How Do You Get a Law Firm Into ChatGPT, Gemini and Google AI Overviews?

A prospective client types "do I have a claim if I was partly at fault for a car accident in Queensland" into ChatGPT, or asks Google and reads the AI Overview at the top without scrolling. The answer names two or three sources. If your firm is one of them, you have been recommended by the most trusted referrer the client has. If it is not, the client may never see your name, however well your service page ranks underneath.

This article covers how a law firm gets into those answers. It is written from running a content system for a Melbourne litigation firm whose organic clicks went from 785 to 5,217 a month this year, and whose new enquiries now include a growing number of people who say an AI assistant sent them. The short version is that the same work produces both results, and most of what is sold as "GEO" or "AI optimisation" is ordinary SEO with a new invoice attached. The longer version, including what is actually different for AI, is below.

What actually happens when someone asks an AI a legal question

Each of the major engines works the same way at the level that matters to a law firm. The model does not "know" who the best family lawyer in Sydney is. It turns the question into one or more search queries, runs them against a search index, reads the top results, and writes an answer with citations to the pages it relied on.

The four engines that matter and where each gets its sources. The page in the middle is the only part you control.

Two consequences follow from this and they settle most of the arguments about AI search.

The first is that a firm with no organic presence has no AI presence. There is no separate door. The litigation firm's pages started being cited by AI assistants in the same months they started ranking, because the engines were reading the same index everyone else was. Our SEO guide for law firms is the foundation this article sits on, and if you are not ranking at all, why your law firm is not ranking comes first.

The second is that the engines read Google's index differently from how Google ranks it. An AI Overview for "how long do I have to make a TPD claim" will often cite a 900-word article from a mid-sized firm over the practice-area page of a national brand that outranks it, because the article answers the question in its first paragraph and the brand page does not. That is the opening for a smaller firm, and it is the part of this that is actually new.

How big a deal is AI search for a law firm right now?

Big enough to plan for and not so big that it replaces anything. The numbers, with sources, because the ones circulating in agency pitches rarely have any.

Read together, these say three things. People use AI tools heavily, and in Australia almost half the adult population uses ChatGPT in a given month. When Google shows an AI answer, roughly half the clicks that used to go to the results disappear. And people still verify a lawyer before calling, which is where your reviews and your Google Business Profile do the second half of the job.

What the numbers do not say is that organic search is finished. Organic still delivers the largest single share of traffic to law firm sites we manage, and the AI-referred visitor, where we can attribute one, converts at a higher rate because they arrive having read a summary that named you. The practical position is that AI citations are an addition to the ranking you already need, and a firm that is not ranking has a bigger problem than AI.

The eight signals that get a law firm cited

The engines are different products but they are reading for the same things. These are the eight we build into every page in the litigation firm's system, in the order of how much difference each one makes.

1. The page answers one question, and answers it in the first sixty words

An engine assembling an answer wants a sentence it can quote. A page titled "Contesting a Will in Victoria" that opens with three paragraphs about the firm gives it nothing. A page titled "Can I contest a will in Victoria if I was left out?" that opens with "Yes, if you are an eligible person under Part IV of the Administration and Probate Act and you apply within six months of probate" gives it exactly what it needs, and the explanation can follow.

The litigation firm's articles that are cited are the ones built this way: one question, a direct answer, then the detail. The articles written as general overviews rank but are rarely quoted. Our content writing guide covers the structure.

2. A named lawyer with a verifiable footprint

Every substantive page carries the name of the lawyer who practises in that area, with a bio page, a photo, their admission and their experience, marked up with Person schema. The engines cross-check this. A named author who appears on the Law Institute register, in Doyle's Guide, on LinkedIn and in a couple of press mentions is an entity the model can trust; "the team at [firm]" is not.

The firm in our example has the advantage of real brand volume and real awards, and lawyers whose names people search. That is not something an agency can manufacture, and it is a large part of why its content is trusted so quickly. If your firm has these assets and they are not on the page, you are leaving your strongest signal off.

3. Cited sources inside the content

The Princeton study on generative engine optimisation found that adding statistics lifted a page's visibility in AI answers by around 40%, and adding citations to authoritative sources by more than that for lower-ranked pages. For a law firm this is natural: link to the legislation, the court's own guidance, the regulator, the ABS. A page that cites the Act it is explaining looks like an authority to a model in a way a page that paraphrases it does not.

4. Consistent firm details everywhere

Name, address, phone, practice areas and lawyer names, identical across the website, Google Business Profile, the Law Society or Law Institute listing, Doyle's, LinkedIn and the directories. Engines resolve a firm as an entity by matching these; inconsistency breaks the match. Our Google Business Profile guide for lawyers covers the profile side.

5. Reviews, with volume and recency

The BrightLocal finding that 97% of people check AI recommendations against reviews is the reason this is in the top five. The engines also read reviews directly when answering "best X lawyer in Y" prompts, and a firm with 40 reviews at 4.9 beats a firm with 6 at 5.0. Reviews on Google first, then on the legal directories the engines index.

6. Outside mentions

Third-party pages mentioning the firm by name in the context of its practice area: local press, a podcast, a bar association piece, a sponsorship page, an industry body. One analysis of Gemini citations found that around 90% went to third-party editorial content rather than firm-owned pages, which means that for "who is a good X lawyer" prompts the engine is often quoting someone else talking about you. The link building guide for law firms covers how to earn these without buying them.

7. Structure a model can parse

Question headings, short paragraphs, tables for anything comparative, numbered steps for anything procedural, a defined term where the law uses one. This is the same formatting that makes a page readable on a phone, and it is why tables appear in every article we publish. Schema on top: Article with an author, Person for the lawyer, LegalService and LocalBusiness for the firm. Schema does not get you cited on its own, and it helps the engine confirm what it has read.

8. Freshness that is real

A visible "last updated" date, a dateModified in the schema, and content that actually changed. Ahrefs found AI-cited pages skew newer than the standard results for the same query. Legal content has a natural reason to update: thresholds, limitation periods, fee scales and case law move every year. A page refreshed with the current figure and the current date beats an identical page from 2023.

The robots.txt mistake that removes you from ChatGPT

This one is worth its own section because it is silent, common and takes five minutes to fix.

When AI crawlers first appeared, a lot of developers and some SEO agencies added blanket blocks to robots.txt to keep sites out of model training. That is a legitimate choice. The problem is that the search crawlers were blocked along with the training crawlers, and the search crawlers are the ones that fetch your page to cite it.

Check your robots.txt now. If it contains User-agent: * followed by Disallow: / under a block meant for AI, or lists OAI-SearchBot or Claude-SearchBot as disallowed, you have opted out of being cited in ChatGPT and Claude. We find this on roughly one law firm site in ten, usually with nobody at the firm aware it was done. A firewall that challenges unknown bots (common on managed WordPress hosting) can have the same effect without a line in robots.txt; the test is to request a page as OAI-SearchBot and see what comes back.

What is different in Australia, the US and the UK

The engines are global and the mechanics above apply everywhere. The differences are in rollout, regulation and what the engines cite.

The Australian point that matters most is the personal injury advertising rules. An AI Overview quoting your page about compensation amounts is republishing your content; the content has to comply in the first place, and ours does because we write to the rule rather than around it. Our piece on whether personal injury lawyers can advertise in Queensland covers the boundaries.

For UK firms, the SRA's price transparency requirement means pages for conveyancing, probate and some other services carry mandatory pricing information; those pages are, in our testing, cited heavily for "how much does X cost" prompts precisely because they contain a concrete figure. For US firms, the directory layer is deeper and the engines lean on it more for "best lawyer" prompts, so the Avvo and Justia profiles need the same attention as the website.

How to measure whether it is working

Until this year, the honest answer was "ask every new client how they found you". That still matters, and there is now data to put beside it.

The litigation firm runs the prompt test monthly against a question list built from Search Console and from the enquiries its lawyers field. The firms cited beside it are, nearly always, the firms with the next-best question articles rather than the firms with the biggest brand, which is consistent with everything above.

What the "GEO" pitch gets wrong

A new discipline has produced a new set of invoices. Some of the advice is sound and some of it is a way of charging for ordinary SEO twice. The claims we would push back on:

The honest version of GEO is a checklist, and most of it was already on the SEO checklist. Where a firm is paying separately for it, the question to ask is which of the eight signals the work actually changes.

The 90-day plan

This assumes the firm already has a site that is indexed and some content that ranks. If it does not, the first ninety days belong to the SEO guide and to how long SEO takes for a law firm.

The firm in our example did not do this as a 90-day project. It did it as the standing way the site is run, with articles going out weekly, each reviewed by the responsible lawyer, each linked into its hub, and the measurement in the monthly report. The AI citations arrived as a by-product of that work rather than as a separate project. Our article on how law firms are using AI in marketing covers the production side, including what to systemise and what has to stay with the lawyers.

What to expect, and when

AI citations lag rankings by a few weeks, because the engines are reading an index that has to pick the page up first. On the litigation firm's timeline, pages that reached the top five for their question were being cited in AI Overviews and Perplexity within the following month, and in ChatGPT somewhat later, since Bing's index moves more slowly than Google's for a mid-sized Australian site.

The leads follow the citations, and they are a different kind of lead. Someone who read an AI answer that named you and quoted your explanation of their problem arrives knowing what you do and trusting that you know it. The first call is shorter and the conversion rate is higher. That is the business case, and it is the same business case as ranking, made stronger because fewer firms are in the answer than on the page.

If you want to know where your firm stands today, book a call and we will run the baseline prompt test and the Search Console check with you. We do this for personal injury, family, litigation and commercial firms in Brisbane, Sydney, Melbourne and Perth, and for firms in the US and UK.