ABP/AU
AI Search8 min readUpdated 2026-09-05

Search Expanded. It Did Not Die.

Google still matters. But discovery now runs through generated answers, conversational queries and recommendation surfaces at the same time.

Contents

Author

ABP Intelligence

Search & answer engine practice

ABP's search practice works across technical SEO, semantic architecture, answer engine optimisation and generative engine optimisation for Australian and international organisations.

Reviewed 2026-09-05 — Published 2026-08-18

The common claim is that AI killed search. It did not. It widened it.

A customer still opens a search engine. They also ask an assistant a complete question, read a generated summary, and act on a recommendation they never clicked through to verify.

That means visibility is no longer a single ranking position. It is a set of conditions under which several different systems decide you are worth surfacing.

One question, asked three ways
Classic ranking
Answer extraction
Generative synthesis
Consistent presence across all three
Figure 01 — one asset, three retrieval paths: the same authority read by different systems.
01

Three surfaces, one authority

Classic ranking, answer engines and generative experiences read the same underlying assets — your structure, your entities, your evidence — and reward them differently.

Strategies that treat these as three separate projects duplicate effort and dilute authority. Strategies that treat them as three views onto one asset compound.

  • Classic search rewards accessibility, relevance and link equity.
  • Answer engines reward extractable, unambiguous passages.
  • Generative engines reward original information and citations.

Where discovery effort now goes

  • Authority, original information & evidence45%
  • Structure, entities & machine accessibility33%
  • Classic technical & link foundation22%
02

What actually changes in practice

The work does not become less technical. It becomes more explicit. Anything implied by design, tone or convention has to be stated in a way a machine can verify.

The most common failure is not thin content. It is excellent content that no system can attribute to a clearly defined organisation.

03

How to measure a surface you cannot rank in

Positions are the wrong unit. The useful questions are whether machines describe your business correctly, whether they cite you in your own category, and whether that citation share is moving.

Monitoring belongs in the operating rhythm, not in a quarterly report. It is an input to the next cycle of work.

Frequently asked

Short Answers.

Does SEO still matter with AI search?
Yes. Answer and generative engines draw on content they can crawl, parse and trust. Technical accessibility and semantic clarity are prerequisites, not legacy work.
Should we build separate AEO and SEO programmes?
No. Build one authority programme with different output requirements. Separate programmes duplicate research and split the signals you are trying to concentrate.