ChatGPT's Citation Shifts Signal a Deeper Change in Search

Search Engine Journal has reported on new Ahrefs data showing that the types of websites ChatGPT cites in its US answers have shifted noticeably. Government health bodies, regulators and independent product testers gained ground, while large consumer health and finance explainer sites lost it. The movement is small in absolute terms but clear in direction: the assistant appears to be reaching for first-hand sources rather than second-hand summaries.

Ahrefs compared millions of answers from an earlier month with millions from a later one, and said the sites it tracked across both periods moved the same way. High-traffic community and encyclopedia domains also appeared in a smaller share of answers, though they remain among the most cited. The company cautions that its data reflects a single model over a limited window, so patterns can change with any update. That caveat matters as much as the finding itself.

Google Rankings Are No Longer the Whole Picture

For two decades, SEO meant earning a position on a page of blue links. The Google algorithm rewarded pages that matched a query well and built trust through links. AI assistants break that model. There is no page of ten results — there is one answer, and a short list of sources underneath it.

That change rewards different content. A well-written explainer can rank on Google and still be skipped by an AI assistant that goes straight to the study, the regulator or the person who actually tested the product. Being useful is no longer enough. Being the origin of the information is what earns a citation.

This does not make classic SEO pointless. Google still drives enormous traffic, and AI systems often draw on the same underlying index. But AI SEO is now a separate job with its own scoreboard: how often your domain is named as a source, on which topics, and who gets named instead.

What This Means for Australian SMBs

Most Australian small and mid-sized businesses will never out-cite a federal agency or a global publisher. That is fine, because the same shift creates an opening. If AI tools prefer original information, then the tradesperson who publishes real job costs, the clinic that shares its own outcome data and the retailer that tests its own stock all hold something no aggregator can copy.

Local relevance adds another layer. Australian buyers ask about local rules, local suppliers and local pricing, and thin overseas content rarely answers those well. Content that is original, specific and clearly dated is easier for a machine to trust — and easier for a customer to trust too.

What You Can Do Now

  • Check which sources AI tools cite for your five most important topics, then compare them with your own pages.
  • Publish at least one piece of original material each quarter — your own data, pricing, testing or customer results.
  • Give every page a named author, a date and a clear source for each factual claim.
  • Keep your Google basics strong: fast pages, clear headings and accurate business details.
  • Track AI citations the way you track rankings, and review after each model update.

MS&VG works with Australian businesses on exactly this puzzle, pairing practical SEO work with an AI-powered SEO service built to help your content get found and cited.