TL;DR: Google’s answer layer now has two distinct surfaces: AI Overviews (the generated summary above classic results) and AI Mode (a fully conversational search experience). Both select and cite a small set of sources. For agencies, they are separate battlegrounds with separate tracking — a client can be cited in one and absent from the other.
Two surfaces, not one
AI Overviews appear on top of the classic results page for queries where Google decides a synthesized answer helps. They cite a handful of pages — and being one of them puts your client above position one. Google documents the feature and its guidance for creators on Google Search Central, and announces changes on the Search blog.
AI Mode goes further: a conversational surface where the user can ask follow-ups and Google answers with model-generated responses, citing sources along the way. It behaves less like a results page and more like an assistant that happens to be built into search.
The practical consequence: “ranking on Google” is no longer one number. A client can hold position three in the classic results, be cited in the AI Overview, and be invisible in AI Mode — three different outcomes on one query.
What earns citations in Google’s answer layer
Google is explicit that there is no special markup that buys AI Overview inclusion — the guidance points back to the fundamentals: helpful content, sound technical SEO, structured data that accurately describes the page. In our experience the same properties that make a page quotable for ChatGPT help here: clear question-shaped headings, self-contained answer paragraphs, honest structured data, and fast, clean raw HTML.
That last part matters more than most agencies expect. The answer layer works from what Google can crawl and understand — and pages whose substance only materializes after JavaScript runs are working with a handicap in every AI surface at once (we covered why in AI crawlers don’t run your JavaScript).
Tracking it like an agency
Position tracking alone misses both surfaces. Preferium tracks, per keyword, whether the client is cited in the AI Overview and whether they are cited in AI Mode — alongside the classic Google and Bing rank. Over time, that gives you three trendlines per keyword, and a defensible answer to the question every client is starting to ask: “are we in Google’s AI answers or not?”
Key takeaways
- AI Overviews and AI Mode are separate surfaces with separate citation sets — track both, per keyword.
- No magic markup exists; citations flow from crawlable, well-structured, genuinely useful pages.
- Server-side clean HTML compounds across every AI surface — Google’s and the assistants’.
- Report citations next to rankings. Clients see the AI answers daily; your reporting should too.
