Google has changed the model that writes the AI answers in its search engine. The new engine breaks each question down into many parallel searches and links to more websites than before. For leadership, the takeaway is clear: visibility is no longer won only on the main keyword, but on each of the sub-questions a customer asks before buying. Companies with clear answers to those sub-questions now have a better chance of showing up.
What happened
On 12 November, Google announced that Gemini 4 Hydrogen is now the default model in AI Mode, the conversational mode of Search, and that it is also used in AI Overviews (the summaries shown above the results) for complex questions. It is available to all users worldwide and in every language AI Mode supports. Google’s head of Search product called it “the biggest model upgrade to Search yet”.
Google highlights four changes:
- Better understanding of intent: multi-part questions, very specific requests and even queries with typos.
- More searches in parallel: the query fan-out technique runs simultaneous queries across every relevant subtopic of the question and, according to Google, links to “a broader range” of websites than before.
- More visual answers: more images, structured responses, comparison tables and custom tools built for each query.
- More nuance in every supported language, Spanish included.
Google has not shared any figures on traffic impact. The change comes a few days after another version of the model was launched for paying AI Mode subscribers, and in the middle of the November core update, during which Google had already explained that it is bringing its classic search signals into its AI experiences.
The bigger picture
Until now, AI in Search tended to rely on a handful of large, well-known sources. With more parallel searches per question, each answer is built from more pieces, and each piece can come from a different website. That opens a door for mid-sized companies that do not win the generic search but are the ones that best explain a specific detail: a delivery time, a technical compatibility, the difference between two materials or the real maintenance cost of a piece of equipment.
The flip side is that the comparison now happens inside Search. If the AI can build a table with your products and your competitors’, customers reach your website with their decision half made. Or they don’t reach it at all.
Why it matters to your business
- Sales: in considered purchases (B2B, home, equipment, mid-range fashion) customers ask long questions. Every sub-question is a chance for you to appear, and a chance for someone else to.
- Product data: comparison tables are fed by what your website and product pages state clearly and in a structured way. If specifications, sizes or terms sit in a PDF or an image, the AI will compare using whatever it finds elsewhere.
- Measurement: organic traffic may move over the coming weeks for two reasons at once, the core update and the new model. Without separating the two, it is easy to make the wrong calls on budget or content.
- Timing: it lands just before Black Friday and the Christmas season, when comparison searches peak.
What to do now
The questions we would ask in a leadership meeting are simple: do we know which questions our customers ask before choosing us? Does our website answer them better than our competitors’ do, with facts rather than slogans? Can our reports tell how much of the traffic change comes from the core update and how much from AI answers?
As an example: an office furniture manufacturer with only a generic “ergonomic chairs” page is competing for a highly contested search. If it also publishes which chair suits which number of hours of use, height or type of workstation, with its own comparison table, it has many more sub-questions where it can be the source that gets cited.
How we approach it at TRILOGI
Here we do not start with the technology, but with the customer. We sit down with sales and customer service and list the real questions that come up before a purchase: in emails, on calls and in sales objections. With that map we build a content plan by sub-question, prioritised by the margin of the product or service behind it, and we turn information that currently lives in catalogues or PDFs into clear tables and answers on the website. That is the work behind our content strategy service: fewer generic pieces, more answers that help people decide. Then we track which questions start citing the brand and adjust.
Every Tuesday, the 5 news stories that matter: subscribe to the newsletter




