Google AI Mode: What 1 Billion Users Means for SEO

Two numbers explain this entire situation, and they point in opposite directions.

At Google I/O on 19 May 2026, the company announced that AI Mode had passed one billion monthly users, with queries more than doubling every quarter since launch. It is now the default search experience rather than an opt-in tab.

Meanwhile, tracking data covering 2025 showed AI Mode referral traffic growing from roughly 1,600 visits to 38.2 million over the year — a remarkable growth rate that still amounts to somewhere around 0.01% of total web traffic.

A billion users. Almost no clicks. That gap is the whole story, and it determines what you should actually do about it.

What Google Announced

The substantive changes from I/O 2026:

• AI Mode became the default, not a tab users opt into
• The search bar was rebuilt around Gemini 3.5 Flash, which also became the default model in AI Mode globally
• The input field now handles longer conversational queries, and accepts images, documents, videos, and Chrome tabs as attachments
• The interface slides between traditional results, AI Overviews, and full AI Mode conversations without leaving the page
• AI Mode is available across more than 200 countries and territories

Google described it as the biggest change to Search in more than 25 years. The framing is defensible: the interface is no longer a search box with links beneath it. It is closer to a conversational assistant with a search engine underneath.

AI Mode Is Not AI Overviews

The distinction matters and is routinely blurred.

AI Overviews are AI-generated summaries that appear above organic results. The results page still exists beneath them. You can still rank, and ranking still produces clicks — fewer of them.

AI Mode is a conversational experience that replaces the results page. There is no list of ten blue links underneath to rank in. The session happens inside the answer.

One tactically important consequence: AI Overviews and AI Mode cite different URLs. Being cited in one does not mean being cited in the other, and they need to be tracked separately.

The Paradox, and What It Actually Means

A billion users generating a rounding error of referral traffic is not a contradiction. It is the design working as intended.

If a user asks a multi-part question, receives a synthesised answer, asks two follow-ups, and leaves satisfied, the session was successful for them and produced no click for anyone. Zero-click is not a failure state in AI Mode. It is the product.

The strategic implication is uncomfortable but clear: AI Mode is not a traffic opportunity to capture. It is a traffic removal event to adapt to.

That reframes the objective. You are not optimising to win clicks from AI Mode, because there are very few available. You are optimising to be the brand named inside an answer that ends the session — because the exposure, the recall, and the resulting branded search are the value you can still capture.

What the CTR Data Shows

Several independent studies now measure the effect, and the figures vary because methodologies differ.

Two caveats worth carrying into any internal discussion.

These are directional, not deterministic. The studies show AI-heavy results pages change click behaviour. They do not demonstrate that every AI Mode impression costs you the same proportion of clicks. Your own Search Console and conversion data is the only reliable evidence for your site.

There is a recovery signal. At least one large-scale study covering millions of queries found organic CTR on AI Overview queries began recovering in early 2026 after a weak end to 2025. The trend is not monotonic, and treating late-2025 as the permanent new baseline may lead you to over-correct.

What Changes When It Becomes the Default

Three structural shifts follow from default status rather than opt-in.

Ranking becomes an input, not an outcome. Strong organic position still matters, because rank functions as an upstream filter on which sources get retrieved and cited. But the ranking itself no longer terminates in a click. It qualifies you for consideration.

Query length and complexity increase. An interface built for multi-part exploratory questions receives multi-part exploratory questions. Content built to answer one narrow query in isolation is a weaker retrieval candidate than content that anticipates the follow-ups.

Impressions and clicks decouple permanently. If Search Console shows impressions rising while clicks and CTR fall, that is not a problem to fix. It is the new relationship between the two metrics, and any team benchmarking against pre-AI baselines will misread its own performance.

Measure It Properly

One genuinely useful development: in June 2026 Search Console began rolling out dedicated Generative AI performance reports, showing impressions from AI features including AI Overviews and AI Mode, initially for a subset of sites.

Check whether your property has access. If it does, that is the closest thing to first-party measurement available for these surfaces, and it is considerably more reliable than inferring impact from aggregate traffic changes.

Alongside it, track:

• Citation rate across a fixed set of buyer questions, run repeatedly, since AI answers are non-deterministic and single checks are noise
• AI Overviews and AI Mode separately, given they cite different URLs
• Branded search volume, which is where AI Mode exposure converts into something measurable
• Conversion rate of the traffic you do receive, which multiple studies suggest arrives with higher intent

What to Actually Do

The tactical answer has not changed much, which is itself informative. Google’s own May 2026 guidance holds that answer engine optimisation is not a separate discipline from SEO — strong fundamentals plus content only your team could produce.

In practice:

• Keep ranking. Position is an upstream filter for citation.
• Answer the follow-up questions, not just the initial one. Exploratory sessions reward content that covers the whole enquiry.
• Structure for extraction — question-style headings, direct answers in the first sentences, self-contained passages.
• Publish something only you have. Original data and first-hand experience are what make a source worth citing when nine other pages say the same thing.
• Give people a reason to click anyway — tools, calculators, proprietary data, things an answer cannot contain.

What Not to Do

• Do not build thin “AI SEO” pages. Volume was already a losing strategy and this accelerates it.
• Do not abandon SEO. Ranking is now a qualifying condition for citation rather than a redundant one.
• Do not benchmark against 2024 traffic. You will misread the results and misallocate budget.
• Do not treat a single favourable answer as a result. Responses vary by user, location, and moment.
• Do not assume AI Overview citation implies AI Mode citation. Different surfaces, different sources.

Final Thoughts

One billion users is a genuine milestone, and the reason it matters for SEO is not that a billion people are about to visit your website through it.

It is that a billion people now have a way to resolve a question without visiting anyone’s website. The work shifts from earning the click to being the source named in the answer — and, where possible, giving people something the answer cannot give them.

Check whether your Search Console has the new AI reports, establish your own baseline, and stop measuring against a version of search that no longer exists.