Google I/O produced a week of coverage of how AI will change the search experience. Most of it focuses on consumer factors, with little attention paid to the emerging pattern of businesses.
Most high-profile consumer demos at I/O move the user from search to action, with Google handling most of the journey in between. While this was a recurring theme throughout I/O, the infrastructure behind those demos was being released months before the keynote.
Last week’s deep dive asserted that the real risk of I/O was more economic than technical. This piece looks at where that economic risk lies and why the business playbook has yet to match the consumer experience that Google showed on stage.
Displayed by Google
Google I/O demos include Global Cart, local service agent booking, and information agents that monitor listings or products in the background.
Universal Cart allows you to add products to a single cart that persists across all Google sites. Agent bookings bring prices and availability together and provide links to complete bookings, bringing the journey closer to completion.
Not all demos were selling, however. Google also demonstrated coding, dashboards, simulations, and research tools.
The infrastructure was already in place
I/O makes the infrastructure visible to consumers, but has been developed for a long time.
In late 2025, Google released a checkout agent, which allows Google’s AI to add items to a merchant’s cart and complete the purchase.
This year, Google launched the Universal Commerce Protocol, an open standard for agent commerce. UCP provides agents and merchant systems with a common language, rather than requiring a separate interface for each agent.
In April, Sundar Photosi told Stripe CEO Patrick Collison that the search would be “agent manager.” SEJ has been tracking this change by leveraging Google’s search patents and job-based search features since the beginning of the year.
Jay Jaffin, CMO and Strategic Advisor at Visor Strategic Advisors, summarized the concerns of businesses:
“Global Cart doesn’t just keep a colon at the bottom of the funnel. It collects everything, from the first search query to the last exit, without your customer ever landing on your site. The window to adapt to this could be much shorter than a decade.”
These Demos are User Built
After watching the I/O demos, it was clear that these features are aimed at a specific type of user. This user does not open ten tabs and compare options manually. They explain what they want and let the AI do the rest.
When they ask real estate agents to monitor real estate listings or track sneaker drops, they’re not looking for it the usual way. They submit research work and wait for notification.
That means businesses are competing for something different. Haroon Qureshi, Global Retail Experience & Partnerships Lead at WPP Media, explains how the goals have changed, saying:
“In the future, are brands competing for clicks? Or are they competing for endorsements?”
At I/O, Google said AI Mode has surpassed one billion users per month, with queries more than doubling every quarter since launch. That gives this new way to search several access points that are the same.
Why This Matters to Search Professionals
Ecommerce
Google ensures that, with UCP, your product remains the best seller of record. Shoppers can check out with Google Pay or transfer items directly to your business website.
However, marketers are beginning to differentiate between owning a purchase and owning the data that led to it.
Armando Roggio, Senior Contributor at Practical Ecommerce, put it bluntly:
“In Google’s model, marketers still own the activity, but not the purchase intent or product discovery.”
That makes the optimization problem difficult to solve without data from Google about how much weight different signals have in the mediator flow.
Aleyda Solís, SEO Consultant and Founder of Orainti, commented on LinkedIn that “ecommerce SEO and AI search optimization cannot be reduced to ‘content in products.’”
His posts featured key features, such as accurate feeds, consistent attributes, clear pricing, and detailed content that give agents something to reference.
Local and Service Businesses
For local businesses, Search aggregates prices and availability with direct links to complete the booking with the provider of your choice. In select categories such as home improvement, beauty, and pet care, users can ask Google to call businesses for them.
If the call goes to voicemail or the staff cannot provide clear answers, the business can lose time before the user visits the site.
In a sense, agency booking turns readiness into visibility. Karim Al Chamaa, Founder of Implemnt, explained the dynamic on his company’s blog, saying:
“If a Google agent is the one making the call, the default is an automatic cancellation.”
Average
If the information agent monitors real estate listings for a week and returns a recommendation, the value is extracted without the traditional click method.
Jake Ward, Co-Founder of Spoken, wrote on X that “we are moving forward into a world of impressions > clicks.” You can track organic sessions and referral clicks, but you can’t track how many times your business’s products were considered and rejected by an agent, or how often your business was recommended in an agent’s booking.
The metrics that have defined search performance for years may not describe this agent’s mediation journey as clearly.
What’s Not Known For Now
Google did not share the selection criteria for Universal Cart recommendations or the agent’s booking results. Marketers are currently developing strategies based on intuition rather than formal guidance. Until Google clarifies the signals its agents rely on to make comparisons and choices, the optimization process is a matter of educated guesswork.
Currently, there are no third-party measurement tools that track agent-initiated activities or how often recommendations are made as separate metrics from organic traffic.
While Merchant Center now provides AI-driven insights that compare voice shares to similar products, businesses can’t say “the agent never considered us” or “the agent considered us and rejected us.”
The connection between paid ads and organic visibility in AI-driven commerce is not fully explained either. Google says it’s “not a retailer” and “not a marketplace,” but Universal Cart combines products from a variety of vendors and offers AI annotations that suggest alternatives. How advertising fits into this is a question Google has yet to answer.
Looking Forward
Google makes it quick for consumers to move from search to action, but at the same time, it makes it challenging for businesses to see and measure their visibility. The shopping experience shared at I/O was shown from the consumer side, giving a few insights that can help businesses emerge within it.
The feedback loop becomes difficult to track. If the buyer leaves the purchase decision to the agent, businesses that are not selected may not know they were part of the process.
Additional resources:
Featured Image: Roman Samborsky/Shutterstock
