At Shoptalk 2026, we scored dozens of established retail brands on AI readiness. The average was 2.7 out of 10. Nearly a third scored below 1.0. These are companies with ecommerce directors, real revenue, and established online stores – and most of them had no idea where they stood.
At Shoptalk 2026 in Las Vegas, we met with ecommerce directors, CTOs, and VPs of digital from dozens of retail brands. Before each meeting, we ran their products through the same AI visibility analysis we use for our customers: how their products rank across ChatGPT, Google AI Mode, and Perplexity, and how rich their structured product data is on a 0-10 scale.
We scored brands across fashion, beauty, home goods, electronics, CPG, food and beverage, and specialty retail. The results were worse than we expected.
Average AI readiness score: 2.7 out of 10. Median: 2.6. Only 12% scored above 5.0. Nearly a third scored below 1.0. About 9% scored zero.
These are not small startups. These are companies with ecommerce directors, established online stores, and real revenue. Most of them had no idea where they stood.
Two things determine whether AI agents recommend your products or your competitor’s. First, AI visibility rank: where the brand’s products appear relative to competitors when shoppers ask AI agents category-level questions like “best [product type] for [use case] under $[price].” We tested each brand against every competitor in its category. A brand ranked #1 out of 51 competitors means AI agents recommend it before anyone else in that category.
Second, AI readiness score on a 0-10 scale: how well the brand’s product data is structured for AI agent consumption. This covers the number of structured attributes per product, data completeness across the catalog, schema markup quality, and how well the data supports the kinds of decomposed sub-queries AI agents run when evaluating products.
A high rank with a low score means the brand shows up today – probably on brand recognition or third-party signals – but the position is fragile. A competitor with better structured data can displace it. A low rank with a low score means the brand is invisible and has no foundation to build from.
Average score: 2.7 / 10. The typical retailer in our sample had product data that could answer roughly a quarter of the sub-queries an AI agent would generate from a shopping request. The other three-quarters went unanswered, which means the AI either skipped the product or ranked it below competitors with more complete data.
Median score: 2.6 / 10. Half the brands in our sample scored below 2.6. This is not a case of a few outliers pulling down the average. The distribution is concentrated at the bottom.
Only 12% scored above 5.0. The brands that cleared that bar had something in common: mature product information management. They had invested in structured catalog data – not for AI, but for marketplace syndication, internal systems, or compliance. Their AI readiness was a side effect of good data hygiene.
31% scored below 1.0. Their product data consisted of a title, a price, and a paragraph of marketing copy. AI agents had almost nothing structured to evaluate.
About 9% scored exactly 0. Their products were functionally invisible to AI shopping agents across every platform we tested. One was a fashion brand. One was a specialty food retailer. One was an industrial equipment company. The common thread: no structured attributes beyond the bare minimum required to list a product online.
The brands scoring above 5.0 shared three characteristics that explain the gap. Understanding these patterns is the fastest way to diagnose where your own catalog stands.
Deep structured attributes. Their product listings had 20 to 30+ queryable fields per product. Not just title, price, brand, and color, but material composition, use case tags, compatibility information, care instructions, certifications, and dimensional data. When an AI agent decomposed a query into 8 to 12 sub-queries, these catalogs had a structured field that answered most of them. The five dimensions of AI-ready product data – attribute depth, specificity, structured format, freshness, and cross-channel consistency – were all present in the top scorers’ catalogs. Category leaders in AI recommendations consistently carry 40 or more attributes per SKU.
Feed infrastructure already in place. They were already submitting detailed product feeds to Google Merchant Center, Amazon, or marketplace partners. They had systems for maintaining feed accuracy and freshness, including data hygiene across their Merchant Center feed and real-time synchronization between their catalog source of truth and every distribution channel. When AI shopping channels emerged as a new surface, their existing feed infrastructure gave them a head start that competitors without that discipline could not replicate quickly.
Product data treated as a cross-functional asset. In these companies, product data was not owned exclusively by the ecommerce team or the marketing team. It was a shared asset maintained by PIM systems with input from merchandising, supply chain, and product development. The data was richer because more parts of the organization contributed to it.
The brands that scored below 1.0 shared a different pattern: product data was a marketing function. Descriptions were written for conversion, not for structured retrieval. Attributes beyond the basics were stored in free-text paragraphs or not captured at all. That approach worked when the audience for your product data was a human reading a product page. It does not work when the audience is an AI agent parsing structured fields to build a recommendation.
The most common reaction when we showed ecommerce directors their score was genuine surprise. Most had never measured AI readiness. It was not on their dashboard. They were tracking conversion rate, AOV, organic rankings, and paid media efficiency. AI visibility was not a metric they had.
Several told us they had noticed AI-referred traffic appearing in their analytics but did not know how to influence it. They could see sessions with referral sources like “chatgpt.com” or “google.com/ai” but had no framework for understanding why some products appeared in AI results and others did not. The traffic was there. The understanding of what drove it was not.
The most striking conversations were with brands ranked #1 in their category who had low scores. When we showed them their competitive ranking, they were pleased. When we showed them the data quality gap between their score and what we had seen from top performers in other categories, the tone shifted. They understood the position was fragile. A competitor who systematically enriches their catalog – adding the structured attributes that AI agents need to confidently recommend products – can close a ranking gap that currently looks comfortable.
The score distribution varied meaningfully by category, and the patterns reveal where the structural work is most urgent.
Fashion and apparel showed the widest score range: 0 to 6.9. The highest-scoring fashion brand had invested heavily in structured size, fit, fabric, and occasion data. The lowest had lifestyle-oriented product pages with almost no structured attributes. Fashion is a category where the gap between “looks great on the page” and “parseable by an AI agent” is especially large. A beautifully art-directed product image tells a human shopper everything. It tells an AI agent almost nothing without structured metadata behind it.
Electronics and tools clustered in two groups: brands with detailed spec sheets already structured from marketplace requirements scored 2.9 to 5.9. Brands without marketplace discipline scored 0.2 to 0.3. The spec sheet discipline that marketplace compliance enforced turned out to be the single most predictive factor of AI readiness in this category.
Home and furniture averaged below 2.0 despite being a high-intent AI shopping category. Shoppers frequently use AI agents to compare furniture options by room dimensions, material, and compatibility with existing pieces. Dimensional data, material composition, and room compatibility were rarely structured in the catalogs we reviewed – exactly the attributes AI agents need most for this category.
CPG produced both the highest single score in the sample (7.9) and some of the lowest. The high scorer was a multi-brand conglomerate with mature PIM infrastructure across its portfolio. Smaller CPG brands had minimal structured data. The gap within a single category between the best-prepared and least-prepared brands is the clearest signal of how much competitive advantage is still available to the retailers who move first.
According to Adobe Analytics data published in April 2026, AI-referred traffic to U.S. retailers grew 393% year over year in Q1 2026. Those visitors convert 42% better and spend 48% longer on site than shoppers arriving from other channels. AI-driven revenue per visit is now 37% higher than non-AI traffic – a complete reversal from just 12 months ago, when regular human traffic was worth 128% more. The channel is growing fast, and the traffic quality is the best in ecommerce.
Adobe’s same report found that roughly a third of retailers’ product pages cannot be properly accessed by AI – meaning the content is invisible to the large language models deciding which products to recommend. That finding matches what we saw in our Shoptalk scoring: the average retailer’s catalog answers only about a quarter of the sub-queries an AI agent generates from a single shopping request.
The fix is not mysterious. It is catalog enrichment: adding structured attributes that AI agents can query, maintaining feed accuracy, and implementing schema markup that makes product data machine-readable. The retailers who have already done this work – for marketplace compliance, for PIM discipline, for any reason – are benefiting from AI traffic today. The retailers who have not are invisible to a channel that is growing faster than any other in ecommerce.
For a complete tactical playbook on closing the gap, the AI search SEO playbook covers the specific schema types, feed configurations, and attribute strategies that move products into AI shortlists within 30 days.
The question for every ecommerce team reading this: what is your score? If you do not know, that is the first problem to solve. You can get your free AI visibility report at paz.ai.
An AI readiness score measures how well a brand’s product data is structured for consumption by AI shopping agents like ChatGPT, Google AI Mode, and Perplexity. The score reflects the number of queryable structured attributes per product, catalog completeness, schema markup quality, and the ability to answer the decomposed sub-queries AI agents generate when evaluating products. A score of 10 means the catalog can answer virtually every sub-query an AI agent would run. A score below 1.0 means the catalog has almost no structured data beyond a title, price, and unstructured description.
The most common root cause was that product data was treated as a marketing function rather than a structured data asset. Product descriptions were written for conversion – to appeal to human shoppers reading a page – not for structured retrieval by AI agents parsing fields. Attributes beyond the basics (title, price, color, brand) were stored in free-text paragraphs or not captured at all. This approach worked when the audience for product data was human. It does not work when the audience is an AI agent that evaluates products across 30 or more structured fields.
AI agents decompose a shopping query into 8 to 12 sub-queries and look for structured fields that answer each one. Beyond the basics, the attributes that consistently determine AI recommendation outcomes include material composition, dimensional data, use case and occasion tags, compatibility information, care instructions, certifications and standards, and comparison-relevant metrics specific to the category. Fashion products need fit and fabric data. Furniture needs room compatibility and dimensional data. Electronics need spec sheets. The target is 30 or more structured attributes per product, with category leaders carrying 40 or more.
Brands that already submit detailed product feeds to Google Merchant Center, Amazon, or other marketplace partners have a structural head start on AI readiness. The discipline required for marketplace feed compliance – clean attribute data, consistent taxonomy, accurate and fresh pricing and availability – is nearly identical to what AI shopping channels require. When AI shopping surfaces emerged as a new distribution channel, retailers with mature feed infrastructure were already positioned to benefit. Retailers without that infrastructure face both the enrichment work and the feed infrastructure build simultaneously.
Meaningful improvement is possible within 30 to 60 days for retailers who focus on their top revenue-generating SKUs first. Start with your top 20 to 50 products by revenue. Apply structured attributes across the five dimensions: attribute depth (30 or more fields), specificity (numbers and standards over adjectives), structured format (JSON-LD schema and feed attributes rather than prose), accuracy and freshness (real-time or near-real-time syncing), and cross-channel consistency. Brands that reach full attribute completion on their top SKUs typically see a 15 to 25% lift in AI recommendation frequency within 60 days, according to benchmarks from catalog enrichment programs.
