Where do US buyers actually meet AI answers in 2026? On separate answer surfaces, not one results page. Google runs AI Overviews and AI Mode as separate surfaces with a Search Console report of their own. Bing publishes citation telemetry that covers Copilot and its AI-generated summaries. Each surface reports different numbers in different units, so each one has to be read on its own rail, and measured there too. The map below is built entirely from official platform docs.
Here is the share backdrop. June 2026 StatCounter puts Google at 86.67% of US search. Google reported AI Mode passing 1 billion monthly users in its first year. Bing's index reaches further than its own bar. It supplies Yahoo Search results under a long-standing partnership and underpins Microsoft Copilot. On the measurement side, Compass reads this map with 20 buyer questions and Panorama with 50, each surface kept in its own unit.
Does one SERP still exist? It doesn't, and here is what replaced it
Google Search Console now ships a dedicated generative AI performance report. Per Google's own docs, it covers exposure of your links in AI Overviews and AI Mode and leaves Search Labs experiments out. It breaks the data down by page, country, date, and device. Official primary support.google.com
The measurement rules are equally specific. When a user clicks an external link inside AI Mode, Google counts it as a click; when a user asks a follow-up question inside AI Mode, Google treats it as a new query. Official primary support.google.com That second rule matters. A single buying conversation can generate a chain of queries, and your brand can enter or exit that conversation at any link in the chain.
The report is a Search Console UI surface. As of July 14, 2026, Google's public developer docs offered no read API of its own for it. That is what makes steady, repeated observation of these surfaces its own operational job rather than a dashboard checkbox.
Scale check (added July 17, 2026). This is no longer a test surface. At I/O 2026 Google reported that AI Mode passed 1 billion monthly users in its first year, with query volume more than doubling each quarter. That is a figure Google reported about itself, attributed to its own announcement. Official primary blog.google
Where does US search demand sit, and why does Bing belong on the map?
June 2026 StatCounter share puts Google at 86.67%, Bing at 8.73%, Yahoo Search at 2.55%, and DuckDuckGo at 1.53%. Market data gs.statcounter.com
Read the small bar carefully. Bing's share is a fraction of Google's, but Bing is the platform on this map that ships explicit citation telemetry. In February 2026, Microsoft announced AI Performance in Bing Webmaster Tools as a public preview, covering citation activity across Copilot, Bing's AI-generated summaries, and select partner integrations. Official primary blogs.bing.com Of the two official reports on this map, it is the one that gives you platform-level evidence of your pages being cited by an AI assistant — whatever the share chart says.
Bing is not one surface either. Its index feeds Yahoo Search results under a long-standing deal and sits under Microsoft Copilot. By several industry analyses it still influences which sources ChatGPT search surfaces, so the small bar carries more answer real estate than it looks. Market data searchengineland.com
What does each surface actually report?
Per-surface units are the part a blended score erases. The two US reporting surfaces do not measure the same event, and neither claims to.
Google's report speaks in impressions. It tells you your links were exposed inside AI Overviews and AI Mode, sliced by page, country, date, and device. It does not tell you which prompts triggered the answers, what the answer text said about you, or which exact URLs were cited in what role. Official primary support.google.com
Bing's report speaks in citations. Total Citations, Average Cited Pages per Day, citation trend, sampled grounding query phrases, and per-URL citation counts, reported for your verified site. Microsoft's announcement is also explicit about what the numbers do not mean: citation counts do not indicate position, rank, authority, or the role your page played inside the answer. As of July 14, 2026, this preview is a UI surface with no public read API found. Official primary blogs.bing.com How do you stand these two reports up as a working stack? What does each one prove, and what does neither of them see? That is the subject of our guide to the official AI-visibility rails.
The decision this map changes. "How visible are we in AI search?" is not one question. It is at least three. Are your links being shown where Google assembles answers? Are your pages being cited where Bing grounds Copilot, and what do the answers actually say when a buyer asks about your category? Each question has a different official data source, a different unit, and a different competitive move attached to it. A team that sees the three numbers separately knows which surface to fight for first. A team holding one blended score knows only that something, somewhere, changed.
How does Google assemble an AI answer? From its own guide
Google's AI features optimization guide, updated July 10, 2026, says how the machinery works. Generative responses are grounded in the Search index using retrieval-augmented generation, and complex questions trigger query fan-out, which means related sub-questions issued in parallel. Eligibility is stated just as plainly: content must be indexed and eligible for snippets to be a candidate, and no special AI-specific schema markup is required. Official primary developers.google.com
Two consequences follow directly. First, you are no longer competing on a keyword. You are competing across a family of related questions the system builds around a buyer's intent, most of which never appear in your keyword tools. Second, there is no schema shortcut to buy your way into these answers; the platforms select sources from the same index your competitors live in. Which questions your brand wins, loses or is absent from is an empirical fact about your market, and it is measurable. What that shift does to demand measurement, meaning question families instead of keyword rows, is worked through in our query fan-out analysis.
Do you need special schema to appear in AI Overviews?
No. Google's own guide sets the entry conditions in plain terms: content must be indexed and eligible for snippets to be a candidate, and no special AI-specific schema markup is required. There is no markup shortcut into these answers. The entry conditions above are the ticket, and the contest after entry happens inside the same index your competitors live in.
| Surface | Official report | Unit it reports | What it leaves out |
|---|---|---|---|
| Google AI Overviews | Search Console generative AI report | Impressions and clicks by page | The prompt, the answer text, the cited URL |
| Google AI Mode | Same report | Same units, follow-ups counted as new queries | What the answer said about you |
| Microsoft Copilot | Bing Webmaster AI Performance | Citation counts per URL, sampled query phrases | Clicks, position, the full answer |
| Bing AI summaries | Same preview | Same units | Same gaps |
| Assistants with no publisher rail | None published | Nothing | Everything, unless you observe it yourself |
Rows restate the official docs linked in the sections above. Send this comparison table to whoever still budgets for one results page.
How do you measure the whole map, not a mood?
CiteAngle measures US AI search the way this map is drawn: per surface, per question, with each result kept in its own unit. Compass takes 20 of your buyer questions to every surface in the table above and asks each one seven separate times, because these answers move between one day and the next. Nothing is sampled and every reading keeps its status, so what comes back is where your brand actually stands, not where it stood once. Panorama does the same on 50 questions. It carries the same evidence rules across the wider engine set and turns it into a prioritized plan: which questions to take back first, on which surface, with which exact URLs. Official platform reports and repeated screen-level observation appear as separate evidence rails in the deliverable, so every number keeps its source.
What that split looks like in practice: our own run over this map, measured July 27, 2026, put 2,675 citations on Google News answers alone, 30.9% of everything the run collected, drawn from 635 separate domains. The same run left 14 source links on the Claude surface across its 50 answers. Same questions, same day, two surfaces that report almost nothing in common — which is the part a single blended visibility score averages away. The per-surface breakdown of that run is published in full.
What a one-screen plan costs. If your plan budgets for a single results page, you are invisible on most of the surfaces below and nothing in your reporting will say so. We read them on one sealed protocol with competitors side by side with your brand, which is why a prescription from us lands somewhere instead of stopping at the surface it was measured on.
The second cost is organisational. Without a map, each team defends the surface it already reports on, so budget follows the loudest dashboard while buyers move somewhere else, and the gap widens quietly for a year before anyone can point at it.
Why read the map with us
A map like this is uncomfortable because it turns a single line item into many. It is also the cheaper discovery: a plan that budgets for one results page is not slightly wrong, it is blind on most of the screens where the shortlist now gets written, and the blindness is invisible from the inside. Every report still renders. Nothing looks broken.
Reading all of it on one yardstick is the part that is hard to assemble yourself. We put the same questions to 15 engines and count the results on 17 surfaces, holding search results and AI answers in the same grid, so a prescription has somewhere to land instead of stopping at the surface it was measured on. Each surface keeps its own denominator, because pooling them would hide the exact unevenness this article is about. And the roster is published surface by surface, so you can hold it against anybody else's before you buy either.
Know where your brand sits on every surface on this map
The 15-surface grid puts a measured status on every cell, and a free AI visibility check shows the format on your own brand before any money moves.