Google publishes first AI Mode usage data for US
Google's first AI Mode usage snapshot signals a distinct citation surface. Optimising for AI Overviews no longer covers your visibility.
Key takeaways
- Google has released its first US usage data for AI Mode, separate from AI Overviews.
- AI Mode and AI Overviews are diverging surfaces with different citation logic and source sets.
- Gated PDFs and definitional explainers are now visibility liabilities in AI Mode.
- Structured, retrievable, comparative content wins citations on the prompts buyers actually use.
- Vendors control the narrative on AI search usage while keeping citation mechanics opaque.
What happened
Per the Google AI blog, Google has published its first look at how Americans are actually using AI Mode in Search. The post frames AI Mode as a behavioural shift, not a feature launch, and uses the United States as the testbed because it is where the product has had the most runway since rollout earlier in 2025.
Google's framing is deliberate. The company is signalling to publishers, advertisers and enterprise SEO teams that AI Mode queries are now substantial enough to merit their own usage taxonomy, separate from classic Search and separate from AI Overviews. The illustrated taxonomy in the announcement (planning, shopping, learning, troubleshooting, creative tasks) tells you what Google believes AI Mode is winning: longer, more exploratory questions that classic ten blue links handled badly.
What Google is not yet publishing: query-level citation share, source diversity data, or how often AI Mode answers without sending a click. That absence is itself the story.
Why it matters for your brand
The first thing to register is that Google is now actively marketing AI Mode as a separate surface. For any B2B brand that has spent the last 18 months optimising for AI Overviews, that is a meaningful split. Overviews sit on top of a results page. AI Mode replaces it. The citation logic, the source set, and the prompts that trigger each are diverging, and your visibility on one does not guarantee visibility on the other.
For financial services brands, this matters because AI Mode skews toward the "help me decide" queries that used to fund the middle of the funnel. A user comparing custody providers, ESG ratings methodologies, or trade finance options no longer scrolls. They ask. If your thought leadership is not in the model's retrieval set for those prompts, you are not in the consideration set. The implication for content strategy: stop publishing 1,500-word explainers that rehearse definitions, and start publishing structured comparative content that answers the second and third question in a research chain, not the first.
For multilaterals and UN-system bodies, the shift is sharper. AI Mode is where policy researchers, journalists and member-state staffers will increasingly land when they want a synthesised view of, say, disaster risk financing or financial inclusion benchmarks. Google's own framing suggests AI Mode favours authoritative, well-structured sources. That is good news for institutions like UNDRR or CGAP if (and only if) their data and frameworks are published in formats the model can ingest cleanly. PDF-only reports are now a visibility liability. HTML landing pages with clear data tables and named methodologies will be cited; locked PDFs will not.
For major industrial groups, the relevant scenario is procurement-adjacent search. Specifiers, engineers and sustainability leads at customer organisations are using AI Mode to scope suppliers before they ever reach a sales conversation. If your technical documentation, EPDs, and sustainability disclosures are not retrievable in formats the model trusts, a competitor's are. The brand-building consequence is that distribution now matters more than production. A whitepaper that lives behind a gated form is functionally invisible to AI Mode.
For philanthropic and policy institutions, the angle is narrative control. AI Mode synthesises. If your foundation has spent a decade shaping the language around, say, climate adaptation finance or pandemic preparedness, the model is now the editor deciding which framing wins. Institutions that publish their definitions, taxonomies and position papers in machine-readable, well-linked form will see their language reproduced. Those that do not will watch their categories be renamed by the model using whichever competing source had cleaner markup.
The signal in context
Google publishing usage data for AI Mode is the latest move in a pattern: AI search vendors releasing curated glimpses of how their products are used, without releasing the underlying citation logs. OpenAI has done the same with ChatGPT search, and Perplexity has done it with its Discover feed. The vendors are setting the narrative about what their AI surfaces are good for, while keeping the citation mechanics opaque. That asymmetry favours the platforms and disadvantages anyone trying to measure share of voice rigorously.
The harder context is that AI Mode is moving from novelty to default in the US faster than most enterprise comms teams have adjusted for. If your 2026 content plan still treats AI search as an experimental channel, this announcement is the signal to reclassify it. The brands that will be cited in AI Mode answers a year from now are the ones publishing structured, attributable, retrievable content today, on the specific questions their buyers actually ask. The rest will keep ranking on classic Search for queries that fewer people are typing.