What AI SEO covers
AI SEO is a practice, not a platform feature. It groups the work that affects whether an AI-assisted search system can use a business as a source. Four questions define the scope, and they sit in dependency order.
- Can the system reach the site? Retrieval crawlers and user-triggered fetchers have to be allowed through robots.txt, firewalls and bot management, and the page has to contain its content in HTML rather than only after JavaScript runs.
- Can the system identify the business? One name, one address form, one description, and agreement between the site, registries, directories and profiles. This is the subject of entity SEO.
- Can the system retrieve the right passage? Answers are composed from passages, so a page that states its answer plainly and near the top is easier to use than one that buries it.
- Is there a reason to cite or recommend this source over another? Corroboration from independent sources, original material that does not exist elsewhere, and coverage of the comparison and decision questions people ask assistants.
DOCUMENTED Google’s own guidance for site owners lists the same foundations for its AI features: crawling allowed in robots.txt and by any CDN, content findable through internal links, important content available as text, structured data that matches the visible text, and up-to-date Business Profile information. It adds that no new machine-readable files, AI text files or special schema.org markup are needed to appear in AI Overviews or AI Mode.
What AI SEO is not
DOCUMENTED It is not a way to make a system name a business. Google states that meeting every requirement and best practice does not guarantee a page will be crawled, indexed or served. No other platform documents a guarantee either. A vendor promising a citation, a mention or a place in a shortlist is promising something no platform offers.
DOCUMENTED It is not a separate discipline from SEO on Google’s surfaces. Google’s page for site owners opens by saying that best practices for SEO remain relevant for AI features and that there are no additional requirements. The AI SEO vs SEO reference goes through what carries over and what is new.
It is not prompt injection, hidden instructions to models, or keyword stuffing aimed at assistants. None of the platforms document any mechanism by which text addressed to a model changes its source selection, and this publication treats such tactics as unsupported.
Why the term exists
TESTED The behaviour that made the practice necessary is measurable. Pew Research Center’s analysis of browsing data from 900 US adults found that 18% of Google searches in March 2025 produced an AI summary, that users who saw a summary clicked a traditional result link in 8% of visits compared with 15% when no summary appeared, and that users clicked a link inside the summary in 1% of visits. Those are US figures; no equivalent Australian panel study is known to this publication.
TESTED The research vocabulary came first. The term Generative Engine Optimisation was introduced in a 2023 paper by Aggarwal and colleagues, who defined a generative engine as a search system that composes answers with a large language model and measured how content changes altered visibility in those answers. AI SEO is the broader market term that grew around the same idea; the GEO reference covers the paper and its limits.
How the platforms document their side
The platforms do not publish ranking factors for their AI surfaces. What they publish is access controls and eligibility rules, which is the part of AI SEO that can be stated as fact.
| Platform | Retrieval or search agent | Training agent | Documented eligibility rule |
|---|---|---|---|
| Google (AI Overviews, AI Mode) | Googlebot | Google-Extended controls Gemini training and grounding, not Search | Page must be indexed and eligible for a snippet |
| ChatGPT | OAI-SearchBot (search index), ChatGPT-User (user-triggered fetch) | GPTBot | Sites opted out of OAI-SearchBot are not shown in ChatGPT search answers |
| Perplexity | PerplexityBot (search index), Perplexity-User (user-triggered fetch) | Not used for foundation model training per Perplexity | Allow PerplexityBot to appear in results |
| Claude | Claude-SearchBot (search index), Claude-User (user-triggered fetch) | ClaudeBot | Disabling Claude-SearchBot prevents indexing for search |
DOCUMENTED Each of these is taken from the platform’s own crawler documentation, read on 9 September 2026 and linked in the sources. Everything beyond access and eligibility, including which sources a system prefers and why, is either observed behaviour or inference, and is labelled that way wherever this site discusses it.
How AI SEO applies in Perth
A Perth business asking “does this apply to me” is really asking whether people in Perth use these systems for the questions that lead to it. That is a measurement question, and it is the reason this publication runs an observation programme rather than restating national claims.
What can be said now:
- DOCUMENTED Google lists Australia among the countries where AI Overviews are available, and Google Australia announced AI Mode’s rollout in Australia in English on 8 October 2025. Both surfaces are live for Perth searches.
- DOCUMENTED Google reports traffic from AI features inside the normal Search Console Performance report under the Web search type, so a Perth site’s AI Overview and AI Mode clicks are already counted in the figures it sees, not separated out.
- OBSERVED The source mix for Perth recommendation prompts has now been counted, and it is platform-specific rather than uniform. In the Perth AI Search Study 2026, across 2,400 AI answers collected 8 to 10 September 2026, Google AI Mode drew 85.8% of its citations from Google-hosted business panels, AI Overviews drew 38.2% to 70.9% from businesses’ own websites and Gemini 42.0% to 77.6% (the lower figure counts only businesses the study identified, the higher every site that looks like a business website), and ChatGPT cited Google zero times across 1,794 citations. Directories, review platforms and forums sit high in the combined cited-domain table. Whether that mix holds beyond this 200-question sample is not established.
The practical Perth checks are the same as anywhere: whether the retrieval agents above are allowed through, whether the business resolves to one entity across ABN Lookup, its Google Business Profile, directories and its own site, and whether its pages answer the questions Perth buyers actually ask. The guides cover each check.
What is documented and what is inferred
- Documented: access controls per platform; Google’s eligibility rule; Google’s statement that no special markup is required; availability of AI Overviews and AI Mode in Australia; the Pew click figures for the US.
- Tested: the GEO paper’s finding that content changes altered visibility in its benchmark, with the paper’s own caveat that effects varied by domain.
- Inferred: the four-question scope above is this publication’s synthesis of the documented foundations, not a platform’s statement.
- Not established: any claim that a particular signal causes citation, and any platform-wide ranking formula. Perth frequencies are no longer unmeasured: the first observation round reports them for 200 commercial questions on four platforms. They are associations observed over three days, not causes, and the measured characteristics account for roughly a quarter of the variation between businesses.