Skepticism grows over AI shopping agents as ChatGPT checkout launches

Independent analyst Andrew Lipsman published detailed analysis on October 6, 2025, questioning the commercial viability of AI shopping agents despite OpenAI's recent launch of instant checkout features for ChatGPT. The analysis, published through Marketecture, examines eight structural challenges facing autonomous shopping systems.

"The fundamental flaw with 'agentic commerce' or 'agentic advertising' is that it violates the motivations of retail outlets to 1) control the customer relationship and 2) monetize their first-party data with advertising," wrote Eric Seufert of MobileDevMemo in analysis cited by Lipsman. Amazon and Shopify collectively control more than 50% of the U.S. ecommerce market. Both platforms block AI agents to maintain discovery ownership and protect their retail media businesses, which have become critical revenue sources as retail media networks race toward $300 billion by 2030.

The analysis arrives weeks after OpenAI launched instant checkout capabilities on September 29, 2025. That announcement introduced the Agentic Commerce Protocol developed with Stripe, enabling transactions from Etsy and Shopify merchants directly through ChatGPT conversations. Payment processing flows through Stripe, with merchants maintaining responsibility for fulfillment and customer service.

Lipsman defines agentic commerce through an AI-generated definition: autonomous AI agents acting on behalf of consumers to handle complete buying processes including product search, option comparison, price negotiation, and purchase completion independently within user-defined parameters. The definition emphasizes that agents handle the entire buying process rather than requiring constant human feedback loops.

"Before we start calling it a 'revolution,' maybe we can find evidence of a single consumer engaging in an agentic purchase?" Lipsman wrote. The analysis challenges technology predictions that emphasize theoretical capabilities over actual consumer adoption patterns.

A probability analysis in the document examines requirements for successful agentic commerce transactions. Lipsman assigned probability estimates to each necessary condition: shoppers must know how to prompt AI agents (50% probability), trust agents with financial information (50% probability), authorize autonomous transactions (50% probability), and accept products selected without direct inspection (25% probability). Products must fall within categories where agent selection makes sense (25% probability). Agents must possess sufficient category knowledge (75% probability), understand purchase context (50% probability), and select products that don't require returns (50% probability).

The cumulative probability calculation suggests successful agentic commerce transactions face substantial mathematical challenges. Any broken link in the chain results in failed or unsatisfactory purchases, making habit formation unlikely when feedback loops routinely produce negative results.

Historical ecommerce hype cycles provide context. Voice commerce predictions in 2018 warned search marketers to prepare for Alexa-first results. Livestreaming commerce projections in 2019 predicted U.S. market growth to $58 billion by 2025 based on Chinese market success. Metaverse commerce in 2022 suggested virtual store shopping would replace physical retail. None achieved predicted market significance. The document includes data showing these emerging channels represent 0.2% or less of U.S. retail sales in 2025.

Google's "I'm Feeling Lucky" button serves as a historical precedent. Google VP Marissa Mayer revealed in a 2006 Washington Post interview that less than 1% of users selected the automatic navigation option, even when the only downside was clicking a back button. The feature allowed Google to bypass search results and send users directly to the top-ranked website.

Ecommerce return rates present additional complications. Apparel maintains mid-20% range return rates for online purchases, while shoes, accessories, and consumer electronics register double-digit percentages. Walmart achieved ecommerce profitability in 2025 after years of operational challenges. Increased return rates from agent-selected purchases would further pressure already challenging economics.

A CivicScience survey from 2020 identified top ecommerce pain points: paying to ship items back (28%), dropping packages at mail facilities (25%), no easy exchange options (21%), repackaging items (12%), waiting for refunds (8%), and printing shipping labels (6%). Agents provide minimal assistance with these physical-world challenges that inflict time, energy, and money costs on shoppers.

The complexity-competence conundrum presents a structural challenge. When shopping complexity is high—such as purchasing furniture, appliances, or fashion apparel—AI agents are less likely to produce satisfactory results due to personal preferences, contextual factors, and physical product inspection needs. When complexity is low—such as buying books or commodity household essentials—agents don't address sufficient pain points to justify adoption. The document includes a visualization showing this relationship across product categories.

PayPal CEO Alex Chriss offered a grocery shopping scenario in a LinkedIn post: a working mother named Sarah tells her AI agent to plan healthy meals for the week accommodating a child's peanut allergy while keeping costs under $150. The agent creates a meal plan and schedules delivery. Lipsman's analysis identifies potential failure points: duplicate pantry items in the delivery, ingredient choices that conflict with the family's actual dietary preferences, meals children refuse to eat, and allergic reactions from ingredients not produced in allergen-free facilities.

AI agents operate without contextual knowledge that influences purchase decisions. The most important variables for many purchases aren't behavioral but contextual: occasion, physical location, emotional state, urgency, or factors that shoppers don't consciously recognize but influence product selection. Even when consumers include context in prompts, agents may not know how to weigh variables appropriately.

Lipsman examined his own recent purchase history to test agent utility. Five recent searches included: a navy blue v-neck shirt (purchased an 8-pack of blue and grey v-necks), a basketball pump (purchased a Spalding ball pump), white and black OnCloud running shoes (bought OnClouds after 7 minutes, remaining ambivalent about the choice), generic Claritin (purchased Amazon Basic Care Generic Claritin), and a compost bin (purchased an OXO compost bin after trusting only OXO brand quality). Transactions where agents might reach the same conclusion—the basketball pump and Claritin—required minimal thought. Other instances would likely produce suboptimal outcomes.

The analysis addresses ChatGPT's announcement directly: "What OpenAI announced isn't really agentic commerce because it requires human intervention." The "buy" button has been attempted multiple times historically, with consumers consistently resisting on-platform checkout. Social checkout has struggled to gain traction because consumers prefer interacting directly with retailers, don't trust products sold on social platforms, and don't trust platforms with payment information.

Most critically, retailers lack incentive to allow OpenAI to intermediate transactions. Retail media networks have become essential revenue sources, with European retail media spending reaching €13.7 billion in 2024 representing 22.1% growth. Brands working with four to six retail media networks doubled from 10% to 24% in 2025, indicating diversification strategies among advertisers seeking first-party data access.

"Retail platforms have no commercial motivation to allow third-party agents, broadly and without restriction, to browse their catalogue or to make purchases directly. If agentic commerce takes shape, it's likely to occur through narrowly-scoped, explicit partnerships that tilt in favor of the platforms," Seufert wrote.

Amazon has implemented comprehensive restrictions against AI crawlers from OpenAI, Anthropic, Meta, Google, and Huawei. The company updated its robots.txt file on August 21, 2025, to block AI bots from major tech companies while developing competing tools including the Rufus chatbot and a "buy-for-me" feature.

Shopify adopted a different approach. The ecommerce platform introduced warning language to merchants' robots.txt files rather than outright blocking, requiring "buy-for-me" agents to include human review steps and directing developers to integrate Shopify's checkout technology. Shopify has partnered with Perplexity and reportedly is preparing integration with OpenAI to enable transactions through AI chatbots.

The competitive landscape has intensified through 2025. Microsoft launched the Copilot Merchant Program on April 18, 2025. Code analysis revealed OpenAI developing native purchase functionality with Shopify integration around the same time. The timing suggests both companies responded to competitive pressure in AI-assisted shopping.

Industry analyst Karsten Weide offered a contrasting perspective: "Ultimately, blocking personal shopping bots will be a losing defensive battle. First it's going to be drip drip drip in niches, giving competitive advantages to smaller, specialized retailers, but eventually it'll turn into a stream, then a torrent." Weide predicted fundamental changes in consumer commerce patterns driven by AI agents, with brands losing direct customer relationships but gaining connections with "machine servants" executing purchases.

The analysis concludes with a challenge: "If you believe in the power and inevitability of agentic commerce, then use it for your next 5 purchases and let me know how it goes. Want to raise the stakes? Do it for a $100+ purchase, or for a gift for somebody you care about."

For retail media networks, the document suggests minimal immediate concern. "RMNs shouldn't waste a single moment preparing for this 'threat,'" Lipsman wrote in the article's subtitle. The analysis positions agentic commerce as a "collective hallucination" driven by technology capabilities rather than consumer behavior realities.

OpenAI has been building infrastructure for potential advertising integration. The company posted a job listing on September 24, 2025, seeking engineers to develop internal marketing technology infrastructure including campaign management and attribution systems. OpenAI appointed Omnicom Media Group's PHD as its global media agency of record. Additionally, Fidji Simo, OpenAI's CEO of Applications and former Instacart CEO who oversaw that company's advertising product development, has been conducting interviews with candidates to lead a team focused on bringing advertisements to ChatGPT.

The platform reached 700 million weekly active users in August 2025, representing a fourfold increase from the previous year. ChatGPT shopping features expanded to German users in July 2025, with the platform processing over 1 billion weekly searches. Product selection operates through algorithmic methods rather than sponsored placements, with ChatGPT's product results selected independently without advertising.

The technical foundation relies on Bing Search API infrastructure. ChatGPT determines products through intent-based evaluation incorporating structured metadata from third-party providers, consumer reviews and ratings, pre-search model responses, and real-time search results. The system weighs factors including technical specifications, pricing information, product descriptions, and quality assessments from public reviews.

Retailers must avoid blocking OAI-SearchBot, ChatGPT's web crawler, to receive consideration for shopping results. The platform automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs for analytics tracking. OpenAI has indicated plans for direct product feed submissions from merchants to ensure more accurate, up-to-date listings.

The document's author, Andrew Lipsman, is an independent analyst and consultant at Media, Ads + Commerce. His industry coverage specializes in retail media, which he coined as "digital advertising's third big wave." He previously served as Vice President & Principal Analyst at eMarketer, focusing on retail and ecommerce, and as SVP of Marketing & Insights at comScore.

The analysis draws from two decades of ecommerce experience observing technology predictions. "I've been in ecommerce for 20 years and watched a dozen supposedly revolutionary trends promising 'the future of retail' come and go," Lipsman wrote. He notes that even ecommerce itself has been overhyped over the years, with predictions that it would make in-store shopping obsolete now seeming unrealistic given that ecommerce accounts for approximately 15% of U.S. retail spending nearly 30 years into its ascendance.

The retail media sector continues expanding through traditional and non-traditional players. Mastercard launched a commerce media network on October 1, 2025, leveraging permissioned transaction data from more than 160 billion annual payments. The platform operates with 25,000 advertisers and reaches 500 million enrolled consumers. Criteo became Google's first onsite retail media partner on September 10, 2025, enabling advertisers to create campaigns across Criteo's network of over 200 retailers directly within Google Search Ads 360.

UK digital advertising is set to reach £45 billion by 2026, with retail media cited as a primary growth driver. Online retail media hit £1.5 billion in the first half of 2025 in the UK market. The sector has "rapidly established itself as a core channel, reflecting how brands are investing closer to the point of purchase to drive measurable outcomes," stated IAB UK.

Technology providers have responded to retail media growth. Topsort partnered with Skai on September 23, 2025, becoming the first API partner to deliver access to retail media networks across 40 countries through unified platforms. Retail media networks embraced real-time bidding for sponsored products through the Pentaleap-Teads partnership announced July 24, 2025, representing the first programmatic solution enabling advertisers to activate sponsored product inventory across multiple retail networks.

The broader industry context includes substantial investment in commerce-enabled advertising formats. PayPal debuted shoppable storefront ads on June 16, 2025, enabling direct purchases within display advertising units. Digital advertising revenue reached $258.6 billion in 2024, with commerce media including retail media networks growing 23% to $53.7 billion.

For marketing professionals, the analysis suggests focusing on proven channels rather than speculative technologies. Retail media networks provide measurable outcomes through first-party data access and purchase-path engagement. The infrastructure supporting these networks has matured from experimental tactics into strategic advertising systems, with standardization efforts progressing through industry organizations and technology partnerships.

The chicken-and-egg problem remains unresolved: retailers won't overhaul site experiences and metadata structures for agentic commerce until it represents meaningful purchase activity, but agents can't achieve meaningful adoption without optimized retailer infrastructure. This creates a circular dependency that further complicates adoption trajectories.

Subscribe & Save represents the closest existing analog to proactive commerce, yet it remains a marginal ecommerce behavior despite years of availability. Product search shifted from Google to Amazon and other ecommerce sites over time. Comparison shopping engines existed for 20+ years and have declined in popularity. Auto-complete checkout flows and mobile wallet integration like Apple Pay have largely resolved purchase completion friction.

The analysis positions current agentic commerce enthusiasm within historical patterns of technology prediction failures that over-attribute technical capabilities while under-attributing human factors in adoption equations. "We over-attribute the tech and under-attribute the human in the equation," Lipsman wrote, summarizing the core argument against current predictions.

Timeline

• 2020 - CivicScience survey identifies top ecommerce return pain points, with shipping costs representing 28% of complaints
• 2022 - Metaverse commerce predictions suggest virtual shopping will transform retail; these predictions have not materialized
• 2006 - Google VP Marissa Mayer reveals less than 1% of users selected "I'm Feeling Lucky" option
• April 18, 2025 - Microsoft launches Copilot Merchant Program for retail integration
• April 21, 2025 - Code strings discovered suggesting OpenAI-Shopify integration in development
• April 28, 2025 - OpenAI introduces shopping to ChatGPT with 1 billion weekly searches
• May 3, 2025 - Detailed analysis published of ChatGPT product selection methodology
• July 2, 2025 - ChatGPT shopping features expand to German users
• July 9, 2025 - Innovid unveils retail media tools for onsite and offsite campaigns
• July 25, 2025 - Retail media networks embrace RTB for sponsored products through Pentaleap-Teads partnership
• August 30, 2025 - Amazon blocks AI bots from major tech companies including OpenAI, Meta, Google, and Huawei
• September 4, 2025 - Omdia research projects retail media will exceed $300 billion by 2030
• September 10, 2025 - Criteo becomes first onsite retail media partner for Google Search Ads 360
• September 23, 2025 - Topsort partners with Skai to expand global retail media reach across 40 countries
• September 24, 2025 - OpenAI posts job listing for Growth Paid Marketing Platform Engineer
• September 29, 2025 - OpenAI launches Instant Checkout and Agentic Commerce Protocol with Stripe partnership
• October 1, 2025 - Mastercard launches commerce media network with $100B market potential
• October 6, 2025 - Andrew Lipsman publishes analysis questioning agentic commerce viability through Marketecture
• October 11, 2025 - European retail media spending reaches €13.7 billion with 21.1% growth

Summary

Who: Independent analyst Andrew Lipsman, an analyst and consultant at Media, Ads + Commerce specializing in retail media, published the analysis. The analysis cites perspectives from Eric Seufert of MobileDevMemo, PayPal CEO Alex Chriss, and industry analyst Karsten Weide. The document also references Google VP Marissa Mayer's historical comments on user behavior.

What: A detailed examination questioning the commercial viability of agentic commerce—autonomous AI agents handling complete purchasing processes—despite recent technology launches including ChatGPT's instant checkout feature. The analysis presents eight structural challenges: retailer incentives against AI intermediation, historical ecommerce hype cycle failures, probabilistic analysis showing low transaction success likelihood, consumer preference for seeing options before deciding, ecommerce return rate complications, the complexity-competence conundrum, AI agents lacking contextual purchase knowledge, and real-world purchase history showing minimal agent utility.

When: Published October 6, 2025, through Marketecture, arriving weeks after OpenAI launched instant checkout capabilities on September 29, 2025. The analysis examines developments throughout 2025 including Microsoft's Copilot Merchant Program launch in April, ChatGPT shopping feature expansion in July, and Amazon's AI bot blocking in August. The document also references historical precedents from 2006 (Google's "I'm Feeling Lucky" feature), 2016-2022 (failed ecommerce predictions), and 2020 (consumer pain point surveys).

Where: The analysis addresses the U.S. ecommerce market where Amazon and Shopify collectively control more than 50% of the market. The broader context includes European retail media markets reaching €13.7 billion in 2024, UK digital advertising projected to reach £45 billion by 2026, and global retail media networks positioned to exceed $300 billion by 2030. The document examines retailer behavior in blocking AI agents and consumer shopping patterns across physical and digital channels.

Why: The analysis matters because substantial technology and financial resources are flowing toward agentic commerce development despite questionable consumer adoption prospects. Retailers lack incentive to allow AI intermediation that would threaten retail media networks, which have become essential revenue sources growing at 21-23% annually. The marketing community faces strategic decisions about resource allocation between proven channels like retail media networks versus speculative technologies like agentic commerce. The analysis provides a counterweight to technology-centric predictions by examining consumer behavior, retailer incentives, economic structures, and historical precedents showing that theoretical technology capabilities rarely translate to actual adoption when human factors and commercial incentives don't align.