The Fields You’ve Been Ignoring Are Now the Ones That Matter

Mark Batson, Head of GTM Technical Operations at Athos Commerce, put it this way: “If traditional SEO practices, when applied to a feed, are about your product being found and discovered, then AEO and GEO optimization is about getting your products selected and recommended.” GEO doesn’t require new infrastructure. It requires treating previously optional fields as first-class contract terms.

Start with product highlights and product details in Google Merchant Center. These fields have existed for years, but most merchants either leave them empty or populate them with the same keyword-optimized copy they use in titles and descriptions. That approach misses what these fields are for. Product highlights are where Google looks for intent signals, the contextual metadata that helps an AI agent recommend your product for a specific use case rather than returning a generic match.

The difference looks like this. A compliance-level product highlight might say “Men’s hoodie, charcoal grey, size L.” An intent-level product highlight would say “Heavyweight breathable hoodie for cool-weather hiking, relaxed fit through the shoulders.” The compliance version tells an agent what the product is, but an AI shopping agent answering the query “What’s a good hoodie for hiking in cool weather?” needs to know what the product is for. Intent-level highlights answer that question directly.

Product highlights and product details are becoming even more important in the era of agentic commerce, because those are the attributes that Google is going to be prioritizing to give indications of intent and suitability.

During the webinar, Brown demonstrated how Athos Commerce's generative AI can extract product highlights from existing data. Using only the title and description of a dining table, the tool generated six highlight values: seating capacity, tabletop material, dimensions, assembly difficulty, weight capacity, and room size suitability. Each of those values gives an AI agent a specific signal to match against a shopper’s query. The extraction took seconds and required no manual input beyond the product data already in the feed.

Athos ran an A/B test around Valentine’s Day that focused on layering intent-driven data cues into product highlights for a Google Merchant Center feed. The results were significant: a 120% increase in clicks for the change set with populated highlights. Impressions saw a more modest uplift, but the products with intent-optimized highlights generated far more clicks, suggesting they were surfaced more prominently across Google’s shopping surfaces. As Batson put it: “It fundamentally boils down to the difference between a product that has been optimized to speak directly to those intent challenges and being surfaced above all of those other products.”

Athos Commerce’s Performance Connector now imports impressions from Google AI Mode and Gemini alongside traditional Shopping data so that merchants can measure the impact of enrichment across AI surfaces. These gains compound because the same enriched data appears across every AI-powered surface that draws from your Merchant Center feed.

The Other Clauses in the Contract

Enriching product highlights gets you recommended, but the contract has more clauses that agents verify before they complete a transaction. AI agents cross-check your feed data against your website, and any mismatch breaks trust.

Price alignment is the most critical. When your feed lists a product at $49.99 but your website shows $54.99, an AI agent won’t display a corrected price or flag a warning. It drops your product from consideration because the discrepancy signals unreliable data. Google is encouraging merchants to adopt the Merchant API to keep feed and site data in sync, and Brown emphasized the stakes: “If the data isn’t accurate, relevant, and ready for action, the agents simply won’t risk the transaction.”

Beyond price accuracy, agents check the data they need to complete a purchase without human intervention. Free shipping annotations, shipping speed per product, and return policy data tell an agent whether it can complete a Google Shopping transaction with confidence. When an agent is comparing similar products from three retailers, the one with explicit trust data wins because the agent can commit to a purchase without sending the shopper to verify shipping cost or check a return window. We covered the full Trust Stack framework in Your Google Shopping Feed is Already Powering ChatGPT, and each of these signals is worth auditing alongside your enrichment work.

Review data is part of the same trust equation. AI agents perform sentiment synthesis, reading the actual text of your reviews to answer specific shopper queries like “Does this jacket run small?” or “Is this table sturdy enough for kids?” Feeds without integrated review data pay what the Athos team calls a “trust tax” in lower visibility because the agent sources social proof from Reddit or third-party blogs instead, and you lose control of the narrative. Three review attributes matter most for GEO readiness: your aggregate product rating, total review count (agents trust 500 reviews far more than 5 reviews), and whether reviews include a verified purchase flag. Including product ratings in your feed drives a 5% increase in click-through rate, according to a Google MerchX case study.

A Practical Sequence for Getting GEO-Ready

You don’t need to overhaul your entire feed operation at once. The better approach is to identify which contract terms you’re failing and address them in order of impact.

Start with an audit. Before you can fix anything, you need to know where the problems are. Athos Commerce's AI-powered feed audit scores every attribute across your entire catalog, ranks issues by their impact on visibility, and generates AI-recommended fixes at scale. If you don’t have a dedicated feed team, this is where Athos can do the heavy lifting, identifying and prioritizing the work so you can focus on the changes that move the needle. In one recent audit, 85% of products had titles too short for AI discoverability. When that much of your catalog is invisible to AI agents, feed enrichment stops being a nice-to-have project.

Next, enrich product highlights on a test set. Pick 50 to 100 products, populate their highlights with intent-driven language, and run the test for two to four weeks. Measure changes in CTR and prominence across Google Shopping and AI Mode through your feed management platform’s reporting. This test requires no new integrations or budget approval, and based on the Valentine’s Day results, you should see measurable changes within weeks. For a pair of waterproof hiking boots, that means replacing “men’s boot, black, size 10” with “waterproof leather hiking boot for wet-weather trail use, cushioned ankle support, Vibram sole.” Brown told a similar story in the same session about searching for a World Book Day costume for her son, where the products that included fit and suitability data were the only ones she could evaluate with confidence.

Then fill the trust signals. Add annotations for shipping speed, return policy, and free shipping at the product level. If you have verified reviews, integrate them into your feed. If you don’t have reviews on your site at all, make review collection a near-term priority.

After that, expand your feed footprint. Set up a Bing product feed, which powers Microsoft Copilot and serves as ChatGPT’s secondary data source. Register for the ChatGPT Merchant Dashboard beta if you’re not on Shopify. If you currently submit your Google feed as a CSV, TXT, or XML file, investigate migrating to the Google Merchant API for programmatic, near-real-time data sync.

Ongoing, test and measure. Run A/B tests at the attribute level. Monitor performance across Google Shopping, AI Mode, and Gemini. Make sure your website content matches your feed data so that crawlers and your feed tell the same story about every product.

The Contract Is Already Being Enforced

AI agents are evaluating your product feed today, whether you’ve prepared for it or not. Seventy-seven percent of mobile searches already end without a click to any website. AI recommendations increasingly guide the shoppers who do click, and the merchants whose feeds meet the contract terms are the ones those agents recommend.

The terms of this contract weren’t negotiated. They were set by the AI agents your customers are already using. Your only choice is whether to meet them. As Batson put it at the close of the webinar: “There is going to be no better time than to do this now, and that learning curve is only going to grow from this point forward.”

If you want to see where your product data stands, Athos Commerce, an intelligent discovery and feed management platform, can run an AI-powered feed audit that scores your catalog across every attribute and identifies the highest-impact opportunities for enrichment. Watch the full webinar for live demos of the audit tool and the product data extraction capabilities discussed in this article.