AI for law firms: what to use it for, what to keep it away from, and how to show up in AI answers

Strategy11 min readReviewed 11 September 2026

AI is already inside your practice whether you have a policy or not, and it is already between you and your next client. Two things are true at once. Generative AI is a genuine productivity tool for the parts of legal work that are drafting, summarising and searching, and it is now a channel through which prospective clients find and shortlist a lawyer, because a growing share of them ask ChatGPT, Gemini or Google’s AI Overview before they ever open a firm’s website. This guide covers both halves: how to use AI without breaching a court rule or a client’s confidence, and how to make sure the AI a client asks knows your firm exists.

It is written for principals and practice managers, not technologists. Nothing here needs a developer.

Where AI earns its place in a practice

The useful applications are unglamorous. They are the tasks that consume junior hours and produce a first draft that a lawyer then corrects.

• First drafts of routine correspondence. Engagement letters, file notes from a dictated summary, plain-English explanations of a process for a client. The lawyer still owns every word that goes out.
• Summarising long documents you supply. A 90-page contract, a bundle of medical reports, a transcript. Summaries are a starting point for reading, not a substitute for it.
• Intake and triage. A structured intake assistant that asks a prospective client the six questions your receptionist would ask, and routes the enquiry to the right lawyer. This is where AI has the clearest commercial return, because it answers at 11pm when a family law client is actually searching.
• Marketing drafts. Outlines for guides, alternative headlines, a first pass at a Google Business Profile post. Everything published still has to pass the advertising rules, which are covered in our guide to legal advertising rules in Australia.
• Internal search. Asking questions of your own precedent bank, in a tool that does not send that bank anywhere.

What AI is poor at, in 2026, is the thing clients pay for: judgement about their specific matter, applied to current Australian law, with accountability attached. It hallucinates authorities, it does not know the case you have not told it about, and it cannot sign a certificate.

The rules that already apply

Australian courts and regulators have not waited for a national framework. The obligations below are live now, and a firm that ignores them is exposed in a way that has nothing to do with marketing.

Court practice notes on generative AI

The Supreme Court of NSW issued a practice note on the use of generative AI in proceedings, in force from early 2025, which restricts how AI can be used in affidavits, witness statements and expert reports and requires disclosure in some circumstances. Other Australian courts have published their own guidance. If your firm files anything, someone in it needs to have read the practice note for that court. The NSW note is linked in the sources at the end of this guide.

Confidentiality and privilege

Pasting a client’s matter into a consumer AI tool is a disclosure to a third party. Whether that breaches confidentiality depends on the tool’s terms, but the safe assumption is that a free consumer product may retain and train on what you give it. Enterprise agreements with a no-training clause, or tools that run inside your own environment, are the answer. The conduct rules on confidentiality have not changed; the number of ways to breach them has.

Privacy

The Office of the Australian Information Commissioner has published guidance on using commercially available AI products, which is direct about the point above: personal information entered into a public AI tool is a disclosure under the Privacy Act, and firms need a lawful basis for it. The guidance is short and worth reading in full.

Competence and supervision

A lawyer remains responsible for work product regardless of what drafted it. Every jurisdiction’s regulator has said some version of this. Practically it means AI output gets the same review a paralegal’s draft would, and a junior lawyer is never left to submit AI-drafted work unsupervised.

A one-page AI policy for a small firm

Large firms have 40-page policies. A five-lawyer firm needs one page that people will actually follow. It should answer five questions.

That is the policy. Review it every six months, because the tools and the practice notes both move.

The other half: being findable inside AI

This is the part most firms have not thought about. When someone asks an AI assistant “who is a good family lawyer in Parramatta” or “do I need a lawyer for a contested will in Victoria”, the assistant answers from what it has read. If it has never read anything that connects your firm to that question, you are not in the answer, and the person never reaches a search results page where you might have ranked.

Google’s own AI Overviews work the same way, drawing on pages Google has indexed and trusts. Google has published guidance on how its AI features select sources, and the short version is that it favours the same qualities as its ordinary results: clear, specific, credible pages that answer the question directly. There is no separate trick.

What makes a firm citable

• Specific pages for specific questions. A page titled “Contesting a will in Victoria: time limits, grounds and costs” is citable. A page titled “Wills and estates” with three paragraphs is not.
• A direct answer near the top. Assistants extract. If the answer is in the second paragraph in plain words, it can be quoted. If it is implied across six paragraphs, it cannot.
• Named authors with credentials. “Written by a partner with 15 years in family law” is a trust signal to a model, as it is to a person. Our guide to SEO for lawyers covers the credibility signals in detail.
• Consistent facts everywhere. The same firm name, address, phone number and practice areas on your site, your Google Business Profile and the legal directories. Assistants cross-check.
• Structured data. Schema markup for your organisation, your people, your services and your FAQs. It is the machine-readable version of your website.
• Letting the crawlers in. Some firms block AI crawlers in robots.txt without knowing. Check yours. A blocked crawler cannot cite you.

How to check where you stand

Do this once a quarter. It takes twenty minutes.

• Ask ChatGPT, Gemini and Perplexity the five questions a client in your best practice area would ask, including your suburb. Note which firms are named.
• Search the same questions in Google and note whether an AI Overview appears and who it cites.
• Ask each assistant directly what it knows about your firm. The answer tells you which of your pages it has read, and whether the facts are current.
• Check your robots.txt for AI crawler rules, and your Google Business Profile for anything out of date.

Firms are routinely surprised by the third step. An assistant that describes a firm using its 2019 practice areas is working from an old directory listing, which is a fixable problem.

What to do in the next 90 days

• Weeks 1 to 2: the one-page policy, an approved-tools list, and a read of the relevant court practice notes.
• Weeks 3 to 6: pick one internal use with a clear return, usually intake or first-draft correspondence, and run it properly with a review step.
• Weeks 7 to 12: the visibility work. Author bios with credentials, schema markup, a robots.txt check, and one genuinely specific guide per practice area written to be quoted. Then re-run the quarterly check.

If you would rather have the visibility half done for you, that is most of what a law firm SEO engagement now involves. If you want to see where your firm currently stands in AI answers before deciding, the free SEO audit includes it.