The short answer
Google shows a page of results and a map with many pins. An AI assistant gives an answer with a few names: in our Sydney check ChatGPT named 2.4 to 3.2 businesses per answer and Claude 5.8 to 7.5. There is no page two. Either a business is in the answer or it is not.
How we measured it: between 27 October 2026 and 29 October 2026 we asked OpenAI’s and Anthropic’s own models (the ones behind ChatGPT and Claude), with web search on, three everyday questions (for example “What’s the best barber in Newtown?”) for 156 trade-and-suburb combinations in Sydney, 936 answers in all (as of 29 October 2026), and stored every answer, every business named and every website each assistant read.
Four differences that matter
- Fewer winners. ChatGPT: 2.7 names on average. Claude: 7. Google: ten results plus a map.
- Different sources. Google ranks pages. The assistants read a few pages and quote them: Reddit, Word of Mouth, Fresha, Yelp, Tripadvisor, review pages. Your Google rank is not what they read.
- Answers move. 16 different businesses typically appeared across six answers for one suburb. One check tells you little; a pattern tells you where you stand.
- Mentioned is not recommended. 18% of businesses that appeared were only mentioned. An assistant can name you in passing and recommend someone else.
What to do
Keep doing what works on Google. Add the pages the assistants read for your trade, keep your details consistent, and check both ChatGPT and Claude for your suburb. The free check does that six times and shows the pattern.