06-reference

stratechery ben thompson michael morton ecommerce ai

2026-06-18·reference·source: Stratechery·by Ben Thompson
e-commerceai-disruptionshopifyproduct-discoveryreferral-vs-distribution

"An Interview with Michael Morton About E-Commerce in the Age of AI" — @benthompson

Why this is in the vault

Michael Morton (MoffettNathanson senior research analyst) makes the analytical case that AI is trending toward being a referral channel rather than a distributor in e-commerce — a distinction with direct implications for how RDCO thinks about agentic commerce, Shopify's moat, and the Universal Commerce Protocol. The interview is data-grounded: Morton measured brand.com traffic from AI platforms vs. traditional search and found brand.com outpaces legacy marketplaces 4-8x via AI referral, compared to ~2x via Google Search. This is one of the cleaner empirical readings of AI-driven commerce disruption available.

Interview highlights

The unfalsifiable bear case. Consumer internet multiples are down ~30% YTD because the AI bear case for e-commerce is structurally impossible to disprove with today's data. Morton frames the challenge: each client meeting surfaces a different apocalypse — AI ordering groceries, AI replacing delivery apps, AI vibe-coding Shopify stacks. The amorphousness is the problem.

Shopify. Morton upgraded Shopify to buy against the "software apocalypse" narrative. His evidence: brand.com (Shopify's customer base) now outpaces traditional marketplaces 4-8x in AI-referred traffic vs ~2x in Google Search. The Shopify Catalog (direct feed integration into AI platforms for product discoverability) accelerates this. Key structural moat: Shopify's negotiated fee structures with Stripe, CAPI integration with Meta (Tobi-Zuckerberg direct deal), and cross-channel API reliability make the "vibe-code your own stack" threat largely fictional for any merchant with meaningful volume.

Referral vs. Distribution. The central frame of Morton's MoffettNathanson report. If AI becomes a distributor (owns checkout, owns customer), incumbent consumer internet companies become API calls — existential. If AI is a referral channel (better, clearer purchase intent earlier in funnel), it is a more-accurate Google Search and a de-risking event. The evidence as of mid-2026: distributor attempts (OpenAI Checkout, ChatGPT commerce) have fallen short; AI is behaving as a referral channel.

Why OpenAI Checkout failed. ChatGPT needed Amazon more than Amazon needed ChatGPT. Catalog coverage was incomplete (consumers noticed), and OpenAI was attempting too many things simultaneously. The referral model is structurally sounder for both parties.

Ads and AI. Google and Meta built empires on referral advertising, not distribution. Google's AI Mode continues that model. Amazon's sponsored-prompt integration into on-site AI search is vertically consistent with their existing model. The ad ecosystem survives; the question is whether the distribution layer (checkout ownership) can ever achieve scale.

Universal Commerce Protocol (UCP). Morton walks back earlier fears that UCP was a Google land-grab on Shopify payments. Key facts: (1) UCP allows bring-your-own-payment-solution; (2) Shopify co-developed it with Google; (3) the UCP council now includes Amazon, Meta, Stripe, major merchants. The consensus layer is forming around incumbents, not around a single disruptor.

Grocery. Amazon's grocery push (Amazon Fresh, Whole Foods) is being repositioned as an AI-enabled replenishment play, but Subscribe & Save's historical friction illustrates the gap between vision and consumer behavior. Morton is watching Amazon's grocery logistics as an underappreciated asset.

DoorDash and autonomous vehicles. Morton is bullish on DoorDash as a founder-led company navigating delivery disruption. Waymo is identified as the standout AV opportunity — geographic density + operational execution creates a defensible moat.

Mapping against Ray Data Co

Reinforces existing vault thesis — UCP/agentic commerce. The vault's Universal Commerce Protocol research brief (06-reference/research/2026-04-24-universal-commerce-protocol.md) analyzed UCP as a verification-layer vocabulary for agents. Morton's update confirms UCP is cross-industry-governed and Shopify-cobuilt — not a hostile Google wedge. Strengthens the case that UCP is the rails, not the threat.

Referral > distribution is the key frame for phData conversations. phData clients selling into commerce verticals should understand that AI-driven traffic is a better referral engine, not a displacement of their existing stack. This reframe is useful in discovery calls where clients fear AI will bypass their digital properties.

Brand.com 4-8x referral multiplier is a data point worth citing. When assessing whether a client's DTC brand is vulnerable to AI disruption, Morton's finding flips the default narrative: AI product discovery favors direct brands with strong catalog quality and trusted signals (price, shipping speed). Catalog completeness becomes a competitive asset, not just an ops task.

Shopify Catalog as analogue to RDCO data infrastructure thinking. Shopify solved the "models don't like raw HTML scrapes" problem by building a structured product feed layer. This is the exact same pattern as the MAC data modeling discipline — structured vocabulary that lets agents operate with conviction. The parallel is instructive when designing data products for commerce clients.

Vibe-code-your-stack threat is a good calibration check. When a client or prospect raises "we'll just vibe-code our way off our current platform," Morton's Shopify analysis provides grounded pushback: the negotiated partnership network (Stripe fees, Meta CAPI, multi-channel API reliability) is not code — it's years of relationship capital. Applicable to any incumbent platform the client is considering abandoning.

Gap surfaced — no RDCO vault coverage on MoffettNathanson e-commerce research cadence. Morton's buy-side data methodology (measuring AI-platform traffic to brand.com vs marketplaces) is a repeatable signal. Worth tracking this series (Morton has now appeared 3x on Stratechery: Apr 2024, Nov 2025, Jun 2026).

Related