Bernstein: AI Agent Disruption Overstated, Emerging U.S. E-Commerce Stocks Unfairly Sold Off
nashnova research
U.S. internet-platform stocks fell a median 7% in two days after Meta, OpenAI, and Anthropic launched consumer AI-agent products in quick succession; but Bernstein's latest note argues agentic commerce is still in its infancy, consumer behavior has barely shifted, and this sell-off is a sentiment over-reaction, not a fundamental deterioration.
How much real traffic are AI agents actually driving?
Bernstein cites Similarweb data for August 2026: AI-chatbot referral share to top e-commerce platforms is tiny — Wayfair 1.4%, Walmart 1.2%, DoorDash 0.4%, Instacart 0.3%.
This means → despite broad adoption of ChatGPT and Gemini, AI referrals convert to less than 1% of actual e-commerce traffic. The "AI agents will disrupt platforms" narrative is running far ahead of the numbers.
In plain terms = consumers ask questions inside an AI chat window, but very few click through and buy — the traffic hasn't turned into transactions.
Are consumers ready to hand their wallets to AI?
Checkout.com's *Agentic Commerce 2026* report shows 25% of consumers say they will never delegate a purchase decision to AI; 27% do not trust any organization to run an AI shopping agent.
The average amount consumers would let an AI agent spend without additional approval is just £177 — below the £200 merchants expect.
This means → trust is a harder bottleneck than technology. Even where AI-agent capability exists, consumers will only let it spend small amounts; big-ticket purchases still require a human decision.
Which platforms have the strongest moats — and which are most exposed?
Bernstein scores platform moats on six dimensions: supply uniqueness, purchase-decision complexity, transaction complexity, competitive positioning, customer loyalty, and excess take-rate. Verdict: ride-hailing and food delivery are the most resilient.
Uber's driver-side liquidity, DoorDash and Uber Eats' restaurant breadth, and Instacart's real-time grocery logistics are structural barriers an AI agent cannot easily route around; subscription-based loyalty adds another layer of defense.
By contrast, eBay and Etsy — asset-light platforms whose core value proposition is product discovery — are the most exposed. AI-assisted price comparison strikes directly at what makes them useful. Wayfair's home-goods category relies heavily on visual and emotional decision-making, and its logistics capability is a moat, but market structure remains a concern.
What are Bernstein's ratings?
Bernstein maintains Outperform on DoorDash, Instacart, Uber, and Wayfair; rates eBay, Etsy, Lyft, and Zillow at Market Perform.
Analyst Nikhil Devnani's team draws an analogy to the long-running "Uber vs. autonomous vehicles" debate: sentiment has swung too far negative, while the fundamental impact still needs time to materialize.
This means → the next 12 months are the critical window. Whether consumers actually shift from ingrained shopping habits to agent-driven purchasing — and whether platforms can convert AI capabilities into measurable revenue growth — will determine if this panic was a mis-pricing or an early warning.
市场有风险，内容仅供研究参考，不构成投资建议。