AI shoppers disrupt Europe’s ad industry
A new era of shopping is emerging in Europe, where artificial intelligence no longer functions solely as a recommendation tool but actively executes purchases. Visa’s European transaction systems already support this capability, while protocols such as OpenAI and Stripe’s Agentic Commerce Protocol and Google’s Agent Payments Protocol (AP2) are accelerating its adoption. However, while the technology is operational, most advertising strategies have not yet adapted to this change.
The transition introduces fundamental challenges for advertising, as traditional models rely on human engagement—metrics like impressions, clicks, and conversions tracked along a predictable path. When software becomes the buyer, much of this framework becomes irrelevant. Advertisers must now rethink every aspect of their approach, from how ads are displayed to how success is measured.
Three distinct approaches to AI shopping are taking shape, each with different implications for advertising. In the first model, AI acts as a surface, similar to how ChatGPT suggests products alongside labeled ads. Here, persuasion continues to function as before. The second model involves AI narrowing choices, as seen with Amazon’s Rufus or Google’s AI Mode, where product data becomes the primary creative asset. The third model involves AI acting as the buyer itself, as enabled by Visa’s European transactions and protocols like AP2. Here, only data and proof count, and no proven ad model yet exists for this scenario.
Read Also: OpenAI Safety Executive Resigns, Warns of AI Risks
Advertisers pivot to AI-driven metrics
Advertisers are already adjusting their strategies. Amazon’s legal action against Perplexity over its shopping agent highlights the core issue: how to assign value to an “impression” when the reader is software. The solution may involve shifting from paying for visibility to securing inclusion on an AI’s shortlist or bidding per verified order. ChatGPT’s ad product, launched in February at a reported $60 cost per thousand impressions (CPM), has already begun this transition, with updates expected in late summer allowing advertisers to bid for conversions on campaigns built from their product feeds.
Creative strategies must also evolve. AI systems do not engage with videos or traditional visual content; instead, they analyze prices, delivery times, specifications, and reviews for consistency. A mismatch between a brand’s website and marketplace listing, something a human might overlook, will damage trust with an agent. The product feed, once a back-office detail, now becomes the most critical creative tool. Additionally, attribution models break down, as today’s last-click tracking assumes a human journey: impression, click, and purchase. With AI agents, the path collapses into a single, instantaneous decision, no visible click trail remains.
Early performance data indicates AI-driven traffic converts at higher rates than traditional sources. Adobe found that in July, visits to U.S. retail sites referred by AI assistants converted 60% higher than other traffic and generated 53% more revenue per visit. The greater challenge, however, remains unresolved: Can advertising even reach an agent? If the AI operates on behalf of the buyer, paying to influence its recommendations risks resembling prompt injection with a media budget.
Three approaches are emerging. OpenAI maintains ads alongside answers, labeled and, according to the company, without altering results. Amazon and Google embed sponsored products directly into their assistants, treating placements as digital shelf space. The third model has no ads at all: the buyer pays for the agent, and brands compete on data, price, and proof.
Read Also: Attar Metals cuts shredder wear with new pre-shredder
Trust and data replace ads in AI shopping
Agents will prioritize brands they can read, trust, and verify. Readability requires clean product data accessible via API or feed, not buried in scraped pages. Trust depends on consistent reviews, policies, and track records, attributes reassessed by machines. Verifiability demands first-party data proving sales occurred. When software makes purchases, budgets will shift toward brands demonstrating a direct impact on orders.
For businesses in Europe, where AI agents are already processing transactions, preparation begins with foundational steps. Align product feeds, site markup, and marketplace listings on price, stock, delivery, and returns. Track visits from AI assistants (via domains like chatgpt.com, perplexity.ai, or gemini.google.com) as a distinct channel. Developers should verify agents using standards like Visa’s Trusted Agent Protocol instead of blocking all bots. Finally, allocate a small test budget for AI placements, comparing results against a control group to measure actual impact.
The companies that prepare their data now will be the ones agents can find, trust, and purchase from, regardless of which future arrives first.
