The shift · Updated July 2026

The buyer is becoming software.

Not eventually. The protocols shipped, the traffic arrived, and the checkout closed — all within about nine months. This page is our standing position on what changed and what it breaks.

01 / What changed

In nine months, agent commerce went from one experiment to a three-way standards contest.

This is the part most merchants missed, because it happened in specifications rather than headlines. The rails were laid while everyone was still debating whether anyone would use them.

StandardBacked byAnswersStatus
UCP
Universal Commerce Protocol
Google, Shopify What to sell and how to build the cart — discovery, cart, checkout, orders, post-purchase. Merchants publish a /.well-known/ucp manifest. Jan 2026
ACP
Agentic Commerce Protocol
OpenAI, Stripe How the payment executes inside an AI surface, without exposing buyer credentials. Now spans cart, feed, orders, authentication and MCP. Sep 2025 →
AP2
Agent Payments Protocol
FIDO Alliance Who authorised the purchase — a cryptographic chain proving the agent acted with the user's consent. Governed 2026

They stack rather than compete — discovery, execution, authorisation — which is why "waiting to see who wins" is the one strategy guaranteed to fail. There is no winner to wait for. There is only whether your product data can answer the questions all three of them ask.

02 / What it breaks

The funnel didn't shorten. It collapsed into one call.

Every optimisation of the last fifteen years assumed a human moving through stages, and assumed you got to influence them at each one. Agents don't move through stages. They resolve a goal.

Then — human funnel
Discover SEO, ads, a beautiful hero image
Consider Reviews, copy, photography, dwell time
Compare Tabs open across five sites
Convert Cart, upsell, checkout
collapses to →
Now — agent resolution
One query "Trail-running shoes under $150 that arrive Friday." Discovery, comparison and checkout resolve in a single pass — against your fields, not your photographs. You are either legible enough to be included, or you are not in the set.

Notice what has no place to happen in the right-hand column: persuasion. There is no stage at which your art direction changes the outcome, because there is no stage. The agent either resolves your product against the constraint or it doesn't. Being included is now upstream of being chosen — and inclusion is a data property.

03 / The evidence

The traffic is already here. The catalogues aren't ready for it.

We'd rather show you other people's numbers than our own adjectives. Each of these is sourced and dated — check them.

DEMAND

693%

Increase in generative-AI-driven traffic to retail sites over the 2025 holiday season. The channel didn't grow — it appeared.

Adobe Digital Insights, Jan 2026
QUALITY

31% higher

Conversion rate from AI referrals versus other channels — and 254% more revenue per visit. Agent traffic arrives pre-qualified.

Adobe Digital Insights, Jan 2026
STAKES

$385bn

US ecommerce spend that could be influenced by AI agents by 2030. Even discounted heavily, the number survives the discount.

Morgan Stanley, via Digiday

04 / Our position

The industry focused on making 3D look better. We’re making it understand better

For years, 3D helped customers see products in new and immersive ways. The next evolution is making those same digital assets understandable to AI. By combining high-fidelity 3D models with structured product metadata, we transform visual assets into intelligent product experiences. This enables AI agents to discover, understand, and recommend products with greater accuracy. The future of commerce belongs to products that are not only visible, but intelligent.

Here's what a 3D model actually is: a dimensional, material-accurate, queryable description of a physical object. Not a picture of a chair — a machine-readable account of what the chair is. Height, materials, tolerances, articulation, mass. It is the most complete description of a physical product that we know how to make.

We built machine-readable product truth and marketed it as eye candy. Then the machines showed up.

// What the industry sold
{ "asset": "hero_render_4k.png",
  "purpose": "look premium" }

// What it always was
{ "geometry":  "verified, dimensional",
  "materials": "named, physical",
  "variants":  "enumerated, resolvable",
  "reader":    "not human" }

05 / What to do about it

What should you do now? Start Monday, not Q4.

01

Find your AI-ready opportunities

Start with your highest-value products and identify where richer product intelligence can drive better discovery.

02

Build the foundation for AI commerce

Let MetaZtech.ai convert your catalog into structured, AI-readable digital assets without replacing your existing ecommerce investments.

03

Be ready when AI agents start shopping

Ensure your products can be understood, recommended, and purchased by the next generation of intelligent commerce platforms.

Start here

How legible is your catalogue?

Paste a URL. We'll crawl it the way an agent would and send back a scored gap report. Free, and yours whether or not you ever talk to us.

Get your Agent-Readiness Score

Aggregated and anonymised, these scores become the annual Agentic Commerce Index.