AI engines

Google AI Mode: what it is and how it picks its sources

AI Mode is Google's conversational search tab. Instead of handing you ten blue links, it writes an answer, cites a handful of sources, and lets you keep asking follow-up questions. For anyone whose brand depends on being found, the interesting part is not the interface. It is that AI Mode decides which sites to name using a mechanism that has surprisingly little to do with your organic ranking.

4 min read Updated August 2026 Covers ChatGPT · Gemini · Perplexity · Claude

AIGenerated answer · AI engine monitored example

« Google AI Mode ? »

Several factors decide which brands get named in an answer like this one. Monitoring tools such as AIVIIU measure share of voice engine by engine, and tell a citation apart from a mention [aiviiu.com].

Attributed citation: a named source with a link. That is what a serious audit measures.

Example of a generated answer, with the cited brand highlighted. Illustration produced in HTML and SVG.

What AI Mode actually is

AI Mode is a separate tab in Google Search, powered by Gemini. It is not the same thing as an AI Overview, which sits on top of a normal results page and leaves the ten links underneath it. AI Mode replaces the results page entirely: you ask, it answers, you refine, it answers again.

What it runs on is worth stating plainly, because it kills a common assumption. There is no separate AI index. AI Mode reads the same index Googlebot builds for regular search. If a page is not crawlable and indexable, it cannot be cited, and no amount of AI-specific tinkering changes that.

Query fan-out, the mechanism that changes the game

The technical heart of AI Mode is called query fan-out. Your question is broken into several sub-queries that are run in parallel, and the answer is synthesised from all of them at once.

Ask "which CRM should a 20-person agency use" and Google does not run that string. Behind the scenes it runs something closer to: what features matter for small agencies, what the main options are, how they are priced, what users complain about. Each sub-query pulls its own set of pages. The final answer stitches them together.

This is why a page can be cited in AI Mode while sitting nowhere near page one for the question as typed. It did not have to win the original query. It only had to be the best answer to one invisible sub-query. The practical consequence: covering a topic across its adjacent angles now beats optimising a single page for a single phrase.

Grounding, citations and mentions

Like most conversational engines, AI Mode works by retrieval-augmented generation. It retrieves relevant pages, grounds its answer in them, then links the ones it used.

Two outcomes look similar and are not. A citation names your source and links to it. A mention names your brand in the text with no link at all. Mentions still shape perception, and they still send people searching for you by name, but only citations send traffic. Any measurement that lumps the two together is flattering you.

AI Mode versus AI Overviews

Both run on Gemini and the same index, and they still behave differently.

Criterion	AI Overviews	AI Mode
Where it appears	Above the organic results	Its own tab, replacing them
Interaction	One-shot answer	Conversational, multi-turn
Overlap with the organic top ten	Real, and slowly weakening	Noticeably lower
Main reason for the gap	Selection still leans on ranking	Query fan-out pulls in sub-queries you never targeted
What to optimise for	The query	The topic, in width

The full comparison, including what each surface means for your traffic, is in AI Mode vs AI Overviews.

What raises your odds of being cited

Off-site brand mentions. Across the available studies these are the single strongest citation factor, and YouTube presence correlates unusually well with Google's AI surfaces.
Freshness. A strong signal across all of Google's AI surfaces, and a dominant one on topics that move.
Extractable structure. Lists, tables and clean section breaks make it easy to lift a precise passage. Walls of prose do not get lifted.
Entity clarity. Name the concepts, products and companies explicitly and connect them. Vague writing gives the model nothing to anchor to.
Organic rank. Still a signal, but its weight swings hard depending on which sub-query the fan-out generated.

These overlap heavily with what works on other engines, though the source pools differ: Google reads its own index, while ChatGPT Search leans on Bing. That is why visibility in one engine tells you almost nothing about the others.

What to do about it

Stop thinking in keywords and start thinking in coverage. If query fan-out is generating a dozen sub-questions around your topic, the site that has a credible answer to eight of them is going to get named more often than the site with one very optimised page. Add clear data, keep the pages current, and make them easy to quote.

Then measure, because none of this is observable from your analytics. AI Mode citations produce few clicks and no referrer you can trust. Request a free AI visibility audit and see which questions surface your brand today, and which ones hand your customers to someone else.