Microsoft logo on the keynote stage at Ignite 2026.

Microsoft says its Bing search engine now has one billion monthly users. The company says those are humans — not AI agents — looking up information on the web. What most people may not realize, though, is that Microsoft also uses the same Bing index to serve agents through Web IQ.

Web IQ is Microsoft’s agentic search service, competing with upstarts like Exa, Firecrawl, and Tavily. DeepSeek opened its own search API to developers in October.

At Microsoft’s Ignite conference this week in San Francisco, The New Stack sat down with Jordi Ribas, Microsoft’s president of Search & AI. Ribas joined the search team in 2008, back when Bing was still called Live Search. Today, his team runs both Bing and Web IQ, which Microsoft introduced at Build in June and opened broadly to Azure customers on Tuesday.

How agents search differently

“Bing is a search engine for humans,” Ribas says. “This is really [a] search engine where the queries come from the AI, and so you do need to design it differently because agents search differently.”

“Bing is a search engine for humans.”

It’s not just that human and agent search queries look different (and Ribas noted that agent searches often look a bit “robotic”), but that they also have different expectations for the results.

On Bing, Ribas says, the order of the results matters because people have to decide which link to click. “An agent doesn’t really need that,” he says, “because an agent will read everything.”

At their core, Bing and Web IQ share the part of the search engine that finds the relevant documents, Ribas says. For agents, Web IQ picks out the passages it deems most relevant and sends them back.

For example, an agent comparing corporate earnings results might need only one row in a table on page 12. Sending the whole report would give the agent more context than it needs, and because the model would have to wade through far more text, the search would become far more expensive.

As Ribas puts it, “the agent will spend money basically by reading all the documents.” “So it’s very important that you extract exactly what you need, and no more.”

Microsoft builds that shared index, always follows the instructions in a site’s robots.txt file, and identifies itself as a crawler, Ribas stressed multiple times. “Not everybody does that in the space.”

How much of the search should the agent do?

Typically, the agent’s orchestrator decides when to call Web IQ as a tool and what queries to send, Ribas notes. “Sometimes the orchestrator is not that smart about knowing what queries to do,” he says, “and it may do multiple iterations.”

At Ignite, Microsoft is addressing this by making Auto the default search mode in Web IQ, which Ribas describes as “kind of like a mini orchestrator that happens within Web IQ.”

With this change, an agent sends Web IQ only one question. Microsoft then tries a few variations of the query in parallel and sends back passages with citations.

Auto Advanced takes this a step further and checks whether the first round turned up enough information. If it finds a gap, it revises the query and searches again. “And we do have a model that says, ‘OK, is that evidence enough to provide a good chat response?'” Ribas says. “And if it’s yes, great, we’re done.”

Unlike some Web IQ competitors that try to create full answers and research reports themselves, Microsoft leaves the agents and models that power them to create the actual answers.

“You can use pretty much any model on top of it,” Ribas says, though he adds that Microsoft is thinking about packaging Web IQ with a model at some point.

What Web IQ costs

On Web IQ, one thousand standard web searches cost $7, compared with $8 for Auto and $9 for Auto Advanced. (The $1 starting price Microsoft advertises is for Browse, where you already have the URL.) If you know exactly what you’re looking for, the standard search may be all you need. But if your agent routinely reads the results, rewrites the query and tries again, a single Auto call could cost less overall.

“If you want to be fast but you want higher quality, maybe Auto is the way to go.”

“If all you care is about being super fast, then you would just use maybe the primitives directly,” Ribas says, referring to the individual search APIs. “If you want to be fast but you want higher quality, maybe Auto is the way to go.”

A crowded agent search market

“People are predicting that in a couple of years there will be 1,000 times more queries done by agents than by humans,” Ribas says. “That’s why there are so many competitors in [the] space.”

Some of those competitors optimize for quality, Ribas says, but “bring a lot of tokens to come up with a comprehensive answer,” which drives up both cost and latency. Others, he says, are very fast, but the quality isn’t there.

“We are able to optimize across three dimensions: quality, latency, and token efficiency,” he says, “and that’s really hard to do.”

Genspark co-founder Ray Zhong says in Microsoft’s announcement that switching to Web IQ cut its voice assistant’s response time from six seconds to 0.6 seconds. “Six seconds in a voice turn is audible dead air.”

The post Microsoft says Bing serves 1B people. Its next search engine isn’t for them. appeared first on The New Stack.