AI Overviews and AI Mode changed the search journey
Search Generative Experience, or Google SGE, was the early label. The live products are AI Overviews and AI Mode. They can summarise information, connect a searcher to supporting websites and extend one question into a longer research session.
For a founder or marketing lead, the practical question is whether Google can find the right page, understand the offer, trust its evidence and surface it while a buyer compares options. Start there before choosing tactics.
Google describes AI Overviews as a way to get the gist of a complex question, with links for further exploration. AI Mode is designed for nuanced questions, reasoning, comparisons and follow-ups. An Overview can also hand the searcher into a conversational exchange in AI Mode. Google said that handoff was live worldwide in its Search I/O 2026 update, which we checked on 10 September 2026.
The commercial consequence is a less linear path. A buyer may learn from a generated answer, open a source, refine the question and search a brand later. A single ranking position tells only part of that story.
One question can create several retrieval needs
Google's documentation for AI features in Search explains that AI Overviews and AI Mode may use query fan-out. The system can issue several related searches across subtopics and data sources before developing a response.
A buyer asking whether an SEO agency is suitable could prompt retrieval around location, services, sector experience, pricing, proof, risks and comparisons. They may type none of those separate searches, and website owners receive no trace of Google's inferred queries. Google needs useful pages and evidence beyond the exact phrase in the search box.
The answer is coherent coverage, rather than a page for every imagined subquery. Service pages, case studies and guidance should each have a clear role, then connect through descriptive internal links. Our guide to planning content for AI Mode query fan-out covers that implementation work while treating inferred queries as hypotheses, rather than keyword data.
Search foundations determine eligibility
Google's current generative AI optimisation guide says its AI features are rooted in core Search ranking and quality systems. Its technical guidance says a page must be indexed, eligible for a Search snippet and included in generative AI features through Search Console. Standard search requirements apply, with no AI-only markup layer, llms.txt file or special schema required.
Important pages must be crawlable and indexable, with useful content visible on the page. Internal links should make related services and evidence easy to find. Structured data should agree with what a reader can see. Page experience and accessibility matter when a person follows a source link.
Eligibility is only the starting point. It doesn't guarantee that Google will crawl, index, select or show a page. If an eligible site is absent while other sources appear, use the AI Overview citation diagnostic to separate an access problem from weak query fit or weak source value. If the structure itself is unclear, the structured content guide shows how to repair headings, page roles, visible evidence and machine-readable support.
For a service business, this foundation has a direct commercial effect. A technically eligible page can enter consideration. A clear page with credible proof gives both Google and the buyer more reason to keep it there.
Source visibility is becoming more explicit
Google said in May 2026 that it was rolling out more direct and inline links to original and relevant web content in AI Mode and AI Overviews. Its August website-owner update also said it had increased inline links and added website previews. We checked both announcements on 10 September 2026.
More visible links create an opportunity, not a promise. A page may be eligible and still not be selected for a particular response. The useful distinction is between making content available and giving Google a strong reason to use it.
Generic summaries are easy to compress because many websites make the same points. Original examples, named expertise, specific service boundaries, current pricing context, first-hand observations and documented outcomes are more useful to a buyer. They also give a generated response something precise to attribute. The guide to creating citation-worthy AI search content develops that evidence standard in more detail.
Put proof beside the decision it supports. A case study should explain what changed. A service page should state who the work suits and where its limits sit. An article should distinguish experience from inference.
Reporting and controls changed in 2026
Google's announcement on new controls and insights for website owners was updated on 31 August 2026 to say the features had rolled out to all websites worldwide. The dedicated Search Console reporting shows generative AI impressions, pages and countries. A separate inclusion control lets an owner decide whether the site can appear in and help ground AI Overviews, AI Mode and AI Overviews in Discover. We rechecked that worldwide status on 10 September 2026.
The control has a real tradeoff. Google says an opted-out site won't receive traffic or impressions from those generative AI features, while the choice isn't used as a ranking signal outside them. The AI Overviews opt-out decision guide explains how to weigh visibility, content use, measurement and commercial exposure before changing the setting.
The report answers a narrower question than a complete marketing dashboard. It can show where pages appeared in Google's supported generative AI experiences. It can't tell you whether the description was persuasive, whether a later branded search was influenced by that exposure or whether a lead was worth pursuing. Our framework for measuring AI search visibility combines the report with citation review, prompt sampling, analytics and sales evidence.
Traffic needs a commercial reading
AI features can change clicks, but their effect varies by query. A generic definition may be satisfied inside the result. A buyer choosing an adviser, checking a provider's work or deciding whether a service fits still needs evidence and judgement. That person may visit a source immediately, return through a branded search or arrive later on a high-intent page.
Sessions and click-through rate remain useful, but they need context. Watch generative AI impressions, branded demand, visits to commercial pages, qualified enquiries, assisted conversions and what prospects say in sales conversations. Those signals help distinguish lost informational clicks from influence that moved elsewhere in the journey.
If traffic or CTR has already changed, the AI Overviews traffic-drop diagnostic provides the right next step. It tests reporting, ranking, demand, feature exposure and conversion evidence before assigning a cause. That is safer than treating every decline after a product rollout as proof that AI Overviews caused it.
Agentic Search remains an emerging direction
Google is extending Search from answering questions toward completing some tasks. Its Search I/O 2026 announcements included information agents, expanded booking capabilities and custom experiences built inside Search.
Availability matters. As checked on 10 September 2026, Google said local-services booking and calling capabilities were rolling out in the United States, while custom Antigravity experiences were due to start with US Google AI Pro and Ultra subscribers. The announcement did not establish those capabilities as generally available to Australian service businesses. The conversational handoff from AI Overviews to AI Mode was described as worldwide.
Australian operators should watch agentic Search as an emerging product direction. Clear service data, accessible interfaces, current availability and well-structured proof are useful website investments today. A feature-specific playbook can wait until the buyer use case and Australian availability are clear.
Choose the next job from the evidence
The right response depends on where the problem appears. Start with AI search measurement to establish whether the business is visible. If Google can access the site but rarely uses it as a source, work through the citation diagnostic. For a performance change, use the traffic-drop diagnostic.
When one buyer question spans several topics, move into query-family content planning. When the concern is whether Google should use the content at all, make the inclusion-control decision before changing Search Console settings.
Some problems cross reporting, content architecture, technical eligibility and internal linking. That calls for joined-up SEO planning. If the constraint is the template, accessibility, proof presentation or page architecture, website design is the more direct route.
AI Overviews and AI Mode raise the value of a website that's easy to understand and verify. Identify whether the current constraint is eligibility, evidence, measurement, coverage or control, then fix that specific job.