A Practical Guide to Getting 'Chosen' by AI Agents – 2026 Edition: Product Data and Protocol Adoption at the Forefront

Forbes Systematizes E-Commerce Strategy for the Agentic AI Era

A 2026 Guide To Getting Agentic AI To Recommend Your E-Commerce Site

As AI agents increasingly influence online shopping decisions, e-commerce brands must adapt their strategies to be recommended by these autonomous systems.

On February 22, 2026, Catherine Erdly, a Forbes contributor and retail strategy expert, published a practical guide for e-commerce brands to get recommended by agentic AI. In an era where AI agents compare, select, and purchase products on behalf of consumers, the guide systematically explains what e-commerce brands need to do to earn a place on the "invisible shelf."

Industry Trends

Agentic commerce is the most closely watched structural shift in the e-commerce industry in 2026. McKinsey's forecast projects up to $1 trillion in agentic commerce sales in the U.S. B2C retail market and $3-5 trillion globally by 2030. Morgan Stanley similarly estimates that 10-20% of U.S. e-commerce ($190-385 billion) will be mediated by AI agents during the same period.

In traditional e-commerce, consumers entered keywords into search engines and visited multiple sites for comparison shopping. However, approximately 39% of consumers are already using AI for product discovery, and generative AI searches related to shopping surged 4,700% between July 2024 and 2025. ChatGPT has begun driving traffic to e-commerce sites equivalent to roughly 20% of Walmart's total referral traffic.

This trend has created a new challenge: if your brand isn't recommended by AI agents, it won't even enter consumers' consideration set.

Machine Readability Determines Recommendations — The Critical Importance of Data Quality

In the agentic AI era, the single most important factor determining whether an e-commerce site earns recommendations is "machine readability of product data." Clickvoyant founder Mia Umanos defines this concept as a "semantic density score" — the amount of high-quality, machine-readable product data that an AI agent can extract from a single query, compared against competitors within the same category.

While traditional SEO focused on keyword optimization, the following elements are critically important for optimizing for agentic AI:

Completeness of structured data. Every product variation needs valid GTINs (UPC/EAN), with attributes standardized using consistent terminology. According to Opascope's analysis, brands that cover all 170 available attributes in Google Merchant Center and apply Schema.org markup and GS1 standards are significantly improving their visibility to AI agents. Brands with core attribute fill rates above 95% hold an overwhelming advantage in agent recommendations.

Shifting from marketing copy to practical descriptions. AI agents evaluate descriptions that address actual customer use cases, not emotional catchphrases. Instead of "stylish running shoes," specifics like "overpronation support, 38N reactivity, sizes 25.0-30.0cm, 8mm heel-to-toe drop" are required. Product titles are recommended to be under 150 characters, using natural language with descriptive expressions.

Real-time inventory synchronization. AI agents prioritize "data reliability" over brand trust. Brands that cannot provide accurate inventory data, reliable pricing information, and precise delivery timelines are excluded from agents' recommendation lists.

The Shift to GEO — Generative Engine Optimization Beyond SEO

Erdly's guide places particular emphasis on the transition from traditional SEO to "GEO (Generative Engine Optimization)." GEO is a methodology for structuring content so that generative AI systems like Google AI Overviews, ChatGPT, and Perplexity can accurately understand, recommend, and cite your products.

Gartner predicts search engine usage will decline by 25% by 2026, as consumers shift to AI agents. Since agents aim not just to retrieve information but to "execute tasks," the focus must shift from "keyword optimization" to "solution optimization."

Specifically, this means evolving from "product schemas" that only communicate "what this product is" to "capability schemas" that convey "this product can be delivered in size M to this zip code by 4 PM today." Measurement metrics are also shifting from traditional "search rankings" to "LLM visibility scores" — how often your brand is cited in generative AI responses compared to competitors.

Supporting Two Protocols Becomes a Survival Requirement

Two major protocols are currently competing as the infrastructure for AI agents to actually purchase products.

Google's Universal Commerce Protocol (UCP), announced at NRF in January 2026, is an open standard supported by over 20 companies including Shopify, Etsy, Walmart, Target, Wayfair, as well as Visa, Mastercard, and PayPal. It employs a layered architecture similar to TCP/IP, covering the entire purchase process from product discovery to payment and post-purchase support.

Meanwhile, the Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe, is designed around conversational purchasing experiences, backed by over 900 million weekly ChatGPT users. Transactions incur a 4% fee plus standard payment processing costs (approximately 2.9% + $0.30).

Microsoft is also deploying Copilot Checkout as "the new front door to agentic commerce," requiring e-commerce brands to support multiple protocols simultaneously. For Shopify merchants, UCP is automatically supported through Agentic Storefronts, and ACP can be applied for at chatgpt.com/merchants. For custom platforms, three components need to be built: product feeds (daily gzip-compressed updates), checkout API (5 REST endpoints), and payment integration with Stripe or other processors.

Impact on E-Commerce Brands and How to Act

The actions e-commerce brands must take immediately to earn agentic AI recommendations are clear.

Product data audit and enrichment is the top priority. The era of competing based on advertising spend is coming to an end. Structured data quality has become the primary lever determining visibility through AI agents. Start by checking your Google Merchant Center attribute fill rate and filling in missing attributes.

Building server-side webhook infrastructure is also urgent. Agentic commerce orders cannot be captured by traditional web analytics. Impressions, clicks, and session data that occur within AI interfaces are not transmitted to the merchant side, requiring server-side tracking infrastructure. However, building a mature attribution framework is estimated to take 18-24 months.

GEO reliability risks also require attention. As AIVO Standard's Tim de Rosen pointed out in Fortune, while AI models accurately describe product features and characteristics, they tend to lack consistency when asked about corporate financial stability or security certifications. Cases of AI "doubling down" on inaccurate information have been reported, making it important not to over-rely on GEO's effectiveness and to thoroughly ensure machine readability of official information.

Summary

E-commerce in the agentic AI era demands a fundamental paradigm shift from "being found by consumers" to "being chosen by AI agents." As data shows that AI-generated recommendations deliver 4.4 times higher conversion rates compared to traditional search, brands that adapt to this change stand to gain significant returns.

Key areas to watch include the expansion speed of both UCP and ACP ecosystems, competitive dynamics with Amazon's proprietary agent strategy (Rufus AI, Alexa+, Buy for Me), and how the trend noted by Bain & Company — that consumer trust in first-party agents is 3 times higher than in third-party agents — will unfold. In a new commerce era where data quality trumps advertising spend, e-commerce brands need to start preparing today.