If AI agents do the shopping, who are the ads for?

I used to try to buy stuff using AI all the time, and it kept telling me, "Oh, I don't have access to this." Which kind of defeats the point of asking an AI to do your shopping.

Then Amazon started blocking AI shopping agents outright. It sued Perplexity over its Comet browser, and it blocked Meta's Muse from shopping on Amazon. Amazon is the biggest example right now, and since it's the world's biggest online store, where it goes the rest follow.

And honestly, the more I looked at why, the less this felt like a story about whether a bot can put toothpaste in a cart, and the more it felt like a story about ads, for the whole industry and not just Amazon. So here's the question I think actually matters. If AI agents end up doing the shopping, will ads die, or will they need to be reshaped?

Amazon keeps blocking AI shopping agents

Amazon filed its complaint against Perplexity on November 4, 2025, alleging that Comet used hidden automated access to get into private customer accounts and get around Amazon's blocks. A federal court gave Amazon a preliminary injunction in March 2026, the Ninth Circuit threw it out in August, and the case was still going in October, so there's still no final answer on whether an agent like that is allowed in.

What did Amazon say it was worried about? In its public statement, it pointed to security, undisclosed access and a "significantly degraded shopping and customer service experience." It said much the same when it blocked Muse in September 2026, that the assistant used undisclosed automated access and could put customer logins at risk. Those are Amazon's allegations, not settled findings about either company.

But buried in the Perplexity complaint is the line that I think gives the game away. In paragraph 61, Amazon writes, "Amazon's advertisers pay for their ads to be shown to humans," and it explains that it has to detect and filter out bot traffic before it bills advertisers. Amazon frames that as a cost and measurement problem, not as lost ad sales. Still, it's right there in black and white: the ads are for humans, and an agent isn't one.

And Amazon doesn't hate shopping agents in general, since it runs Buy for Me, its own agent that buys from other brands' websites. The question is whose agent gets in, and under whose rules.

Why I think Amazon is scared

I think Amazon is scared, and I think it's mostly about ads. Why? Well, look at the numbers. Amazon reported $19.6 billion in advertising revenue for the third quarter of 2026, up 11% from a year earlier. For all of 2025 it was $68.6 billion, up 22%.

Now picture the shopper who sends an agent to get a good pair of running shoes under $120. That agent doesn't scroll, so it never sees the sponsored listing at the top or the banner for a brand it's never heard of. Why would Amazon be relaxed about that? It wouldn't be, especially after spending years building an ad business around the moment people shop.

To be clear, Amazon has never said ads are the reason it blocks these agents. That's my read. But I'm not the only one reading it that way, because analysts at EMARKETER argue Amazon has more to lose than most retailers, since outside agents threaten both its ad business and its grip on the purchase.

The other clue is that Amazon is happy to put ads inside its own assistant. Sponsored prompts went fully live in the US on March 25, 2026, billed per click like its other ads. Then it launched Alexa+ Agentic Ads, which take you from an ad to a finished purchase inside the conversation. So Amazon clearly wants the agent, as long as it's Amazon's agent and Amazon's ads.

Everybody wants agents on their own terms

Is Amazon the only one doing this? Not really. Its CEO Andy Jassy has said he expects, over time, to find ways to partner with third-party agents, but he has also talked about needing a value exchange that works for both sides. Modern Retail's read is that this is about protecting a browsing-dependent ad business while Amazon figures out its terms.

Google has been at it longer, putting ads right inside its AI answers. It put ads in AI Overviews on US mobile in 2024, began testing ads in AI Mode in 2025, and announced a wider rollout in October 2026. It is also piloting Direct Offers, where a paid deal can surface when Google's AI decides it's relevant.

OpenAI started testing ads in ChatGPT on February 9, 2026, for logged-in US adults on the Free and Go plans. OpenAI says the ads don't influence the answers, though that's OpenAI's promise, not an outside audit. Perplexity tried sponsored follow-up questions and then backed off, because its executives worried ads would make people doubt the answers.

Then there's Walmart. In March, its Daniel Danker told WIRED that in-chat checkout converted "three times lower" than sending people to Walmart's own site. In October, Walmart announced an Agent Gateway with its own rules for agents. So everybody wants agents, but nobody wants to hand over the purchase without getting to set the terms.

Three ways this could go

The first possibility is that agents mostly ignore ads. An agent can compare price, stock, shipping and return policies faster than you can, and it has no reason to care about a flashy sponsored spot. That's the best version for shoppers, and probably the worst for every company that sells your attention.

The second is that ads get aimed at the agents themselves. Time sold sponsored FAQ blocks in machine-readable versions of its pages for agents to read, Digiday reported. Perplexity later said it was blocking those ads from its index and agents as deceptive. So we already have ads aimed at robots, and robots refusing to read them.

The third is that people pay extra for an agent nobody paid to sway. ChatGPT sells ad-free plans, and Anthropic says Claude will stay ad-free. That doesn't prove there's a market for an unbought shopping agent, but the promise is already on sale.

I don't know which one wins, but something's going to happen. McKinsey argues that the money moves from buying attention to influencing decisions and tracking actual sales, which sounds right to me, even if the road there is messy.

And it is messy, because even the buying part keeps changing. ChatGPT launched Instant Checkout in September 2025, and by March 2026 OpenAI said it was "moving away from a standalone Instant Checkout experience." Google launched Universal Cart in October. The biggest companies in retail are still writing the rules in public.

What this means for you

So what does all of this mean for the rest of us? I think retailers, ad platforms and AI companies are going to have to adapt in ways they never planned for, and honestly, I think they created this problem themselves. They stuffed shopping and search results with ads and sponsored listings until people wanted a helper to cut through it, and then they raced each other to build and hype AI agents. Now the agents are here, and they're angry about it. That's my opinion, not something any of them would admit.

What does that mean for your future? It could mean the ads stop being aimed at you and start being aimed at your agent, and you might never know it got nudged. It could also mean you end up paying for an agent nobody paid to sway, the same way people already pay to skip ads on ChatGPT. Either way, I think the real fight over the next few years is whose agent gets let into which store, and that's going to decide a lot about how you shop.

To be fair about the limits, nobody has published audited totals for orders placed by agents across the big assistants, so nobody can tell you yet how big this gets or how fast. But I think the direction is clear, because the companies that made the internet harder to shop with ads now have to live with the helper people wanted because of it.