Which sources do AI assistants cite about electricians in Sydney?

Measured 2026-09-08 · Second collection

The top 3 cited sources, by engine

Sample: 43 prompts, responses collected 2026-09-08 to 2026-09-08 (43 from ChatGPT, 43 from Perplexity, 35 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: 22 prompts naming a specific business, per the TOP Score rules, and 67 prompts that stated no location, per the method. Percentages are the share of that engine's responses containing at least one citation of the domain. Ties: on ChatGPT, second place is a three-way tie at 6 citations between baysidelectricalgroup.com, calibreconnect.com.au and connectelectric.com.au. On Perplexity, third place is shared at 14 citations by dmelectric.com.au and thequoteyard.com.au. On Google AI Overviews, second place is shared at 3 citations by empireelectricalcontractors.com.au and nsw.gov.au.

What the data says

Electrical work is licence-gated, so ChatGPT leans on nsw.gov.au here the way it leans on Moneysmart in financial advice: verify first, recommend second. Google AI Overviews does the same thing with energy.nsw.gov.au at the top. Below the regulator it's electricians' own websites rather than directories, and the firms being cited are the ones with real content on switchboards, EV chargers and compliance. Perplexity is the outlier here: Reddit threads lead at 46.5%, ahead of any electrician's own site. Same lesson as plumbing: the trades categories are directly winnable, and the early winners are mid-sized firms that write.

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.