Google’s Next Big Bet Is a Business It’s Never Won

When the AI answers the question, the money doesn’t always follow.

Of course, Google didn’t say it quite that way. What Philipp Schindler, Google’s chief business officer, did say on their Q3 2026 earnings call was that “the strong growth in AI Mode queries had an impact [on Search growth] because of the lower monetization we were seeing in AI Mode.”

In plain English, people still like asking Google’s AI questions. But Google is finding out that it doesn’t make as much money as easily from those answers as it does from a page of blue links and sponsored ads.

If this storyline sounds familiar, it should. It’s the same one that Google first handed to a lot of other businesses that relied on Google to push search traffic to their own sites about two years ago.

In an odd, perhaps even ironic sort of way, Google now knows what all that feels like.

Then again, please don’t start shedding tears for Google.

What Google Actually Said on the Q3 Earnings Call

Google reported on that call that search revenue grew 12% from a year ago to $63.4 billion. Google’s CEO, Sundar Pichai, said that people are searching more than ever; saying that query growth was “the highest it’s been in the last five years.”

Next comes the “but,” and the part of the story that got the headlines, tanked its stock and dented its market cap to the tune of nearly $500 billion in a single day.

That part of the story is that the growth of the crown jewel search business, the core business that puts the ad revenue numbers on the board, is slowing.

In Q2, Google reported that search grew about 17%. Ninety days later, that number was 12%, and revenue barely moved quarter over quarter. Google pointed to currency swings and changes in EU customs rules. Then a tough comparison with a World Cup quarter.

But then, for the first time, it named its own AI.

That’s the part worth digging into. Macro forces like tariffs and currency swings affect every business that operates on the global stage. It’s also what CFOs and treasurers manage day in and day out.

It’s when that happens on top of a slowdown in the core business that investors start to get a little antsy.

Read More: Why Generative AI Is a Bigger Threat to Apple Than Google or Amazon

Google spent two years selling everyone on the idea that AI answers would be good for the web. On this call, it told investors that maybe it’s not all that great for Google.

When the Google AI Overview Chickens Come Home to Roost

In the good old days of three or four years ago, businesses obsessed over their Google rankings. Optimizing them spawned an estimated $68 billion SEO industry to help companies put their best organic Google-search foot forward. Companies produced content built to rank, not just to read. They hoped to match whatever the latest algorithm update rewarded. And when that wasn’t enough, they bought Google ads.

Sometimes all of that still wasn’t enough. It was reported that HouseFresh, a small site that reviews air purifiers, watched its Google referrals fall from about 4,000 a day to about 200 after a single round of search changes in 2024. It reportedly cut most of its staff.

Then came AI Overviews, and the bottom fell out for a lot more people.

Pew Research tracked the real searches of 900 U.S. adults and found people clicked on a result 15% of the time when there was no AI summary. With one, that fell to 8%. Only 1% clicked a link inside the summary itself.

It’s been widely reported that Business Insider’s search traffic fell 55% between 2022 and 2025. Chegg’s fell 49% in a single month, and its stock dropped to about a dollar. Both relied on selling advertising to visitors Google search traffic once pushed to their sites. The Planet D, a travel blog that had been around for years, lost half its traffic after AI Overviews launched and has since shut down.

Businesses that counted on search for inbound leads also felt the consequences of the AI overview takeover on the Google homepage. No traffic, no lead capture.

Read More: Who Will Power the GenAI Operating System?

When Penske Media and Chegg took Google to court, the judge threw out the case. Google had no obligation to send anyone traffic, he wrote, because “an expectation is not an agreement.”

True. And that may be the letter of the law. But now we see the impact of Google’s AI Overviews cutting both ways. Publishers expected that if they played by Google’s rules, the traffic would keep coming. Google expected that AI answers would earn like the search pages they replaced.

And, as it turns out, everyone was wrong.

Why Agentic Commerce Needs to Work for Google

Surely, this couldn’t have caught Google by surprise. I’d be stunned if Google hadn’t whiteboarded a scenario where AI Overviews cannibalized search and had an impact on its core source of revenue.

What maybe wasn’t on that whiteboard is how fast consumers and business executives moved their searches to smarter agents. And how fast businesses adapted their own business models to out-Google Google.

Two other things may also have been overlooked. Agentic commerce hasn’t moved as fast as everyone expected a year ago. And Amazon, Google’s longtime rival for product searches, still captures the sale no matter where the search starts. Google included.

Read More: When Chatbots Replace Search Bars, Who Wins at Checkout?

That’s why agentic commerce — AI agents that find, compare and buy things for people — has become such a big bet for Google. Shopping ads are only as valuable as the number of people who start product searches on Google.

And as we have seen from Google’s own numbers, that number is under pressure.

PYMNTS Intelligence finds that about 66 million U.S. consumers now begin their online shopping research with an AI assistant. Among people who shop with AI, 64% say they now research products on AI platforms that they used to look up on Amazon. Google is still the most common starting point, but the share starting elsewhere is growing fast. In September, 25% of all consumers named AI as their main starting point for retail research, up from 21.4% in June. And among AI power users, 57.1% now prefer a standalone AI platform for product research, compared with 28.2% who prefer search.

When shoppers start somewhere else, Google’s ad business has nothing to sell against. That’s the pressure Google acknowledged for the first time in Q3.

Google’s Big-Bucks Bet on AI Agents and Agentic Commerce

So, naturally, Google is spending like it knows what’s at stake.

It raised its 2026 capital spending plan to between $195 billion and $205 billion and signaled more than $300 billion for 2027. Most of that goes to AI infrastructure that also powers its booming cloud business.

But the commerce push on top of it has been nothing short of relentless. In a little over a year, Google launched the Agent Payments Protocol, the Universal Commerce Protocol, Universal Cart and Direct Offers. It signed up Walmart, Target, Shopify, Etsy and Wayfair along the way.

Read More: Google’s Agentic Wallet: Is the Fifth Time the Charm?

It’s also paying to be where people already ask their questions. Apple confirmed in January that Google’s Gemini will power the new Siri, under a deal reported at roughly $1 billion a year. Good luck to both of them with that.

And when an analyst asked on the earnings call how agentic shopping would make money for Google, Schindler said Google is “confident that we can actually provide really interesting, compelling advertising models for agentic use cases.”

Yes, you read that right. Google’s plan to fix the ad business in an agentic world?

More ads.

What “More Ads” Looks Like Inside an Agent

It’s worth being precise about what Google means, because its plan doesn’t insert ads into the agent. It turns the agent into the ad.

Google has stacked four things on top of each other this year.

The first is the layer merchants will see right away. Since February, Google has been serving shopping ads inside AI Mode, pulled from the same Shopping and Performance Max campaigns merchants already run, marked as sponsored. Same budget, same product feed, new surface. Shoppable ads moved into the answer.

The second is where the model changes. Direct Offers is a Google Ads pilot that drops a merchant-funded discount into AI Mode when Google decides the shopper is ready to buy. Google’s own ads liaison described it as less like an ad and more like a salesperson negotiating a deal on the shopper’s behalf.

For Google, that’s a moment they can monetize. The merchant funds the discount and, at the same time, pays Google for the placement that surfaces it. And since an agent doesn’t read ad copy or look at creative, what the merchant is really buying is a seat in the agent’s consideration set and the money to win the comparison once it’s there. Double the pleasure, double the spend.

Read More: What Happens to Stores When AI Agents Do the Shopping?

The third is Business Agent, a branded AI sales associate that merchants can run right in Google’s search results. It is more or less a sponsored conversation with Google capturing the lead.

The fourth is checkout. Through Google’s Universal Commerce Protocol, a shopper can now buy inside AI Mode or Gemini using Google Pay, with the merchant still the merchant of record. Universal Cart, introduced at Google I/O in May, ties all of it into one basket the shopper carries across Google’s surfaces.

Notice what Google isn’t doing.

So far, it isn’t charging merchants a cut of the sale. Then again, it doesn’t have to. Google’s ad business was built on the click. Someone clicks, lands at the merchant, the merchant logs the conversion and the ad gets the credit. An agent that buys on the shopper’s behalf removes the click, and with it the thread that ties the sale back to the ad. So Google wants to own the checkout. The merchant still gets the order and stays the merchant of record. Google gets the moment the conversion is proved, which is what lets it sell Direct Offers against outcomes instead of clicks. Checkout is the proof that the agent ad engine worked. That’s the business.

Merchants will treat this as the agentic version of shoppable ads, at least at first, and that’s by design. It runs off budgets they’ve already approved with no new line item. And according to PYMNTS Intelligence research, merchants are prepared to pay.

But two things could sink in over the next few quarters.

Google’s lever is price, not brand, because the agent compares structured attributes and discounts, not stories.

And the merchant is paying for the privilege of being auctioned against everyone else in the feed.

Just like the good old days.

Google Wants to Be the Mall, the Anchor Store and the Cash Register

I wrote in August that AI shopping agents are becoming retail’s new anchor store, and the landlord too. In a physical mall, the anchor drew the traffic, the operator leased the space and tenants could negotiate because those were three different parties. In an agentic world, one system creates the demand, decides which merchants appear and sets the terms. The prompt becomes the anchor. It’s the reason the shopper shows up at all. Agents become the anchor and the landlord all rolled into one.

Google’s stack is that argument made concrete.  AI Mode is the anchor. The Shopping Graph is the leasing office. Direct Offers is the rent. And UCP checkout makes Google the cash register in every store in the mall.

Read More: Why the Mall of the Future Won’t Have a Parking Lot

I also noted in that piece that a 2005 study of shopping centers found most anchor tenants paid no rent, while everyone else paid rent tied to their sales. Free referral traffic is how a platform’s monetization starts, not how it ends. Merchants are getting free discovery inside AI Mode today. Direct Offers is the first invoice.

Merchants have leverage right now because agents need their catalogs, their real-time inventory and their checkout. That leverage lasts only as long as no single agent becomes the only place those three things live.

And for Google, as long as consumers and businesses keep showing up to AI Mode for the answers to their search. Not a given.

Agentic Isn’t Google’s First Commerce Rodeo

Google invented search advertising, and it’s still what Google does best. It’s tried commerce plenty of times. It just hasn’t moved the needle.

Google Checkout launched in 2006 to take on PayPal and shut down in 2013. Google Shopping Express, later Google Express, launched to take on Amazon with same-day delivery. Walmart reportedly pulled out, and Google folded it into Google Shopping in 2019. Buy on Google, which let people check out right in search results, ended in the U.S. in September 2023.

Each time Google tried to own the purchase instead of selling the ad, retailers held back and shoppers bought somewhere else. Shopping ads, meanwhile, became a huge business.

Google is great at getting paid to point shoppers to the store. It’s never been good at being the store. And pointing only works if shoppers start on Google, which more and more of them don’t.

Read More: Why 30 Million US Consumers No Longer Search

This time there’s one real difference. Checkout, Express and Buy on Google all asked shoppers to form a new habit and come shop at Google. Agentic checkout asks for the opposite: keep doing what you’re doing, just don’t leave.

That’s a smaller ask of shoppers and a much bigger ask of merchants. It’s also a bigger risk for Google.

Google’s Agentic Commerce Competition

In an agentic world, Google is up against companies that speak commerce as a first language. The marketplaces and platforms that already sit between merchants and shoppers aren’t waiting for Google.

Start with Amazon. When Meta launched Muse, its personal AI agent, in September, people immediately sent it shopping on Amazon. Within about two weeks, Amazon blocked it, saying an “unauthorized AI agent” was violating its terms. Amazon says it’s in “direct conversation” with Meta, and it already has an advertising deal with OpenAI, so some have wondered whether a Muse deal is next.

Read More: The Protocol Power Struggle Reshaping AI-Driven Commerce

I’d be surprised if Muse ever gets past Amazon’s front door.

I’ve used Muse. I’ve also talked to plenty of people who raved about it at first and now say never mind.  My experience was search results that were incomplete, slow to return and often irrelevant. (The one thing it did well was shortlist a name for my Muse: Jewel.) That may be by design. According to instructions the research firm Profound says it pulled from Muse in September, the agent searches no more than three stores on the open web and returns no more than ten products. That’s a short list for a shopper that, in my experience, isn’t all that curated. And a very short shelf for a store the size of Amazon.

So why would Amazon let it in at all?

If it does, my bet is it’ll look like the deal Amazon already has with Meta. Since November 2023, shoppers who link their accounts can buy from Amazon’s product ads on Facebook and Instagram without leaving the app, checking out with Amazon. Meta gets the ad. Amazon gets the sale, the customer and the data. Instagram is a shop window. Amazon is still the store.

It’s the same deal Amazon has with Google. Amazon lets Google’s search engine show its product pages, because every one of those results sends the shopper back to Amazon’s store, where Amazon’s checkout and Amazon’s ads are waiting. Google gets to point. Amazon gets the sale. But even that relationship has cooled. In July 2025, Amazon pulled out of Google Shopping ads worldwide, where it had been the biggest bidder by far, with a U.S. impression share of about 60% against the median competing retailer. A month later it turned Shopping ads back on everywhere except the U.S., and more than a year on, it still hasn’t come back. That same summer, Amazon added Google’s browsing agent, Project Mariner, to the list of AI bots its robots.txt tells to stay out. So Google can still point U.S. shoppers to Amazon. Amazon just stopped paying Google for Shopping ads to do it, and it’s told Google’s agent it isn’t welcome.

Read More: The Great Connected Consumer Paradox

Amazon still owns the checkout. A PYMNTS Intelligence survey of 2,191 consumers in August found about 60% of AI-assisted purchases close on Amazon’s site, and 58% of AI shoppers usually buy on Amazon even when AI helped them find the product. AI is changing where shoppers start. So far, it hasn’t changed much about where they finish.

Merchants are hedging the way a tenant does when the landlord also owns the anchor store. Walmart lets Gemini show its products, then hands the basket to its own assistant, Sparky. QVC and HSN send their catalogs to both OpenAI and Google. Kohl’s doesn’t let agents buy at all. Nobody is betting the store on Google alone.

Shopify goes furthest. It co-developed Google’s Universal Commerce Protocol, then kept going. On October 29, it launched tools that let shoppers send any AI agent to search, buy and track orders across its merchants, paying with Shop Pay, from a catalog of more than a billion products. The agents already running on Shopify’s checkout include ChatGPT, Copilot and Meta’s Muse, the same Muse Amazon just locked out.

Shopify doesn’t care which agent a shopper picks because it owns the merchants, the products and the checkout. Gemini is just one door into Shopify’s store. Google helped write the rules. Shopify is turning them into everyone’s on-ramp. Shopify can afford to open the door to every agent, even a weak one, because it makes its money from merchants’ software and payments, not from ads on their shelves. Amazon makes a lot of money from those ads.

Read More: As AI Assistants Fight Over Shoppers, Infrastructure Gets Paid Either Way

Everybody wants to help people shop. The list of those who’ve got a shot at being both the place search starts and where the purchase ends is short.

So How Good at Commerce is Gemini, Really?

On paper, Gemini appears to have a shot. Google says the Gemini app now has about 1 billion monthly users, up from 950 million last quarter, close to ChatGPT’s roughly 1.1 billion.

Look closer and the picture changes. Sensor Tower puts ChatGPT at about 46% of AI users and Gemini at about 28%. Similarweb has Gemini’s share of AI web traffic slipping from 26.7% in April to 25% in September.

The clearest clue is where Gemini users come from. Among people who use AI platforms, 73% of Android users use Gemini, compared with 49% of iPhone users, according to PYMNTS Intelligence.

Ask consumers which AI is most helpful for a task and ChatGPT wins in all nine task areas PYMNTS Intelligence tracks. Among Android users Gemini leads 65% to 52% because it’s already on their phone. It’s hard to avoid using Gemini if you are an Android phone user.

That’s Google’s real advantage: being the default. It’s why Google pays Apple to be the default search engine on the iPhone, and why Apple now pays Google to power Siri.

But default is a distribution play. It puts Gemini in front of shoppers. It doesn’t guarantee Gemini reaches iPhone users, who have historically spent more per order.

What matters is where shoppers start and where they finish. Millions of people starting with an AI assistant is real demand for discovery. Most of those purchases still closing on Amazon means the payoff lands somewhere else. That gap is the market where Google is trying to build an ad business.

Right now it’s a gap, not a funnel.

What’s Next for Google in an Agentic World

Agentic commerce is in its very early days, and Google has multiple moves.

The first is the one Google is betting on. The agent becomes the ad. Shopping queries stay on Google because Gemini is on most Android phones and inside Siri, merchants pay for placement and fund the discounts that win the comparison, and checkout on Google’s surface proves the ad worked so Google can price it on performance. On this road Google doesn’t need most people to buy inside Gemini. It needs the commercial question to start on Google instead of ChatGPT or Amazon, and it needs enough purchases to close there to make the attribution math work. Ads in the agent are a defense of the search ad business with a proof of purchase receipt attached.

Read More: Department Stores of the Future Are AI Agents

The second road is the landlord without tenants. Merchants keep doing what Walmart, Shopify and Kohl’s are doing now: let Gemini show the goods, keep the basket and the checkout for themselves, and spread their catalogs across every agent so no single one holds them hostage. Amazon keeps the register. Google ends up with a lot of AI shopping research traffic that monetizes worse than search did, which is the Q3 problem at scale. It still gets paid for pointing. It just gets paid less, by merchants who have more options than they used to.

The third road looks a bit like Microsoft’s.

The Microsoft Scenario

There’s a version of this story where Google is fine in an agentic commerce world. Just playing a very different role.

Microsoft missed mobile. Windows Phone never caught on. Microsoft became one of the most valuable companies in the world anyway by turning Azure into a cloud giant that powers everyone else’s apps.

Google shares many of those attributes. Google Cloud grew 103% last quarter to $30.7 billion. Gemini powers Apple’s Siri. Google’s AI chips and data centers run other companies’ AI. If agentic commerce ends up being owned by Amazon, OpenAI, Meta and the retailers themselves, Google can still make a lot of money selling them the computing power to do it.

That’s not a bad business.

Read More: The Existential Threat That Microsoft Missed — and Could Put Its GenAI Future at Risk

Which road Google ends up on isn’t Google’s call. It’s up to the merchants and the consumers.  They decide whether to hand over the checkout. And they’ve already shown, with Checkout, Express and Buy on Google, what they do when the landlord asks for the register. And consumers are showing everyone, every day, where they like to start their search for the things they really want to buy, their most considered purchases.

It’s Google’s Turn Now

Of course, none of this means Google is in trouble today. It’s still pulling in $82 billion a quarter in ad revenue, and its CEO says that people are searching more than ever. And it’s still a $3.4 trillion company, even after last week’s market rout.

But the people Google squeezed over the last two years already know how this story plays out. When someone else’s AI answers the question, you have to find a new way to get paid. Publishers are trying subscriptions and licensing deals. Businesses are rethinking how they find customers, including how to show up in LLMs.

Read More: Gen AI: The Technology That Broke the Adoption Curve

Now Google has to do the same thing. Its answer is agentic commerce with an advertising twist. Let AI agents find, compare and buy for shoppers, and keep selling ads and placements along the way. It’s a bet that runs straight through commerce, the one business Google has never made work on its own, while leaning on the one it’s always been great at.

For two years, Google’s AI made everyone else find a new way to get paid. Now, it seems, it’s Google’s turn.

Welcome to the club.

Until NEXT time.

Join the 21,000 subscribers who’ve already said yes to what’s NEXT.