Victorian lawyers’ AI use rises to 55%, but only 26% report lower client fees

AI use in Victorian legal practice rose from 37 per cent to 55 per cent in a year, according to the 2026 census of 2,221 of the state’s lawyers, conducted by the Victorian Legal Services Board and Commissioner’s legal services research centre and reported by The Guardian. The same census found that 26 per cent of responding lawyers said their use of AI had resulted in lower client fees. Together the figures describe a profession that has adopted the technology faster than it has changed what it charges, a gap that shapes how the result should be read.

The Guardian’s report is the source for every census figure here. The census report itself, including its methodology, response rate and question wording, was not available in the material reviewed. The percentages are self-reported responses from a survey whose sampling cannot be checked, and should be read as a snapshot of what lawyers told researchers rather than as audited or population-wide measurements.

Adoption by practice area and tool

Use was uneven. The census put it at 85 per cent in trade practices, 78 per cent in banking and finance and 78 per cent in information technology and telecommunications, against 29 per cent in criminal law, 45 per cent in family law and 46 per cent in administrative law. Most respondents reported using general-purpose tools: 79 per cent used consumer chatbots such as ChatGPT, Claude, Gemini or Copilot, while 22 per cent used legal-specific tools alone. The age gap recorded in 2025 had largely closed, though use remained low among lawyers aged 60 and over.

Why lower fees are the exception

Just 26 per cent of respondents said AI had led to lower client fees. The figure was higher among lawyers working for business clients, at 34 per cent, than among those serving individual fee-paying clients, at 23 per cent. A further 18 per cent said AI had allowed them to take on work that would previously have been uneconomical.

Dr Georgie Rychner, the census’s principal researcher and a co-author, attributed the stronger effect in corporate and large-firm settings to the clients themselves. As reported by The Guardian, she said they are repeat users of legal services with high legal capability who may use AI in their own work, leaving them better placed to argue for lower prices, and that the pattern was less evident in the community legal sector. That is her interpretation of the data, not a causal finding the census measured.

The question matters too. It records whether lawyers said AI had resulted in lower fees, not what clients were actually billed; savings could arrive as lower hourly rates, shorter bills, fixed fees or not at all. College of Law commentary on time-based retainers argues that only time quantifiable and attributable to a matter can be charged, and that the benefit of time saved belongs to the client. The commentary addresses conduct rules rather than a regulatory ruling, but it explains why efficiency need not appear as a headline fee cut.

The work clients bring with them

The census also records a burden flowing the other way. Some 46 per cent of lawyers said clients were arriving with inaccurate or misleading information, and 41 per cent said clients presented AI-generated documents that required correction. Nearly half said client use of AI increased the time and effort required, against 19 per cent who said it reduced them.

Rychner told The Guardian that lawyers more often encountered misleading claims about their legal problems than evidence of better understanding, and that more research was needed into client perspectives. She also raised a possible explanation for weaker than expected efficiency gains: the 2025 survey found 89 per cent of practitioners said they took steps to verify AI output, work that might have offset the time saved. That link is a hypothesis, and no net-hours measure is available to test it.

Billing as a conduct question

Guidance treats fees as part of a lawyer’s obligations rather than a purely commercial choice. The Law Institute of Victoria guidance summarised by Barry Nilsson requires practitioners not to put confidential or privileged information into open-source or commercial AI systems without client consent, to verify AI-generated outputs including case citations, to disclose AI use to clients, and to ensure fair billing for work completed with AI assistance.

Enforcement has followed misuse. The Guardian reported that a Victorian lawyer became the first in Australia to face sanctions over false citations and was stripped of his ability to practise as a principal lawyer, that regulators in Western Australia and New South Wales opened similar investigations, and that the federal court issued new rules on generative AI in April. Barry Nilsson’s account of the 2024 Dayal case says the Victorian Legal Services Board varied the practitioner’s certificate on 19 August 2025, barring principal practice and trust money and requiring two years of supervised employee practice. College of Law commentary describes a cancellation; the accounts conflict and primary records were not available.

On fees, College of Law commentary describes it as foreseeable that regulators may see more complaints about lawyers failing to pass on AI-related reductions, and suggests such complaints would be examined through fiduciary, ethical and costs disclosure obligations rather than as a purely commercial dispute. It records minimal reported cases so far, and nothing in the material reviewed shows that complaints of this kind have risen.

Other surveys, different questions

Two other surveys offer comparison points, though neither is directly comparable with the Victorian figures. A Best Lawyers press release for its third Australia Legal Market Report, based on responses from 211 law firms, said 72 per cent of firms reported no impact from generative AI on billing practices, and that active communication of AI use was far more common at large firms than at small ones. The Law Society Journal reported an Axiom survey of more than 600 senior legal leaders globally: 79 per cent actively using AI, 6 per cent of firms charging less for AI-assisted work, 34 per cent charging more and 58 per cent not cutting rates. The samples and questions differ, so the figures cannot be merged into a national trend.

What the census leaves open

The strongest conclusion the evidence supports is narrow. AI use among responding Victorian lawyers has become mainstream, and most do not report lower client fees as a result. What the census cannot show is whether savings were absorbed by verification and correction work, retained through billing practices, passed on in forms the question did not capture, or some combination of all three.

Three things would sharpen the picture: the response rate and exact question wording from the primary report, billing data rather than perceptions, and client-side research, which Rychner said was needed. Until then, the correction work is the part that most complicates the productivity story, because the census records more lawyers reporting extra time and effort than reporting savings, and it offers no net measure to settle the difference.

A confident legal citation and a polished AI draft can fail in the same quiet way, which is why the Victorian census figures invite a second look at who checks the output. Adoption has grown quickly, yet the time spent verifying material and correcting client-supplied documents complicates any simple efficiency story. That habit of separating a plausible result from a verified one, and asking where human review belongs, is part of cybersecurity fundamentals, where systems, access and evidence are examined rather than assumed. Readers who want to make that scrutiny deliberate can explore The Hack Academy’s online cybersecurity training programme, which lists cybersecurity fundamentals among its subject areas and sets self-paced learning and practical exercises alongside the theory.