Microsoft Made Human-Only Bing Traffic Official Before Anyone Asked Google

Microsoft confirmed on May 5, 2026 that Bing's freshly announced 1 billion monthly active users are humans, not AI agents. Principal engineering manager Krishna Madhavan replied on LinkedIn to SEO practitioner Mike King, "I can confirm no agents in these numbers - only humans." Google has not made an equivalent public statement about Search MAU.

How Microsoft ended up answering a question Google hasn't

The 1 billion figure landed in Microsoft's Q3 FY2026 earnings call on April 30, when Satya Nadella told analysts "Bing monthly active users reached 1 billion for the first time." It was the kind of round number that gets reported clean and never explained again.

That lasted about a week. Search Engine Journal flagged it first: Matt G. Southern wrote "it's unclear whether Copilot interactions count toward that number," and noted that the gap between 1B MAU and Bing's roughly 5% global search share suggests "a lot of those users are low-frequency or showing up through default settings rather than choosing Bing." Fair caveat. But it didn't ask the harder question.

Mike King did. On LinkedIn, under the official Microsoft post celebrating the milestone, he asked the four-word follow-up: "Does this include agents?" Krishna Madhavan, principal engineering manager at Microsoft, came back with eight words that may matter more than the 1B headline: "I can confirm no agents in these numbers - only humans."

That's a small statement with bigger implications, and worth flagging for what Google hasn't said. There is no equivalent public confirmation that Google's published Search figures, AI Overview reach, or AI Mode usage exclude AI agent traffic. Liz Reid talks about reach. Sundar Pichai talks about Gemini. Nobody at Google has, as far as I can tell, drawn the same line.

Why this stops being a humble-brag and starts being a marketing position

Cloudflare blocked 416 billion AI bot requests since mid-2025, and HUMAN Security's State of AI Traffic 2026 reported agentic AI traffic grew 7,851% year-over-year through that period. The line between "user" and "agent fetching on a user's behalf" is moving fast, and most of the established measurement systems weren't built to tell them apart.

An impression that turns out to be an AI agent fetching for a human elsewhere is still an impression. It is just not the same product the buyer thought they were paying for.

For media buyers, that line is about to become a billing question. If 12% of your Microsoft Ads inventory turns out to be agent-impression rather than human-impression, your CPM math is wrong by roughly the same margin. That isn't hypothetical. Experian launched Agent Trust last month specifically to split AI shoppers into verified humans and reportable waste. The demand side is already moving, which means the supply side is going to have to answer the same question whether they want to or not.

My rough prediction: by the end of Q4 2026, at least one major platform RFP template will include a "human-verified MAU" line item, and Microsoft will be the only one of the big three that can fill it in without a footnote. Probably a 10% to 15% pricing premium attaches to that segment within the same window. I could be off on the exact margin. The direction is hard to argue with.

The supply side just got its first public answer. By stating that the headline 1B is humans only, Microsoft turned a metric definition into a product positioning. "We sell access to a billion humans" is a different pitch than "we sell access to a billion sessions, some unknown share of which are AI agents browsing on someone's behalf." From what I've seen at agencies pricing Microsoft Ads against Google, that distinction is going to come up in Q3 RFPs whether Microsoft puts it on a slide or not.

The footnote nobody's pressing on yet

The "humans, not agents" confirmation only addresses one of the two soft spots in the 1B figure. The other one is still wide open.

Bing's 1 billion humans almost certainly aren't 1 billion people who chose Bing. Statcounter has Bing's global search market share at roughly 4% to 12% depending on device, with desktop pulling closer to 12% and mobile under 1%. Most of Bing's surface area reaches users through default integration: Windows search, Edge's address bar, Copilot's research mode, the search box inside Outlook. Plenty of those touches register as "monthly active" without registering as "I picked Bing this morning."

That isn't unique to Bing. Google has the same problem in reverse, where Chrome and Android route a billion sessions a day through what is technically a chosen default but functionally a tax on attention. The point is, Microsoft answered "are these agents?" but not "did these humans actually search Bing on purpose?" Both questions matter for how an ad buyer should think about Bing inventory.

Ulta's 95% AI-driven sales claim was a similar move: a true number that, on inspection, has been roughly 95% since 2023 and isn't telling you what it sounds like it's telling you. I'm not saying Microsoft's 1B is wrong. I'm saying one verified caveat doesn't close out the rest, and you should still ask the second question.

The follow-up questions worth asking your reps this quarter

For Microsoft Ads buyers, the right move this week is to ask your account manager for a placement-level breakdown of human MAU versus Copilot and agent traffic on the campaigns you're already running. This is the rare moment when Microsoft has publicly committed to the distinction, which means your rep can probably get the data without escalating. Treat it as a one-time information-gathering window. Once Microsoft figures out how to monetize the human-only segment as a tier, the breakdown becomes harder to get.

For SEO leads, Mike King's eight-word LinkedIn comment is now a template. The next time a Google rep cites SERP impression growth, AI Overview reach, or AI Mode adoption in a quarterly update, ask the same thing: does this include agents? If they don't have an answer, that is the answer. Make a note in the deck and move on.

For agencies running both stacks, the pragmatic ask for Q3 reporting is a single column in your client dashboards labeled "human-verified impressions" with a footnote explaining the source. It will be empty for most channels right now. That emptiness is a feature: it surfaces, in client meetings, which platforms have committed to the distinction and which haven't. By Q4, that column is probably the new line of questioning from CFOs auditing media spend.

One messier thought before I close this out

I don't fully believe Microsoft's 1B number will hold up unmodified for long. Either Microsoft refines it down (excluding very-low-frequency Edge default sessions) or up (including some agent traffic as a separate disclosed tier), and either way the figure is going to get a footnote. What sticks is the precedent. A search engine that publishes a humans-only headline figure has implicitly conceded that other figures, theirs or competitors', might not be. That precedent is harder to walk back than the number itself.

Microsoft probably didn't plan to be the company that drew this line. They got asked, and Krishna Madhavan answered honestly. The next move belongs to Google, and the longer the silence runs, the more it sounds like an answer too.

Notice Me Senpai Editorial