A few weeks ago, I made the case that the Learn and Use phases of the shopping journey are the no-regret bets for Holiday 2026. Consumer readiness is proven, the ROI compounds regardless of pace, and the 90-day window between July 1st and the pending holiday code-freeze in October is enough time to deploy to production. That argument still holds. What it did not have was a scorecard.
Now it does. Last week (July 15), ReFiBuy and Digital Commerce 360 published their inaugural AI Commerce Rankings, a quarterly benchmark that scores the Top 1000 retailers 0 to 100 on how ready their product catalogs are for AI shopping agents to read, interpret, and recommend. It is the first hard measurement of Learn phase readiness at scale, and it confirms both halves of the no-regret argument at once. The opportunity is real. Almost nobody has captured it yet.
The Diagnosis: 42 Out of 100
Cognizant’s Comfort Quotient research, which I cited last time, put Learn phase comfort at 47, the highest of the three purchase phases, with Early Adopters and Accelerators scoring 58 and 50 respectively. Consumers are ready. The AI Commerce Rankings show what retailers have built to meet them there. The average score across all 1,000 retailers is 42, with a median of 44.1. Only 20 retailers score above 60. The highest score recorded, out of 1,000 of the largest retailers in the country, is 72.
Scale does not buy readiness. Online Labels ranks #1 on AI readiness with a score of 71.8 and sits at #814 by online sales. Everlane ranks #4 in readiness at #264 by sales. Several of the retailers with the largest online sales volumes in the country rank in the bottom half of the readiness index. The retailers most exposed to AI-driven discovery traffic today are, in some cases, the ones AI agents can least see. Deloitte’s Stephan Ritter put a finer point on it recently on the Retailgentic podcast: most products will never be seen by AI, because most catalogs weren’t built for AI to see them.
There’s a concentration risk layered on top. ChatGPT accounts for more than 80% of AI-referred traffic to retail sites right now. Adobe Analytics clocked 693% year-over-year growth in generative AI traffic during Holiday 2025, then 393% year-over-year growth in AI-source traffic in Q1 2026. That traffic is real, it is growing fast, and most of it is flowing through a single door.
What to Build in the Next 90 Days
The Learn phase playbook is not a strategy exercise. The AI Commerce Rankings methodology is, in effect, a build spec. It scores exactly what an AI shopping agent experiences when it hits your catalog.
Bot friendliness first. Can ChatGPT, Gemini, and Alexa for Shopping actually reach, read, and transact against your product data, including support for emerging agentic commerce protocols like UCP and ACP? Only 26% of retailers in the index have verified UCP status. That’s the single highest-leverage fix available in the next 90 days for most retail IT teams, and it’s an engineering problem, not a marketing one.
Second, build for more than one engine. A single-engine dependency on ChatGPT is a single point of failure the moment a competing model changes its retrieval behavior or a new entrant like Perplexity gains share. Diversity of AI sources is one of the four signals the rankings measure for a reason.
Third, treat product content as multimodal infrastructure, not marketing copy. Cognizant’s research is direct about this: businesses will need multimodal approaches to conveying product and service information as AI-powered discovery becomes the default research path. That means structured attributes, image and video content agents can parse, and the metadata layer that turns a SKU into something an agent can reason about.
Fourth, know where your category actually sits. The rankings break out readiness by category, and the spread is wide. Office Supplies leads at 47.4. Food & Beverage trails at 37.2, the lowest score and the lowest AI traffic penetration in the index, in a category that touches nearly every consumer every week. One top-10 retailer by online sales in that category ranks below #550 on AI readiness. For CPG and grocery brands, that gap isn’t a warning. It’s the opening.
The Infrastructure Argument Hasn’t Changed
None of this ships from a slide deck and a vendor demo. It requires the same technical foundation I described in the first piece: unified product data, a semantic layer that lets an agent reason across your catalog instead of guessing, and guardrails that keep that reasoning inside the lines. AWS makes the case plainly in its own executive guidance on agentic AI. Organizations that have already operationalized generative AI with production-grade rigor, on a foundation like Amazon Bedrock, are the ones positioned to turn isolated catalog pilots into governed, reusable capability instead of another one-off integration.
That’s applied engineering work, not a repositioning exercise. It’s also the same discipline we walked through recently in how a marketing function runs on Amazon Quick with generative and agentic AI: real data, real guardrails, real production deployment. That’s what separates a working system from a slide.
90 Days, One Scorecard
The October constraints haven’t moved. The pending holiday code-freezes still lock down the production environments for most retailers and consumer brands ahead of peak season, and Amazon’s Fall Prime Event still opens the early holiday promotional calendar in early/mid-October (exact dates still pending, which makes preparation for brands even more challenging). What’s changed is that there’s now a public, quarterly, refreshed scorecard measuring exactly which retailers used this window and which ones didn’t.
An average score of 42 means the race is still open. Twenty retailers have separated from the pack. The other 980 have less than 90 days to decide which side of that line they land on before the next edition publishes.
Almost every retail and consumer goods brand I speak with agrees the opportunity is real. The AI Commerce Rankings just gave everyone a number to measure it against. If your team wants to know where your catalog stands and what to fix first, I’d like to be part of that conversation. Reach out on LinkedIn or directly.
Next up: the Use phase, and what “taking care of itself” actually means in production.
Let’s build what’s next in retail. Request an AI Briefing.
Saul Delage is SVP Client Partner at Robots & Pencils, focused on the Retail and Consumer Goods vertical. Robots & Pencils is an applied AI engineering partner, all in on AWS. Connect with Saul.
