Amazon settles what TikTok starts. An assistant that reorganizes the catalog does not change that

TL;DR

TikTok makes people want things, and Amazon is where they buy them: of the 10,000 largest sellers on each platform, only 498 sell in both top lists, about 5 percent, and a top TikTok Shop seller is five times more likely to be big on Amazon than the reverse. Amazon settles what TikTok starts, and an assistant that reorganizes the catalog does not change that, because every question it answers starts from a want the shopper already has. The thing that will carry the want to the buy is the shopper's own agent, not a smarter catalog, and it will belong to neither platform.

Where the two platforms stand

Three things a shop owner can repeat from memory. TikTok makes people want things. Amazon is where they buy them. The thing that will carry the want to the buy is the shopper's own agent, not a smarter catalog.

The first two are now measured. Marketplace Pulse matched the 10,000 largest sellers on Amazon.com with the 10,000 largest on TikTok Shop, by business name, and published the result on 10 September. Only 498 businesses appear on both lists, about 5 percent of either. One in five of TikTok Shop's hundred largest U.S. sellers ranks among Amazon's top 10,000. One in twenty-five of Amazon's hundred largest ranks among TikTok Shop's top 10,000. A brand that succeeds on TikTok is five times more likely to also succeed on Amazon than the other way around, and among the 498 that do both, rank on one platform says almost nothing about rank on the other.

The two platforms reward different things. Success on Amazon, in the same analysis, is search rank, review count, advertising budget and inventory in Amazon's warehouses. Success on TikTok is a product whose result can be shown on camera in fifteen seconds, sold by a brand that treats content as its main channel and paid for through creators earning a commission on each sale. A replacement charging cable, a set of storage bins, a pack of printer paper: people search for these and nobody films them.

The bridge today is a brand name typed by hand

This was true before TikTok Shop existed. In January and February 2021 a $30.99 pair of Seasum leggings was Amazon's best-selling clothing item, and "tiktok leggings" was among the hundred most-searched terms on Amazon, per Marketplace Pulse's report at the time. The videos that sold them could not link to a product page; TikTok did not allow it then. So millions of people watched a video, opened Amazon, typed the name, and bought where they always buy. TikTok Shop has since added checkout, fulfilment and a creator commission, and some of those sales now stay in the app. The reflex did not go away. A brand that wins the feed inherits an Amazon business it did not build.

The road does not run back. Nobody who finds a product on Amazon goes to TikTok to buy it, and Amazon's own attempts to become a place where people discover things have a record: Amazon Live could not hold a thousand viewers on a normal day, per Marketplace Pulse, and the feeds it built went the same way.

Meanwhile the feed is a room that has already sorted itself. On TikTok Shop the top 1 percent of sellers drive 60 percent of tracked U.S. sales, and the top 0.1 percent, fewer than 90 sellers, more than a quarter, in Marketplace Pulse's June estimate over nearly 100,000 tracked sellers. TikTok's own claim about itself points the same way: 83 percent of shoppers say they have discovered a new product on TikTok Shop, in research it published with GlobalData in June 2025. The feed is where the want is made, and the making is concentrated in a few hands.

What Amazon's assistant does, in Amazon's words

Amazon's answer to the split is an assistant. In May 2026 it folded Rufus into Alexa+ and named the result Alexa for Shopping. On Amazon's count, Rufus "helped over 300 million customers in 2025 research, compare, and buy the products they want and need." Read the capability list Amazon publishes for it: ask a question in the search bar, build a shopping guide for a big purchase, compare products, see a year of price history, set a price alert and buy automatically at a target price, schedule reorders, add to cart in conversation, find deals from your browsing history, and, through Shop Direct, shop other retailers across the web.

Every item on that list starts from a want the shopper already has. Compare, price, reorder, alert, cart: each one takes a thing you already decided you want and reorganizes the catalog around it. That is a search box with better manners, and it is good at settlement. It does not make anyone want a thing. Marketplace Pulse put it in one sentence, and the sentence is right: "reorganizing a catalog is still a search-first idea."

The one line on the list that reaches outside Amazon is the tell. Shop Direct shops other retailers on the customer's behalf. Amazon's own assistant now allows that the best buy is sometimes somewhere else, and goes there. Settlement is loosening its grip on the buy at the same time the feed has taken over the want.

Where agentic commerce stands in 2026: the want moves into agents

So here is the picture. The want is made in a feed. The buy settles in the store with the card on file and the delivery date promised. The bridge between the two is a person's memory and a search box, and the bridge is what is about to change.

Nobody knows what the shopper's agent will look like in six months, in a year, in two years. It might sit in a chat window, in the phone's assistant, or in a host the person runs themselves. What is already visible is that the demand side moves first. Among consumers who already use answer engines, 38 percent use them to discover products and 17 percent complete a purchase through them, in Forrester's mid-2026 read of agentic commerce. The want is starting to arrive carried by software, with a spec attached, instead of a brand name typed by hand.

The supply side moves second, and it has started: a store answering for its stock, its price and its order, to whatever asks. At some point the two meet in the middle, the shopper's agent talks to the store's agents, and the buy settles between them, wherever the fit and the terms are best. Until then a shop does not get to choose the agent. It works with whatever the demand side adopts.

An agent that carries the shopper's spec is the first thing that can take the want to wherever the buy is best, and it belongs to neither platform. It is not Amazon's, so it has no reason to keep the buy on Amazon. It is not TikTok's, so it has no reason to keep the buy in the app. It holds what the person asked for, reads what is on the shelf, and settles where the product matches. A shop that is readable by that agent can take the buy without owning the feed.

That is the approach we take as well. We do not own the agent. SIL is a shopping capability the person's own agent loads, on a host they run, and it stays theirs when they change hosts: it takes the spec, holds every candidate against it, and returns a buyable variant with a stated reason. We are building toward the day the two sides meet, and we are not guessing which agent gets the shopper there.

What to do on Monday

1. Count where your wants start against where your buys settle. For a shop already on Amazon, TikTok's value shows up as branded searches on Amazon, not as TikTok Shop orders, in Marketplace Pulse's reading. Look at branded search volume the week after a creator post. That number is the bridge as it works today, and it is the number an agent will replace.
2. Make the shelf readable by an agent that carries a spec, not a brand name. A person types a name. An agent arrives with what the product has to be: the material, the size, what it fits, what it does, in stock, at what price, shipped by when. Publish those as structured product data, per product and per variant, where software can read them without a person in the loop. Machine legibility already has a price tag, and an agent with a spec reads nothing else.
3. Do not build a feed to fix a discovery problem, and do not build a chat window to fix a settlement problem. Amazon tried the feed three times. The chat window is the second bet in Instacart's playbook, the one it placed after plugging into every agent it could reach. Neither moves the want. Put the effort into the shelf and into being present wherever the agent looks.

FAQ