Which sources do AI assistants cite about commercial lawyers in Sydney?

Measured 2026-09-21 · Previous measurement 2026-09-07

The top 3 cited sources, by engine

Sample: 30 prompts, responses collected 2026-09-21 to 2026-09-21 (30 from ChatGPT, 30 from Perplexity, 24 from Google AI Overviews). Google AI Overviews has a lower response count because some prompts returned no AI Overview, and those are excluded from that engine's denominator. Excluded from citation counts: 20 prompts naming a specific business, per the TOP Score rules, and 70 prompts that stated no location (see the correction note below). Percentages are the share of that engine's responses containing at least one citation of the domain. Google AI Overviews answered all 30 questions but linked no source at all in 16 of them, which is why its counts sit so low. Ties: ChatGPT's second place is shared at 5 citations by business.gov.au and lawconnect.com; Google AI Overviews' second and third places are both cut from a tie of 25 domains on a single citation each, ordered alphabetically.

What the data says

This category is a live demonstration of what wins citations: the firm at the top is not Sydney's biggest or oldest commercial practice, it's the one that publishes the most plain-English answers about contracts, structures and disputes, and the firms around it, LegalVision, Prompt Law and Lazarus Legal, fit the same pattern. The engines reward the publishing, not the pedigree. For an established firm, that cuts both ways: the incumbents' brands aren't protecting them in AI answers, and the gap that let younger, content-heavy firms take the top spots is a gap anyone willing to write can contest.

A fortnight on, that finding has held and firmed up rather than moved. Sprintlaw is still first on both ChatGPT and Perplexity, and on Perplexity it has gone further, from fourteen of thirty answers to sixteen, with LawConnect and LegalVision moving up alongside it. The one real change is on ChatGPT, where Wikipedia has dropped out of the top three and business.gov.au has come up into it, so the reference layer in this category now leans a little more towards the government explainer and a little less towards the encyclopedia.

How to read this page

• 1Citations are not mentions. This page measures which sources the engines link when they answer, not which businesses they name. Who gets named is the TOP Score, and it's measured separately.
• 2Engines disagree, so we never pool them. Each engine is reported on its own row because an average hides a total absence on the engine your buyers actually use.
• 3This is a measurement with a date on it. Models and retrieval change; a source moving between collections has not necessarily done anything. The method, cadence and limits are on the standard page.
• 4Ties are shown alphabetically and single-page sources are named. Where sources tie on citations, the table orders them alphabetically and the tie is stated under it. A domain's count can rest on a single page, often a "best of" listicle, and that concentration is itself part of what this site measures, so where it's true of a listed source we say so. Domains are counted by exact hostname, so government subdomains count separately.