Zero-Click Shopping Is Coming: Ascend AI on Agentic Commerce, AEO/GEO, and the New “Feeds War”

Kevin’s answer is refreshingly grounded: people are already experiencing the first phase of it every day through LLM interactions, asking for recommendations, comparisons, shortlists, trip plans, and shopping options. That’s “agentic” in the sense that the model is doing a form of search + synthesis that used to require clicking through websites.

The more futuristic version is where the model (or a dedicated agent) doesn’t just recommend, but actually executes: selecting an item, creating the transaction, generating the purchase order, and finalizing everything without a human ever opening a checkout page.

Kevin expects this to land earlier in B2B workflows than in consumer shopping, for a simple reason: businesses often don’t care about the browsing experience. If procurement needs 10,000 units of something, the “shopping” part is overhead. In those environments, it’s easy to imagine agents triggered automatically by inventory systems, quietly sourcing, ordering, and processing paperwork.

Consumers, on the other hand, often enjoy choosing. Kevin’s skeptical that most people will fully hand over the steering wheel for personal shopping anytime soon.

Zero-Click Commerce: The End of the Checkout Page

Damien frames what many listeners are already imagining:

Kevin says yes: that’s the direction. The phrase he uses is zero-click commerce—the idea that the entire transaction happens inside the LLM interface.

But he points out the non-obvious consequence: even if OpenAI or Google becomes the “shopping surface,” the merchant still has to run reality:

• payment must be accepted securely
• customer details must be transmitted
• inventory + fulfillment must remain correct
• identity has to carry through, or fraud explodes

And critically: the model can’t reliably make purchasing decisions off “human text on a product page” alone. When products are similar (red vs blue variants, slight differences in SKU, sizing nuance), the model needs richer, structured, machine-friendly data, otherwise hallucinations and mismatches become a real operational risk.

Why “Agents Browsing the Web” Is a Bad Near-Term Plan

Damien asks whether companies can just let agents use the current web the way humans do.

Kevin says: they can, but it’s painfully inefficient.

He points to OpenAI’s agentic browser efforts and describes the typical approach: the agent loads a page, takes screenshots, runs internal reasoning to interpret what it sees, then tries to click around. It works, but it’s slow, expensive, and fragile.

The fix is to stop making models “look at pictures of websites” and instead give them direct data interfaces that expose product truth cleanly.

That’s where the acronym avalanche begins.

AEO vs GEO: Discoverability Has Split Into Two Jobs

Kevin distinguishes two layers of “showing up” in model answers:

AEO: Answer Engine Optimization

This is the near-term, relatively structured world: helping systems like Google’s AI answers return your brand because your site clearly answers common questions. It’s often about schema, page structure, and making key facts easy to extract.

GEO: Generative Engine Optimization

This is the harder, fuzzier layer: how your brand exists across the broader training signal landscape, reviews, sentiment, citations, forum discussions, community references.

Kevin’s point: GEO isn’t just “do SEO better.” It’s “the entire public narrative around your product becomes part of discoverability.” And because LLMs treat sources like community content as validation, brands are now trying to influence places like Reddit, one reason many communities are seeing a flood of bot activity.

AEO is increasingly a technical exercise. GEO is a brand + community + reputation exercise at internet scale.

The Real Shift: Feeds, Protocols, and “Traction Points”

Damien pushes the builder question: “What do we actually have to build?”

Kevin’s framing is useful: models need traction points, fast, high-confidence, structured signals they can latch onto instead of expensive screenshot interpretation.

That traction comes from a growing ecosystem of commerce data standards and protocol-like feeds:

Google’s Path: UCP Through Merchant Center

Kevin explains that Google’s Unified Commerce Protocol (UCP) is administered through Google Merchant Center, extending product feeds with richer attributes beyond basic availability and price, things like:

• occasion
• recipient
• sentiment
• contextual descriptors that map to why someone wants the product

The tricky part isn’t the endpoint, it’s the enrichment. If you have 10,000 SKUs, who decides sentiment for each? How do you keep it accurate? This is still extremely new and rapidly evolving.

OpenAI / Cross-LLM Reality: No One Feed Wins Yet

Kevin’s blunt take: there’s no single standard “feed to rule them all” right now. Short term, brands may need to support multiple parallel representations depending on where customers are shopping:

• platform-specific feeds (like Google’s ecosystem)
• on-page structured data (like schema.org JSON)
• syndication gateways that push enriched product truth wider

He suggests the reason standards aren’t converging quickly is strategic: incumbents want to control the data pipes because those pipes become monetization.

Google wants its feed ecosystem because it anchors future ad and commerce leverage. OpenAI doesn’t want to hand that control to Google. And that tension makes fragmentation likely, at least for a while.

What About MCP-Style “Shoe Buying Tools”?

Damien asks a natural question from the agent world: why not just expose an MCP server as a commerce tool?

Kevin doesn’t dismiss it, but points out the missing piece: discovery.

An agent can’t use an MCP tool it doesn’t know exists. Unless there’s a widely adopted way for agents to find and trust these endpoints, the “every merchant ships their own tool” idea doesn’t automatically scale.

Where Kevin does see this idea emerging is in distribution-controlled environments, large brand “mini-app” experiences inside LLM platforms (he mentions rumors around OpenAI hosting small brand app experiences). In those cases, the platform can make the tool discoverable because it owns the marketplace surface.

That’s great for the platform and big aggregators, but it’s messy for merchants, especially when intermediaries take huge commissions.

Adoption: Low Today, But Pressure Is Rising Fast

Kevin says adoption of these newer enrichment layers is still very low.

If you’re on Shopify, some pieces (like ACP-related functionality) may “just happen” behind the scenes—part of Shopify’s advantage. But if you’re on more complex enterprise setups, you may be stuck fabricating these pipelines yourself.

And the urgency is real: Kevin cites brands seeing organic search traffic declines up to 40% (and sometimes more) depending on demographics and category.

Some brands won’t feel it yet. Others are already near an existential cliff, so they’re leaning in aggressively, hoping early mover advantages matter as models reinforce “what gets chosen.”

Analytics Breaks in a Zero-Click World

Damien nails the scary implication:

If people don’t visit your website, your analytics tools lose their visibility.

Kevin agrees. Brands may still see referral traffic from LLM surfaces, and they’ll track “share of voice” by running prompt panels and measuring how often they appear. But the classic metrics (scroll depth, time on site, behavior flows) fade away.

Even ad products, Kevin notes, are immature by traditional standards. If a platform only gives impressions + clicks, sophisticated marketers won’t trust it with serious budget. Eventually, advertisers will demand richer attribution (like prompt context) because without it, you can’t build a strategy loop.

But platforms also want to appear neutral. If they expose too much, it invites “buying the answer,” which risks trust collapse. So the analytics future looks like a tug-of-war between platform neutrality and advertiser demands.

The Identity Question: Merchants Will Insist

One thing Kevin is confident about: identity has to pass through.

Merchants can’t run sustainable operations without knowing who they’re selling to, where the order is going, and how to handle support and follow-up. And many businesses rely on the post-purchase relationship, email, SMS, loyalty loops, replenishment cycles.

If zero-click interfaces block that relationship, it weakens brands structurally.

Kevin hints at Google’s transaction security layers (like AP2) as part of how identity and payment security will be handled—but the bigger point is that merchants will push hard to keep customer continuity intact.

The Big Takeaway: The Web Is Becoming Machine-Negotiated

This episode isn’t really about acronyms.

It’s about a deeper shift: commerce discovery and purchasing are moving from “humans browsing pages” to “models selecting options.”

That forces brands to compete on a new axis:

• structured truth (feeds, schema, enrichment)
• narrative truth (GEO: what the world says about you)
• platform truth (which ecosystems you integrate with first)
• operational truth (security, identity, fulfillment reliability)

In the short term, it’s messy. Multiple protocols, unclear winners, shifting specs, awkward measurement.

But Kevin’s underlying message is simple: if your business depends on people finding and choosing you online, you can’t treat this as optional anymore.

Learn More

Kevin shares that Ascend Labs works with mid-market companies on AI literacy and implementation—especially where digital commerce is core.

You can find Ascend Labs at ascendlabs.ai, and Kevin on LinkedIn