Your Living Room Is Now an AI Shopping Surface

The biggest screen in your house just became a place where an AI agent recommends products. On June 23, Glance and Samsung put a two-way agentic shopping experience on millions of US smart TVs. If you sell physical products, there is now one more surface where an agent decides whether shoppers see you or your competitor.

TL;DR: Glance and Samsung launched an agentic shopping agent on all 2020-and-later Samsung TVs in the US, adding the living-room screen to chat, search, and marketplace as places AI recommends products. The brands that win this surface are the ones whose product data is structured enough for an agent to retrieve and trust. Most catalogs are not, and the gap is invisible until you measure it.

The pattern here is familiar. A new AI surface opens, it uses its own opaque retrieval logic to pick which products to show, and brands find out months later that they were never in the running. It happened with ChatGPT shopping. It happened again when Meta turned its apps into a shopping surface. Now it is happening on the couch.

What did Glance and Samsung actually launch?

Glance and Samsung launched an AI shopping agent that runs inside Samsung's Tizen TV operating system, letting viewers browse and shop fashion, accessories, and lifestyle products directly from the TV. It went live on all 2020-and-later Samsung models in the US, surfaced through the "For You" and "Apps" tabs, with no phone or second device required.

The release describes "two-way, interactive shopping" with personalized feeds built per viewer. According to PYMNTS, Glance (owned by InMobi) gets distribution few commerce platforms could reach alone, and Samsung gets a shopping layer on the screen it has spent years trying to turn into more than an entertainment device.

The mechanics that matter for brands are not the interface. They are what sits underneath: an agent that ingests product data, evaluates it against a viewer's inferred intent, and decides which items make the personalized feed. That decision is the whole game.

Why a TV surface matters more than it sounds

The TV counts as a genuinely new retrieval context, with its own rules and an enormous installed base, where an agent makes the recommendation before any human comparison shopping starts. That changes who gets seen.

Here is the strategic point. Every AI surface that has opened so far behaves the same way: it decomposes what a shopper wants, retrieves candidate products from structured data, ranks them, and shows a short list. This is the core mechanic of agentic commerce, and it does not change from surface to surface. The shopper rarely sees past the first few results. When Glance builds a "personalized shopping feed" on the TV, it is making the same retrieval-and-rank decision ChatGPT makes when someone asks for the best running shoes.

When a new agent surface opens, brands do not lose visibility loudly. They lose it silently, one unretrieved product at a time, and the loss is only visible if you are measuring that surface specifically.

Naveen Tewari, CEO of Glance and InMobi, framed it as commerce that "understands context, anticipates intent, and meets consumers in the moment." Translated for a brand: the agent is guessing what each viewer wants and matching it against your catalog. If your product data does not answer the questions the agent is asking, you are not in the feed.

The surface set keeps fragmenting

Two years ago, "AI shopping" meant ChatGPT. Now it means chat assistants, Google AI Mode, Perplexity, marketplace agents, and the living-room TV. Each surface has different data requirements, different retrieval logic, and a different competitive set. Optimizing one feed for one surface no longer covers you.

The common thread across every row is the same: the agent reads structured product data, not marketing copy, and picks winners from what it can parse. A brand that is well represented on chat can be invisible on the TV if its feed never reaches that surface or arrives with thin attributes. You cannot assume the work you did for one agent carries to the next.

This is why the question "are we AI-visible?" is the wrong question. The right question is "are we visible on each surface our shoppers use, and against whom?" Those are different answers, and only one of them is measurable per surface.

Why most catalogs lose before the agent even ranks them

Most product pages carry five to eight attributes. AI agents evaluate products against far more, because they decompose a single shopper need into many sub-questions. Google calls this query fan-out: one query becomes 8 to 12 parallel sub-queries the system answers separately before assembling a recommendation.

A Surfer SEO study of 173,902 URLs found that 68% of AI-cited pages are not in the top 10 organic results, and content matching fan-out sub-queries saw a 161% citation lift. This is passage-level retrieval at work: the agent pulls the specific data that answers each sub-question rather than ranking whole pages. The takeaway for commerce is direct: the products that win agent recommendations are the ones whose data answers the specific sub-questions, not the ones that rank highest in classic search. We walk through this end to end in our query fan-out for commerce breakdown.

On a TV shopping feed for fashion and lifestyle, the agent's sub-questions might cover fit, material, occasion, season, price band, and style match to the viewer's prior browsing. A catalog that lists a jacket with a title, price, and three bullet points cannot answer those. A catalog enriched to 30-plus attributes can. The difference decides who lands in the feed.

This is the work behind AI readiness: making sure your product data is complete and structured enough that any agent, on any surface, can retrieve it and trust it. Paz.ai, an agentic commerce optimization platform, monitors how products surface across these agents and flags exactly where the data falls short.

A note on trust, because the TV raises the stakes

There is a second reason this surface matters. New PYMNTS Intelligence research from its May 2026 Consumer AI Benchmark found that shoppers happily let agents handle discovery and comparison, but pull back when it comes to payments and irreversible decisions. Consumers want help finding and comparing products, and they want to keep control of the final call.

Read that the right way and the implication is clear. Consumers will let an agent surface and compare products for them, which means the recommendation step is where the value concentrates. Being the clean, complete, trustworthy option the agent picks is worth more than ever, because the shopper is delegating exactly that step. A stale price or a missing attribute is not a cosmetic flaw on the TV. It is a reason to get filtered out.

What to Do This Week

1. List your surfaces. Write down every AI surface where your category shows up today: ChatGPT, Google AI Mode, Perplexity, marketplace agents, and now connected TV. You cannot measure what you have not named.
2. Audit one product's attribute depth. Pick a hero SKU and count its structured attributes. If it is under 15, it will struggle with query fan-out on any agent surface. Map what is missing against the questions a shopper would actually ask.
3. Check your feed reach. Confirm which AI channels actually receive your product feed. Many brands optimize a feed that never reaches half the surfaces their shoppers use.
4. Verify price and availability sync. Agents quote whatever your feed says. Stale data gets you filtered or, worse, recommended wrong. Confirm your sync frequency is hourly or better.
5. Measure per surface, not in aggregate. A single "AI visibility" number hides the surfaces where you are losing. Track found rate and position on each agent separately, and watch who beats you on each.

Frequently Asked Questions

Is shopping on a TV actually new?

Buying through a TV is not new. An AI agent building a personalized product feed and deciding what each viewer sees is. The shift is from browsing a fixed catalog to an agent retrieving and ranking products for you, the same model that reshaped chat and search shopping.

Does this mean people will buy directly on the TV?

The agent's job is discovery and recommendation. As with ChatGPT, the high-value step is getting surfaced and recommended; the purchase typically completes on the merchant's own site. PYMNTS research shows consumers want human oversight on payments, so the recommendation, not the checkout, is where brands compete.

How is a TV agent different from ChatGPT for my product data?

The underlying requirement is the same: structured, complete product attributes an agent can parse. The difference is reach and context. A TV feed infers intent from viewing behavior and leans on visual, lifestyle attributes, so fields like fit, occasion, and style matter as much as specs.

What product data do AI shopping agents actually need?

Agents evaluate far more than the five to eight attributes most pages carry. They decompose a need into many sub-questions covering use case, compatibility, materials, price, and context. Catalogs enriched to 30-plus structured attributes consistently answer more sub-questions and surface more often.

How do I know if I am visible on a new surface like this?

You measure it directly. General brand awareness does not predict whether an agent retrieves your specific products on a specific surface. Tracking found rate, position, and competitor share per surface is the only way to know whether a new channel is sending you shoppers or skipping you.

A new screen just joined the list of places where an agent, not a person, decides which products get seen. The list will keep growing, and each new surface arrives with its own retrieval rules and its own competitive set. The brands that treat AI visibility as a per-surface, measurable discipline will keep showing up as the surfaces multiply. The ones treating it as a single box already checked will keep finding out too late that they were never in the feed.