ShoppingPartnerLab — Blog

Universal Commerce Protocol & AI Shopping Agents: The Future for CSS

2026-02-19 · Marcus AI · General

Universal Commerce Protocol & AI Shopping Agents: The Future for CSS

Universal Commerce Protocol (UCP): Google's New Standard for the AI Shopping Revolution

The world of e-commerce is on the verge of its biggest transformation since the introduction of the smartphone. Where we previously scrolled manually through rows of product advertisements, the market is now shifting towards agentic technology. Agentic technology refers to autonomous systems, in this case, AI agents, that are capable of independently making decisions and performing actions based on established goals and available data. With the introduction of Google's 'Search Generative Experience' (SGE) and the rise of these AI agents, the way transactions are made is fundamentally changing. The Universal Commerce Protocol (UCP) serves as the invisible backbone that enables AI agents to not only find products but also to autonomously make purchasing decisions based on real-time data and complex user context.

In this article, we delve deep into the technical shift from 'clicking to conversing,' where leading CSS partners such as Klarna, Sembot.com, and Channable CSS play a crucial role in this new, AI-driven auction. We highlight the necessity of UCP, its implications for retailers and consumers, and how you can prepare for these transformative changes.

The Problem: The 'Paradox of Choice' and Inefficient Checkout

Online stores are currently grappling with a growing problem: consumers are overwhelmed by the amount of data available online. Daniel Kahneman's concept of "cognitive overload" is directly applicable here; too many options do not lead to more satisfaction, but rather to indecision and frustration. A traditional Google Shopping search often yields hundreds, if not thousands, of results. The user must manually set filters, compare prices across different providers, read product reviews, check shipping costs and times, and review return policies before an informed decision can be made. This laborious and time-consuming process leads to:

  • High Bounce Rates: Consumers abandon complex comparison processes. According to research by the Baymard Institute, the average bounce rate for e-commerce websites is around 45-55%, partly due to an overly complicated decision-making process.
  • Information Fragmentation: Relevant product information is often locked in static feeds that cannot capture the dynamic context of the user. Different platforms and websites present information in various ways, making comparison difficult.
  • Friction Losses in the Conversion Funnel: Every additional step in the funnel (from clicking an ad to adding to cart, to payment and confirmation) leads to conversion loss. Research shows that every extra mandatory step in the checkout process can reduce conversion by an average of 10%. Unnecessary fields, mandatory account creation, and unclear shipping costs are common stumbling blocks.

In the old world, the 'click' was the endpoint for Google; it signaled that the user had found the desired information and would potentially proceed to the advertiser's website. In the new AI mode, the click is merely an intermediate step towards a fully completed transaction, executed by an AI agent on behalf of the user, which requires a much higher degree of automation and efficiency.

The Solution: Universal Commerce Protocol (UCP) and Agentic Shopping

The solution lies in facilitating AI agents that act on behalf of the consumer. This concept is known as 'agentic shopping.' To this end, Google is introducing conversational shopping in the AI mode of Google Search, also known as the Google Search Generative Experience (SGE). Instead of entering specific search terms, the user can describe a complex scenario in natural language, for example: "Find a sustainable winter coat for a trek in Norway next year, with a budget of no more than €300, suitable for my height of 1.80m and weight of 75kg, and deliverable by Friday in two weeks."

AI agents use the Universal Commerce Protocol to retrieve structured data that goes beyond just price and basic descriptions. UCP is an advanced set of standards that includes not only product characteristics but also dynamic, real-time context. Consider:

  • Current stock levels: Preventing an AI from recommending a product that is out of stock.
  • Locally applicable laws and regulations: Crucial for products such as supplements, cosmetics, or electronics, where rules can vary greatly by country or even region. The mention of new regulations for supplements in France and Germany from November 2025 is a perfect example of data types that an AI agent must be able to interpret and apply.
  • Real-time shipping times and costs: Essential for urgent orders or international shipments.
  • Return policy and warranty conditions: Important for building trust and reducing uncertainty.
  • Sustainability scores and certifications: Increasingly important for consumers who want to make conscious choices.
  • Compatibility information: For technology products, for example, whether an accessory fits a specific model.

This in-depth data enables the AI agent to not only find the *best* product based on the user's explicit search query but also based on implicit preferences and contextual factors.

CSS advantage remains

A crucial aspect for advertisers is that, even in this AI-driven auction, the 20% discount on CPC bids from the EU's €2.42 billion competition case remains in effect. This discount regime, which arose from the antitrust case in which Google was found guilty of abusing its dominant position to the detriment of competing comparison websites, encourages open competition and lower advertising costs. CSS partners such as shopping24, IdoSell Shopping, and Shopmos act as the data gateways that feed these AI agents with the most accurate, technically optimized feeds. By using a CSS partner, advertisers not only ensure that their products are visible but also that they do so at a more favorable rate, thereby maximizing their Return on Ad Spend (ROAS). They are not just technical links but strategic partners in unlocking this AI-driven marketplace.

Comparison: Traditional Shopping vs. AI Agent Shopping

The shift is fundamental and can best be illustrated by a direct comparison:

| Feature | Traditional Google Shopping | AI Agent Shopping (UCP Era) | | :--- | :--- | :--- | | Input | Keywords. User must specify what they are looking for. | Context & Intent (Natural Language). AI interprets complex queries and context. | | Decision Maker | The human user. Makes the final purchase decision after comparison. | The AI agent (on behalf of the user). Autonomously executes transactions based on criteria. | | Data Processing | Static XML feeds. Updates are often periodic, less dynamic. | Real-time API / UCP Protocol. Direct, up-to-date information via APIs. | | Remuneration | Cost Per Click (CPC). Payment per click, regardless of conversion. | Mixed (CPC + Transactional Friction Reduction). Potentially hybrid models, value of reduced friction is recognized. | | CSS Role | Traffic Channel & Savings. Ads are displayed cheaper via CSS. | Data Validator & Return Optimizer. Ensures optimal, AI-interpretable data. | | Goal | Display products that match keywords. | Find the most suitable solution and complete the transaction. | | Decision Complexity | Low to medium. User filters themselves. | High. AI weighs multiple criteria simultaneously, including implicit preferences. | | Personalization | Limited, based on search history and cookies. | Highly personalized, based on a deep understanding of individual needs. | | Speed | Relatively slow, dependent on human decision-making. | Very fast, transactions can be completed in seconds. |

Practical Use Cases

The implementation of UCP and AI agents opens up a range of possibilities for both consumers and retailers:

Scenario 1: The Specific Consumer with Time Pressure

A user asks Google AI: "I have a wedding in Amsterdam tomorrow and my shoe size is 44. Which black, leather lace-up shoes of high quality can I pick up today at a store near Dam Square, or have delivered tomorrow morning?"

The AI here looks not only at the lowest price or the widest online availability but at omnichannel availability and specific logistical requirements. Via the UCP, the AI agent retrieves data on: 1. Product availability: Are there size 44 black leather lace-up shoes in stock at physical stores in Amsterdam Center, and if so, which ones? 2. Pickup/Delivery options: Do these stores offer 'click & collect,' or fast ('same-day'/'next-day') delivery in Amsterdam? 3. Quality criteria: By enriching attributes (leather type, brand perception, customer reviews), the AI can interpret 'high quality.'

A partner like Klarna Comparison Shopping Service can directly facilitate the checkout here by processing the payment on behalf of the AI agent, forwarding the order to the chosen store, and notifying the user of the pickup or delivery details. This significantly reduces friction; the consumer no longer needs to visit websites.

Scenario 2: Dynamic Pricing Strategy via CSS and AI Optimization

Imagine a scenario where an online retailer uses a leading CSS partner like Sembot.com to adjust feeds in real-time. This retailer has an advanced pricing algorithm that takes into account competition, stock levels, and margin requirements. When the consumer's AI agent compares prices in the SGE, the agent recognizes that this shop, thanks to dynamic feed optimization via Sembot, not only has a competitive price but also meets all other criteria (fast delivery time, good sustainability score, excellent customer service). The consumer's AI agent receives a complete picture of the offer via UCP. This results in a higher 'agent preference,' leading to more recommendations and ultimately more sales for the retailer. This is an example of how CSS partners evolve from mere traffic sellers to strategic optimization partners.

Scenario 3: Niche Products and Complex Specifications

An architect is looking for a specific type of light fixture for a project: "I need a dimmable LED strip, 5 meters long, with a color temperature of 2700K-3000K, CRI > 90, IP65 certified, suitable for outdoor use, and compatible with a Philips Hue Bridge."

Previously, this would lead to hours of searching technical specification sheets. With UCP, AI agents can directly match these complex requirements with enriched product data from various suppliers. The AI can even analyze installation manuals and compatibility lists made available via UCP to find the perfect match.

Step-by-step: Preparing your shop for AI agents

The transition to an AI-driven commercial landscape requires a proactive approach. Ignoring these developments will lead to reduced visibility and competitiveness.

1. Optimize your Merchant Center According to the 2025 Guidelines: Ensure that all your product information, including logos, company names, VAT numbers, and contact details, is meticulously correct and up-to-date at the campaign level in Performance Max. Google regularly announces updates to Merchant Center specifications, often aimed at improving data quality and facilitating advanced search capabilities. The 2025 guidelines will likely place an even greater emphasis on the robustness and granularity of product data to meet the needs of AI agents. Correctly setting up company information is crucial for the 'trust score' that an AI agent will assign. 2. Enrich your Attributes: This is the heart of agentic data. Add more than the standard fields `title`, `description`, `price`, and `image_link`. Think of detailed `material` specifications (e.g., "recycled polyester," "organic cotton"), `sustainability_id` (referring to certifications such as GOTS, Fair Trade), specific `size_type` (e.g., "slim fit," "oversized"), `color_family` (e.g., "dark blue," "pastel green"), `occasion` (e.g., "casual," "formal wear"), and detailed `shipping_label` data that not only says "free shipping" but also precise delivery times per region and options for expedited delivery. The more structured, semantically rich data you offer, the better the AI agent can match your product with complex user intentions. Consider unique product identifiers (GTINs, MPNs) that are consistent across all channels. 3. Choose an AI-driven CSS partner: The choice of your CSS partner is more strategic than ever. Work with partners like Channable CSS or Booncy who not only optimize your feeds for the regular Google Shopping auction but also support API integrations instead of just static feeds. API-based integrations enable real-time data exchange, which is essential for UCP. These partners can help you with: * Data Transformation: Converting raw product data to the UCP format. * Attribute Mapping: Correctly mapping your internal product attributes to the required UCP specifications. * Dynamic Feed Updates: Ensuring up-to-date stock, prices, and shipping information in real-time. * AI Interpretation Optimization: Advice on how to best present your data so that AI agents can interpret it effectively. 4. Monitor Auction Insights and AI-specific Reports: Use the auction insights report in Google Ads and any new reports that Google will launch for AI mode. These reports will provide insight into how you are performing compared to other advertisers specifically in the AI-driven environment. Analyze which attributes and descriptions lead to successful recommendations by the AI. Continuously adjust your feed optimization based on these insights. This is an iterative process; AI behavior evolves, so your optimization must too.

Frequently Asked Questions (FAQ)

What exactly is the Universal Commerce Protocol and why is it needed? The Universal Commerce Protocol (UCP) is an advanced set of technical standards and protocols, essentially an extension of existing standards such as Schema.org and the Google Merchant Center Product Data Specification, specifically designed to enable AI agents to universally understand product data and autonomously initiate transactions. It is needed because traditional product feeds and web pages are not structured enough to provide the contextual nuances and dynamic information that AI agents need to make truly intelligent purchasing decisions. UCP solves data fragmentation and ensures a consistent, machine-readable interpretation of complex product information, including real-time stock, personalized prices, and regulations.

Will the 20% CSS discount remain with AI Shopping and how does that work in practice? Yes, the structural separation between Google Shopping and other Comparison Shopping Services remains, stemming from EU competition rules. This means that when an AI agent makes a product recommendation and the final purchase is facilitated through a CSS partner (such as Shopmos), the advertiser still benefits from the more favorable auction conditions – i.e., the 20% discount on CPC bids. This is because the AI agent essentially performs a structured search query that results in a selection of providers that meet the criteria, with the CSS partner emerging as the 'advertising entity' in the auction. The AI optimizes for the best deal for the user, and the CSS structure helps price that best deal competitively.

How does the Search Generative Experience (SGE) affect my CTR and conversion? Research and early tests suggest that the Click-Through Rate (CTR) to individual websites in the SGE can potentially change. Users receive direct answers from the AI, which may mean fewer people click directly to a website for basic information. However, the quality of traffic that *does* click through is expected to increase significantly. The AI agent filters out 'window shoppers' and people with low purchase intent, as it has already performed a large part of the comparison process for the user. The clicks you receive will therefore come from users with a much higher purchase intent, who are closer to making a purchase, which can lead to a lower CTR but a higher conversion rate and a more efficient use of your advertising budget.

Do I need to adjust my feeds for AI agents and if so, which are the most critical? Absolutely. Adjusting your feeds is not optional, but essential for survival and growth in the era of agentic shopping. AI agents rely on 'agentic data,' meaning that the data must not only be available but also usable and interpretable for autonomous decision-making. The most critical adjustments are: * Structured Attributes: More detailed and standardized attributes such as `size_system`, `pattern`, `material_composition`, `features` (e.g., "waterproof," "breathable"), `sustainability_certifications`. * Real-time availability: Precise and up-to-date `in_stock` status including `fulfillment_options` (in-store pickup, delivery options). * Regulatory Information: Any `disclaimers`, `allergen information`, `warnings`, or `licenses` relevant to the product and region. * Use Cases: Explicitly describe for which `purpose` or `occasion` a product is suitable. * Linked Data: Linking to relevant `reviews`, `user_manuals`, `video_demonstrations`, and `compatibility_lists`. The more structured information about fit, use cases, sustainability scores, and real-time availability you offer, the greater the chance that the AI will recommend your product as the most suitable solution for the user's (complex) query.

Conclusion

The transition to Universal Commerce and AI agents is no longer a distant future but a rapidly approaching reality that will become the standard for online retail in 2025. This shift is as profound as the one from desktop to mobile. In this era, CSS partners, who bridge the gap between the complexity of product data and the demands of AI, will be the winners. The companies that proactively optimize their data infrastructure and product feeds for AI interpretation will be the ones that flourish in a market where the machine makes decisions on behalf of the consumer. Ensure that your products are not only 'discoverable' but especially 'readable' and 'actionable' for Google's AI mode. It's time to rethink your e-commerce strategy and prepare for the next generation of digital shopping.

Want to know if your feed is ready for the AI revolution? Let us show you how we optimize your data environment for the future and discover your savings potential!