than they first appear

This isn't a problem unique to legal tech. Accountants advising on tax consequences, real estate agents drafting purchase contracts, HR professionals interpreting employment law statutes: all of them live near this same boundary. What makes tech platforms different is scale. A real estate agent crosses the line once, with one client. A platform crosses it, if it crosses it, thousands of times simultaneously. The document-prep category deserves particular attention here, since customizing a legal document to fit someone's specific circumstances often tips a "general form" into UPL territory, exactly the behavior South Carolina's Bill 4819 captures under its definition of "drafting."

How AI sharpens every existing UPL tension

AI does not introduce a new problem so much as it takes every existing UPL tension and turns up the pressure.

The multi-jurisdictional conflict gets worse, not better, with AI. A tool permitted inside Utah's regulatory sandbox might constitute UPL the moment a user opens it in California or Texas, and a nationally deployed AI product runs into every one of those conflicts at once, not one state at a time

• AI systems struggle with the information-versus-advice distinction for a structural reason: their core function, applying trained knowledge to whatever a user types into the prompt box, maps almost exactly onto the legal definition of advice
• The DoNotPay case remains the clearest cautionary precedent in this space:
• The company once marketed itself as the "world's first Robot Lawyer"
• It faced UPL lawsuits and separately settled a deceptive-practices action brought by the FTC
• The FTC voted 5-0 on January 16, 2025, to approve a final order barring deceptive AI-lawyer claims and imposing monetary relief
• The lesson travels beyond DoNotPay itself: marketing language implying lawyer-equivalent capability draws regulatory scrutiny independent of whatever the underlying product actually does
• Large AI platforms have started building guardrails accordingly:
• As of October 29, 2025, OpenAI prohibits using its services for "provision of tailored advice that requires a license, such as legal advice, without appropriate involvement by a licensed professional"
• As of September 15, 2025, Anthropic classifies legal questions as a "High-Risk Use Case," requiring professional review of outputs and disclosure that AI was involved
• Neither policy eliminates the underlying risk. Both simply shift liability downstream, onto whoever deploys the tool in front of an actual user

The regulatory reform movement and what it has produced so far

Reform talk in this space has been loud for years. What's changed recently is that some of it is turning into actual rule text.

A judicial policy research organization published a policy paper arguing that UPL regulation needs modernizing so AI-fueled legal service tools can responsibly meet unmet legal need. It lays out three paths forward: revise UPL statutes to explicitly permit vetted AI tools under disclosure, data security, and transparency requirements; build regulatory sandboxes for controlled testing; or narrow UPL definitions to focus on who may hold themselves out as a lawyer, rather than who may provide legal assistance. That third option, if adopted broadly, would open real room for marketplace operators to build products outside the courtroom context entirely

• States are already moving, unevenly, in this direction:
• Arizona modernized its rules in 2020 and 2021, permitting alternative business structures
• Utah began piloting 21 Community Justice Advocates in July 2025 to provide limited legal help in debt cases
• Indiana has been examining attorney shortage issues through a dedicated commission
• Colorado's Access to Justice Commission asked the state Supreme Court to revise UPL rules to account for new technology, and the court has since formed a subcommittee to weigh AI-related amendments
• Courts, meanwhile, are not settled on the constitutional boundaries here. The Second Circuit in 2025 ruled against Upsolve, a legal-advice nonprofit that pairs software with trained non-lawyer volunteers, vacating an earlier free-speech ruling that had favored the group. That outcome signals genuine, unresolved judicial uncertainty about where enforcement of UPL rules can constitutionally reach
• The Thomson Reuters Institute has pushed back on the reform push from another angle, arguing that existing consumer protection and product liability law already provides meaningful safeguards, and that demanding tech-based legal services prove they cause zero harm before they're allowed to operate risks strangling supply during what amounts to a genuine access-to-justice shortage

What the LegalZoom–Perplexity partnership signals for AI-native legal distribution

On June 4, 2025, LegalZoom (Nasdaq: LZ) announced a strategic partnership with Perplexity to bring legal solutions to Perplexity Pro subscribers, a notable partnership between a legal services provider and a major generative AI platform.

• The deal represents a legal marketplace choosing to distribute through an AI surface instead of continuing to fight for visibility inside traditional search results, a direct response to how people are starting to look for legal help in the first place
• LegalZoom brings a long history with UPL scrutiny to this move, including a consent judgment in North Carolina and years of state-by-state negotiation over what it can and cannot offer. That history is what makes the partnership a credible signal rather than a stunt: this is a company with real compliance depth choosing AI-native distribution. When Google shows an AI summary above the results, according to Pew Research Center data, users click through to a traditional organic result on only about 8% of visits when an AI summary appears, compared with 15% when no AI summary appears. Traffic flowing from generative AI sources to consumer sites has grown dramatically over the past year

The AI visibility gap facing legal brands and the cost of waiting

Adoption has already happened on one side of this market and not the other. That gap is where the real cost is accumulating.

79% of legal professionals now use AI tools in their own work, showing the professional side of the industry has already adopted the technology. Consumer-facing discovery, how ordinary people find legal help through AI systems, is the piece still lagging behind

• Individual firms are largely invisible at the exact moment a prospective client is asking an AI system for help, while established directory platforms have moved quickly to establish presence in AI-generated responses. Generative Engine Optimization, or GEO, is the discipline built around this gap: optimizing content to increase visibility, citations, and mentions inside AI-powered generative search engines. It targets something fundamentally different from traditional SEO, which chases ranked links. GEO chases inclusion inside a synthesized, generated answer

UPL compliance in content and marketing for legal clients

The UPL boundary doesn't stop at product delivery. It runs straight through marketing copy too, and South Carolina's Bill 4819 makes that explicit by covering advertising or implying the ability to perform legal services, not just performing them.

• Agencies managing legal marketplace clients need to draw the same line in content that regulators draw in product: information about the law is permissible, advice applied to someone's specific situation is UPL exposure. The distinction governing the product is the same distinction that governs how it gets marketed
• Practical operating principles follow from that:
• Bring in licensed attorneys to review any content that edges toward the information-advice line, rather than relying on marketing judgment alone
• Build jurisdiction-specific review into every multi-state campaign, since copy that's fine in one state can be a violation in the next
• Treat AI-generated content for legal clients with the same scrutiny OpenAI and Anthropic apply to their own outputs, both companies require professional review of legal content, a standard that reflects the documented risk AI outputs pose in high-stakes contexts
• Train account teams to explain the UPL framework to clients in plain terms. Legal marketplace clients who already know the rules better than most vendors do quickly lose trust in agencies that can't articulate why a certain content approach is off-limits. AI visibility monitoring deserves treatment as its own ongoing function. Tracking where a legal brand shows up in AI-generated responses, and what those responses actually say about it, requires continuous measurement. The stakes are not abstract: the 5W PR Legal AI Visibility Index 2026 recorded 487 instances of AI errors or hallucinations in court filings in 2025 alone, a concrete demonstration that AI surfaces get facts about legal matters wrong at real scale, and that nobody serving legal clients can afford to treat AI output as self-correcting.