Jassy: Agentic AI Shopping Not Yet Impacting Amazon Ads Revenue

Andy Jassy said outside AI agents are "not meaningfully impacting Amazon Ads." He added "yet." Amazon Ads is approaching $20 billion a quarter, grew around 19% last quarter, and nothing in that number is breaking.

The "yet" is the only interesting word. Amazon is building quietly: Rufus, Alexa+, the Buy with Prime API surface, and its product graph are all groundwork for Amazon's own agent. When an agent layer does arrive, Amazon's closed-loop data gives it a relevance edge no horizontal agent from Perplexity or OpenAI can match on Amazon's own goods.

Who gets squeezed, and why

The first real losers are open-web price comparison and affiliate sites. If the agent is trained on Amazon's data, it routes purchase intent directly to a product page, and comparison sites get cut out before they can charge for the traffic. That is a structural problem, not a near-term one.

Mid-tier retail media networks (Walmart Connect, Kroger Precision Marketing, Target Roundel) face a different version. They carry smaller catalogs and less structured product data than Amazon. When agents do start comparison-shopping across stores, the richer product graph wins the query. For SSPs and DSPs without retail media exposure, there is no direct exposure here.

The bottleneck Jassy named

He said the limit on adoption is convenience, not AI capability. Agents aren't easier than one-click yet. That is a design-and-trust problem that moves at the speed of consumer habit, not model releases. For agents to eat Amazon's ad revenue at scale, shoppers have to hand an AI their credit card for real purchases, the agent has to read intent better than a search box, and someone has to own liability when it buys the wrong thing. None of that exists at volume.

What operators should actually do

Nothing in your Amazon playbook breaks this quarter or next. Sponsored product bids, keyword strategy, ACoS targets: all intact. The one investment that pays either way is clean product data. Structured titles, attributes, and specs help your search ranking today and make you legible to an agent tomorrow. Leave the "agentic creative" budget at zero until a client will pay for it.

Our call: Amazon's advertising revenue for Q4 2026 grows year over year. Confidence is high. Agentic shopping has near-zero adoption, the convenience gap Jassy describes rules out any meaningful shift in a single quarter, and Q4 holiday demand pushes the line higher. We'll revisit when Amazon reports Q4 2026 results in late January 2027.

Andy Jassy, Amazon's CEO, just told the market that AI agents shopping on your behalf are not denting Amazon's ad business. His exact words: outside AI agents are "not meaningfully impacting Amazon ads, and the answer is no." Amazon Ads is closing in on $20 billion a quarter. Then he added one word: "yet."

What's being decided: Nothing, for Amazon, this quarter. The real question for an operator is how long the open-web search-and-affiliate model has before AI agents start routing purchase intent around it. This is easy to undo as a planning assumption, because nobody has to commit capital today. There is no deadline. Amazon's earnings call set the clock only in the sense that it reset the hype.

The Market Analyst. Jassy saying "no" moves nothing on Amazon's stock, because the market already assumed agentic shopping was years out. Everyone else should take note. Retail media networks without Amazon's product graph (Walmart Connect, Kroger Precision Marketing, Target Roundel) carry a smaller catalog and less structured product data. When an agent does start comparison-shopping across stores, the richer graph wins the query, and the thin ones lose. For a generalist: the store with the best-labeled shelf gets picked by the robot. Magnite and PubMatic have no direct exposure. Identity and clean-room players like LiveRamp could gain, because an agent still needs a way to personalize.

The Skeptic. Jassy is answering the question every analyst has been panic-asking, and the plain answer is that this threat is not real at scale yet. For an agent to eat Amazon's ad revenue, you need shoppers to hand an AI their credit card for real purchases, the agent to read intent better than a search box, and someone to own the liability when it buys the wrong thing. None of those exist at volume. The gap between a slick conference demo and something that works for millions of people usually runs years, not months. Amazon Ads grew around 19% last quarter. Model that number. Ignore the LinkedIn clips.

The Operator. Nothing in your Amazon playbook breaks this quarter or next. Sponsored product bids, keyword strategy, ACoS targets: all intact. The useful signal is Jassy's framing: the bottleneck is convenience, not AI capability. Agents aren't easier than one-click yet. At 90 days, expect agencies to wave this quote at clients who demand "agentic-ready" creative, buying time on spec work nobody is paying for. The thing to actually track is whether Amazon quietly tightens product-listing rules to make catalogs machine-readable. That change shows up in your feed requirements before it shows up in any headline.

The Strategist. Jassy's "yet" is doing quiet work. Amazon is calm in public because it is building the walls in private: Rufus, the new Alexa+, the API surface around Buy with Prime, and the product graph are all groundwork for Amazon's own agent. The two-to-three-year setup is winner-take-most. Amazon's closed-loop data gives its agent a relevance edge no horizontal agent from Perplexity or OpenAI can match on Amazon's own goods. The business that gets squeezed is open-web price comparison and affiliate content. If the agent layer is trained on Amazon's data, those comparison sites get cut out of the purchase before they can charge for the new traffic.

The CFO. The spend question is simple: do you fund "agentic" creative and feed work now, or wait? Wait. There is no billable demand, no measurable return, and the convenience gap means consumers aren't using these agents to buy. The one cheap investment that pays off either way is structured product data. Clean titles, attributes, specs. That helps your current search ranking today and makes you legible to an agent tomorrow. Everything else is spec work against revenue that doesn't exist.

Where the council splits. The Skeptic says five-plus years and stop worrying. The Strategist says Amazon is already building, so the clock is running even if the sales line is flat. They are not really in conflict. The Skeptic is right that horizontal agents won't dent Amazon's revenue soon, and the Strategist is right that the losers are the comparison and affiliate sites outside the walls. The second split: the Market Analyst thinks mid-tier retail media networks face a real squeeze; the Operator thinks all of this is a 2027-plus problem nobody should staff against today. Both can hold, because the squeeze is coming and the staffing decision is still premature.

What this hinges on. One belief: whether the limit on agentic shopping is the AI's smarts or the plain fact that it isn't more convenient than one-click. Jassy says convenience. If he's right, adoption stays slow until the experience beats one tap, and that is a design-and-trust problem that moves at the speed of consumer habit, not model releases. De-risk it the cheap way: clean up your product data now, and leave the agentic creative budget at zero until a client will pay for it.

Prediction: Amazon's advertising revenue reported for Q4 2026, on its late-January 2027 earnings call, will grow year over year rather than shrink.

Confidence: High — agentic shopping has near-zero adoption and Amazon Ads just grew about 19%.

Why: Jassy said outside AI agents are not meaningfully affecting Amazon's ad revenue, and he tied the reason to convenience. Shoppers aren't finding agents easier than one-click, so they aren't using them to buy. Amazon Ads grew around 19% last quarter and is closing on $20 billion in run-rate revenue, driven by sponsored products that agentic shopping has not touched. For the revenue line to fall, consumer habit would have to shift in a single quarter, which the convenience gap Jassy describes rules out. The far more likely outcome is continued double-digit growth, with Q4 holiday demand pushing it higher.

Revisit by 2027-02-15: We're right if Amazon reports higher year-over-year advertising revenue for Q4 2026. We're wrong if that line comes in flat or down versus Q4 2025.

The more interesting call sits one layer out, but it needs a longer clock: the first real victims of agentic shopping are open-web comparison and affiliate sites, not Amazon. That one resolves over years, so it stays off the board today.