Tech • AI • Robotics • Game
Stop seducing your customers. Optimize for their agents
8/10TL;DR
As consumers increasingly delegate shopping decisions to AI assistants, merchants are losing control over persuasion at the point of sale and must instead make products machine-verifiable, discoverable, and easy to transact.
KEY POINTS
A new commercial gatekeeper
When a buyer asks an AI assistant to choose a product, part of purchasing power shifts from the customer to the system that organizes the options. A strong brand still matters when shoppers ask for a named product such as Nike, but far less when they ask for the best running shoe for a bad knee or the right vacuum for a thick carpet and a fixed budget. In those cases, the assistant becomes the practical decision-maker.
Why better products can be ignored
A merchant can lose a sale even with the best product if the assistant cannot verify key facts. In the vacuum example, four unanswered questions blocked recommendation: whether it works on thick carpet, whether it is in stock now, whether it can arrive before Friday, and whether it is returnable. A slightly weaker competing model won because those answers were explicit.
Discovery shifted sharply toward feeds
Product discovery inside AI assistants is increasingly driven by structured merchant data rather than traditional web pages. On June 10, 2026, the share of ChatGPT recommendations pulled from merchant feeds reportedly jumped from 8% to 62%. Across 687 monitored markets, 450 lost at least a third of visibility in three days, while the 10 biggest brands rose from 22% to 42% of mentions.
Selection is driven by verifiable facts
Once found, products are filtered by factual criteria, not by branding copy or interface polish. Required merchant data includes availability, price, seller, and brand, with additional weight given to shipping cost, delivery time, returns, and condition. For identical products sold by multiple merchants, ranking factors include availability, price, quality, whether the seller is the manufacturer or a primary partner, and whether payments are integrated with the platform.
Brand equity still exists, but in archived form
AI systems still rely on a brand’s accumulated reputation through reviews, tests, forums, and past web traces. What weakens is the merchant’s ability to persuade at the exact moment of decision with design or storytelling. The customer may still trust the brand, but the assistant increasingly determines whether that trust is converted into a recommendation.
Major platforms are shaping access
Amazon blocked bots from OpenAI, Anthropic, and Perplexity in its robots.txt, choosing not to be broadly indexed under outside terms. Yet it later agreed to let advertisers buy placement under ChatGPT answers, showing that access is becoming negotiable and commercial. Large merchants can refuse some channels and pay for others, while smaller sellers often lack that leverage.
Checkout matters less than conversation control
Direct in-chat payment has not yet proved decisive. In November 2025, Walmart made 200,000 products available for direct payment inside ChatGPT in the United States, but purchases made mid-conversation converted three times worse than flows redirected to Walmart’s own site. The strategic choke point is not only the checkout rail, but who receives the customer query, interprets it, and writes the recommendation list.
Open protocols do not neutralize power
Standards for product, ordering, and delivery data may be open, including commerce protocols compatible with Shopify, Wix, Target, and Walmart. But open transport does not decide which product best fits a “fragile knee” or a dislike of hard returns. The platform that interprets intent still controls ranking, explanation, and visibility.
Customer knowledge is migrating to assistants
Assistants now hold richer, sentence-level context about buyers than many retailers do. Instead of static CRM fields such as age, basket size, and return history, systems can retain priorities such as running habits, past knee pain, or a dislike of complicated returns. With ChatGPT reportedly exceeding 1 billion weekly users, that conversational memory may become more commercially valuable than the merchant’s own customer database.
Smaller firms can still win if they structure first
The shift favors scale, an example of the Matthew effect, but it is not irreversible. Among 67 merchants that gained more than a third in visibility during one measured shift, success went not only to the largest players but to those whose feeds were ready. A smaller seller with clean stock data, shipping times, return rules, warranty details, and bot-accessible pages can outperform a bigger rival still relying mainly on persuasion and brand polish.
CONCLUSION
The rise of AI shopping agents is turning commerce into a negotiation over data quality, platform access, and machine-readable trust signals. Merchants that adapt early may still gain share, but those that fail to become legible to assistants risk disappearing before customers ever see them.
Ask a question