AI use in the legal profession has moved from “experiment” to “everyday workflow” for a meaningful—and growing—share of lawyers. The clearest pattern across high-quality sources is this: adoption is rising quickly, but it’s uneven by firm size, practice type, and access to training and approved tools.

What “using AI” usually means in law

In most studies and industry surveys, “AI use” includes things like:

  • We Legal research support (finding authorities, summarizing results)
  • Drafting assistance (emails, clauses, outlines, first drafts)
  • Contract review and issue-spotting
  • Litigation support (timelines, document summaries, deposition prep)
  • Operational work (knowledge management, intake, admin workflows)

Just as important: reputable sources repeatedly stress that AI output still requires human review due to reliability and hallucination risks. (dho.stanford.edu)

The Best Prevalence Numbers (Recent, Highly Cited Sources)

ABA Legal Technology Survey / ABA TechReport

The American Bar Association’s AI-focused TechReport shows a substantial jump in adoption in recent years. Reporting around the 2024 survey results, ABA-related coverage indicates AI adoption rose from about 11% (2023) to about 30% (2024), with larger firms adopting faster (e.g., much higher usage in firms with 100+ lawyers than in solos/small firms). (LawSites)

This “size gap” is one of the most consistent findings: big firms often have more budget, formal policies, and vetted tools—making adoption easier and safer.

Thomson Reuters: Generative AI in Professional Services (legal-specific)

Thomson Reuters’ reporting on its 2025 findings commonly cites about 26% of legal organizations actively using GenAI (up from a lower figure the year before). (Thomson Reuters Legal)

That number is especially useful because it tries to measure active use inside organizations, not just curiosity.

Peer-reviewed field evidence: legal aid study (real-world usage, not just opinions)

A standout peer-reviewed contribution is a field study + companion survey focused on legal aid professionals, often cited as among the first real-world studies of lawyers using paid GenAI tools in practice. It reports strong perceived productivity benefits and intent to continue using tools after the study period—while also highlighting how access and training shape who adopts AI. (digitalcommons.lmu.edu)

This matters because it moves beyond “Would you use AI?” and looks at what happens when lawyers actually get tools, support, and time to try them.

What Lawyers Are Doing With AI Most Often

Across the best-regarded reporting and research, the most common uses cluster around text-heavy work:

  1. Research acceleration and synthesis
  2. First-draft writing (memos, letters, clauses, checklists)
  3. Summarization (long docs, deposition transcripts, case files)
  4. Contract review support (spotting missing/odd clauses and obligations)

At the same time, careful sources warn that legal AI can still generate confidently stated errors—so workflows increasingly emphasize verification, citations, and audit trails. (dho.stanford.edu)

Why Adoption Is Uneven (And What Predicts It)

Firm size, policy, and “approved tool” access

  • Larger firms and corporate legal departments more often have governance: policies, training, vendor reviews, and secure environments. (American Bar Association)

Training closes gaps

The legal aid field study reported differences in use that narrowed when people received access plus support—a strong hint that “non-adoption” can be a resourcing problem, not an attitude problem. (Reuters)

Reliability concerns slow full integration

Even when lawyers use AI, many organizations hesitate to make it “central” without better confidence in accuracy, confidentiality protections, and clear professional responsibility guardrails. (Thomson Reuters Legal)

What This Means for Clients and Businesses

If you work with lawyers (or you’re hiring one), the prevalence data points to a practical reality:

  • More lawyers are using AI already (especially for early drafting and research support). (Thomson Reuters Legal)
  • The quality difference often comes down to process: tool choice, privacy safeguards, and mandatory verification. (dho.stanford.edu)
  • Expect growing emphasis on AI policies and disclosure norms, especially in high-stakes litigation and regulated matters.

A Simple “Responsible Use” Checklist (What the best sources imply)

If you’re evaluating whether a lawyer or firm uses AI responsibly, look for:

  • A written AI policy (confidentiality, allowed tools, prohibited inputs)
  • Human verification steps (citations checked, quotes validated)
  • Secure environments for sensitive data
  • Training for staff (not just “try ChatGPT”)

These aren’t just nice-to-haves—they’re responses to documented error modes in legal AI. (dho.stanford.edu)

Bottom line

So, how prevalent is AI use among lawyers right now? Based on the strongest, most commonly cited sources available:

  • Roughly a quarter to around a third of lawyers/organizations report active GenAI use in recent surveys, with higher usage in larger organizations. (Thomson Reuters Legal)
  • Peer-reviewed field research suggests that when access + training are provided, lawyers often adopt quickly and report meaningful productivity gains—while still needing guardrails for accuracy and confidentiality. (digitalcommons.lmu.edu)