Last Wednesday Google announced two things on the same day. AI Overviews in the US now run on Gemini 2.0 and will appear for more complex queries, including for people who aren’t signed in. And there’s a new experiment in Search Labs called AI Mode.
The first is an incremental change to something we’ve lived with since last May. The second is the one I’d pay attention to.
What AI Mode is
It’s a separate mode of Search that answers your question as a long, conversational response with links, rather than a page of results with an AI summary perched on top. You can ask follow-ups. It’s built on a custom version of Gemini 2.0, and for now it’s opt-in, US-only and limited to Google One AI Premium subscribers.
So I can’t try it from here in the UK, and nor can almost anyone reading this. Everything below is based on Google’s own description and the screenshots doing the rounds.
The phrase in Google’s announcement that caught my eye is “query fan-out”. Instead of running your one query, AI Mode breaks it into a set of related searches, runs them in parallel across the web, the Knowledge Graph and shopping data, then pulls the findings into one answer.
Think about what that means for a moment. For twenty-odd years SEO has been built on a simple model: someone types a query, Google ranks pages for that query, you try to rank for the query. Fan-out puts a layer in between. The user asks one thing, Google asks itself ten things, and your page might be retrieved for one of those sub-queries that the user never typed and you’ll never see in any report.
Why it matters even though you can’t use it
Labs experiments don’t always graduate. Plenty of things Google tests quietly vanish. But AI Overviews followed exactly this path; Labs in 2023 as the Search Generative Experience, then rolled out to everyone in the US, then other markets. I’d assume AI Mode is on a similar track unless something goes badly wrong.
The commercial question is obvious. AI Overviews already sit above the organic results and answer a chunk of informational queries before anyone clicks. AI Mode goes further: there’s no list of ten blue links underneath to fall back on. If it becomes a default, or even a popular tab, informational content that relies on search traffic has a harder time.
Google’s line is that AI features drive “higher quality” clicks and help people discover a wider range of sites. Maybe. I haven’t seen anything from outside Google that measures that properly yet, and until I do I’d treat it as a claim rather than a finding.
What I’d do now
Not a lot, and that’s deliberate. There’s no point rewriting a content strategy around a US-only beta for paying subscribers. But a few low-cost things make sense:
- Watch the US. If you have US traffic, segment it in Search Console and keep an eye on informational query click-through over the next few months.
- Make your content easy to extract. Clear headings, direct answers near the top, proper structured data where it fits. That was good practice before any of this, and it’s the sort of page a system doing fan-out can pull a clean answer from.
- Know which pages earn money. Split your organic traffic into pages that drive sales or leads and pages that drive only visits. The second group is the one most exposed if answers move onto Google’s page. If that group is most of your traffic, you want to know now rather than later.
- Stop reporting rankings as the headline. If the results page keeps changing shape, a ranking position tells you less and less about traffic. Clicks, conversions and revenue from organic hold up better as a measure.
I’ll be honest that I find this one quite interesting as well as slightly worrying. Fan-out is a genuinely sensible way to answer a complicated question. It’s also a very effective way to answer it without sending anyone anywhere. How that balances out for site owners probably depends on decisions Google hasn’t made yet.
I’d like to actually use it before forming a firmer view. When it reaches the UK, I’ll come back to this.