AI engines

Google AI Overviews: how they work and how brands get cited

AI Overviews are the Gemini-written summaries that now sit above the organic links on a large share of Google searches. They are the biggest AI surface in existence by volume, they already suppress clicks on informational queries, and the rules for getting cited in one are not the rules you learned for ranking.

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

AIGenerated answer · AI engine monitored example

« Google AI Overviews ? »

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 an AI Overview is

An AI Overview answers the question inside the results page itself, with links to a handful of sources it drew from. Mechanically it is retrieval-augmented generation: Google pulls relevant pages, grounds a generated answer in them, and attributes what it used.

The user can read the whole answer and never click. That is the zero-click problem, and it is why the goal has shifted from "rank and get the click" to "be the source that gets named".

One index, and it is Googlebot's

The most common misconception first: there is no separate AI index. AI Overviews are built from the same index Googlebot maintains for regular search. A page has to be crawlable and indexable before anything else is worth discussing.

This is a structural difference from ChatGPT Search, which leans mostly on Bing. Same broad logic, different plumbing, and it explains why a brand can be everywhere in one engine and invisible in another.

How much your organic ranking still matters

The link between organic position and AI Overview citation is real: well-ranked pages get cited more often. But the correlation has been weakening as Google refines how it selects sources for generation.

In practice that cuts both ways. Ranking first helps and guarantees nothing. A page sitting on the second results page can be cited if it answers the intent better than the ones above it. The tie to ranking is still stronger here than it is in AI Mode, where query fan-out breaks the link far more aggressively, and stronger than in Perplexity or ChatGPT.

The signals that move the needle

E-E-A-T. Experience, expertise, authoritativeness and trust. The credibility signals Google already used, now doing double duty as source-selection criteria.
Freshness. Recently updated content wins, especially on topics that change. A page dated three years ago is a weak candidate whatever it says.
Extractable structure. Lists, tables, FAQ blocks. These formats let the model lift one precise passage instead of guessing at your meaning.
Entity density. Concepts, brands and products named explicitly and linked to each other, rather than described in the abstract.
YouTube presence. A strong observed correlation between a brand's YouTube footprint and its citation rate in AI Overviews. It fits the broader pattern that off-site mentions often outweigh what is on your own site.

What it does to your click-through rate

When the Overview answers the question outright, a share of users stop clicking anything. Organic CTR drops mechanically, and hardest on simple informational queries. Gartner has forecast a 25 percent fall in traditional search volume by 2026 as assistants absorb the demand.

The response is not to fight it. It is to change what you count. A citation inside the answer is now worth measuring in its own right, and the brands that adapted early are tracking share of voice per engine rather than sessions.

The Google-Extended myth, killed properly

A persistent belief holds that blocking the Google-Extended crawler in robots.txt removes you from AI Overviews. It does not.

Google-Extended governs one thing only: whether your content can be used to train Gemini models. AI Overviews are generated from the Googlebot index. Blocking Google-Extended changes nothing about your presence in Overviews. It only opts you out of model training, which may or may not be what you wanted, but it is not a visibility control.

The same goes for llms.txt, which is a proposed convention rather than an adopted standard, and which Google ignores. Treat anyone selling it as a requirement with suspicion.

How to actually get in

No single signal is enough. What works is the combination: aim for solid organic positioning on the query, structure the content so a passage can be lifted cleanly, keep it current, name your entities precisely, and build presence on the platforms Google cites willingly, YouTube first among them. The detailed playbook is here.

Before optimising anything, find out where you stand. AIVIIU measures your visibility across the major engines, shows which prompts surface your brand, your share of voice and the sentiment attached to it. Request a free audit and see what Google is telling your customers right now.