The head of Google, Sundar Pichai, stated that he expects there to be two methods in how AI agents access website content; the first is through a method similar to how humans browse the Internet and the second through protocols specifically created for use by agents. Pichai believes both browsing and using protocols are going to grow, however he also feels that protocols will ultimately scale much better than browser-based methods but take longer to develop.
The above comments were made by Pichai during an interview he gave to Alex Heath of the Sources podcast. Google has been backing a protocol to enable agents to shop online called the Universal Commerce Protocol (UCP). Additionally, Google lists Google-Agent as the user agent it uses when hosted agents browse the web.
After making the above statements, Pichai discussed how he expects agents to allow users to have one unified experience across Search and Gemini. Roger Montti covered this portion of their conversation.
How Pichai Sees Agents Reaching Websites
Alex Heath wondered whether the web needs to evolve for AI agents that use browsers much like people do to get things done, noting that the current setup can feel “like trying to fit a triangle into a square.”
Pichai called it “a complex moment” and said:
“When we say agents working at scale, that means many, many different parts of the economy interoperating with each other in some way, in a way that has to make sense for all the players involved.”
He added that he thinks the technology will get ahead of the ecosystem in how those parts come to work together.
Pichai pointed to “a lot of momentum behind MCP or agent protocols or payment protocols,” citing UCP. He said “there’s going to be a chaotic evolution there,” hopefully “working towards more interoperable standards.”
He described browser-based computer use as “a very elegant way to work too, because that’s what you as a user would do,” adding that it allows controls that keep users in the loop.
“I think you have both ways and they’re going to develop,” Pichai said.
He called better protocols “a far more scalable approach” that would work anywhere and instantly, but said he expects them to take longer to develop.
What Pichai Said When Asked About Ads
Heath also asked Pichai how Google protects its own interfaces when AI agents do the browsing. He noted that Google has “a very large ads business that requires engagement” and cited Amazon blocking Muse.
Amazon cut off Meta’s Muse from shopping on Amazon.com in September, saying the agent doesn’t identify itself when it browses. Muse is a personal agent that can open a browser and fill out forms.
Pichai answered:
“The right answer longer term is you put users first and say what would users want.”
He added that ultimately it has to work for all the businesses providing value. He said the market will reach an equilibrium and “that equilibrium will shift.”
He described this period as being similar to the Internet’s infancy; that is when individuals believed buying and selling homes would change almost instantly. Today most of these same methods have remained the same for approximately 20 years, he added, citing “real constraints in the real world.” He stated further he believes “micro equilibriums” will emerge from AI agents, which he expects to be “even more disruptive.”
He described putting himself in users’ shoes, adding, “That’s how I would approach our product development.” His answer didn’t address how Google’s ads would work when an agent does the browsing.
Google’s Protocol And Browsing Agent
UCP, one of the protocols Pichai named, is an open standard for agent commerce that Google built with a group of companies. Google’s UCP page says merchants can adopt it to enable agentic actions in AI Mode and Gemini, starting with direct buying.
When I covered UCP in September, Google had added a Merchant Center option for merchants to turn on cart transfer, which sends a shopper’s items from Google to the merchant’s site.
On the browsing side of things, Google includes Google-Agent as one of the user triggered fetchers in Google’s list. According to Google, agents that are located on Google’s servers use this method “to navigate the web and perform actions upon user request.” Since an end-user requested the fetch, according to Google, these fetchers usually do not follow robots.txt directives. Also, Google notes that user-agent strings may be spoofed, and it publishes the IP address range(s) these fetchers operate out of.
Why This Matters
Pichai expects that developing protocols will be more time-consuming, therefore, the agents that visit pages similar to how an individual would, are going to remain visible while common standards are established, such as UCP. The agents running from Google’s server access the web with a Google-Agent user agent. This user-agent is classified by Google as a user triggered fetcher, therefore a robots.txt blockage generally won’t prevent these agents.
Looking Ahead
Pichai said Google plans to make Gemini’s agent features, currently available mainly to Ultra and Pro subscribers, more widely accessible over the next few months. Google’s UCP overview page also highlights lodging, with booking in AI Mode, and food features marked as coming soon, with a waitlist for each.
Featured Image: Tetiana Yurchenko/Shutterstock
Source: Google.
