Teardowns, benchmarks and field notes on Agentic Commerce Readiness — how stores get seen, compared and recommended by AI buyers.
A shopping agent shortlisting twenty parkas does not download twenty hero images. It reads the text around them. For visual categories — fashion, furniture, jewelry — the information your photos carry has to be translated into machine-readable fields, or the evaluator never sees what makes your product worth recommending.
Read →The most expensive structured data is the kind that is present but wrong. When a price or stock status changes and the JSON-LD does not, AI buyers quote the old number verbatim — and the shopper blames the store, not the model. Three fields stop the drift.
Read →An AI assistant cites your rating before it names a feature, but it does not read reviews the way a person does — it reads the structured rating block and the few snippets retrieval surfaces. Here is how the evidence layer works, the three ways stores break their own proof, and how to make your reviews quotable.
Read →Eleven rescans of four stores: one store moved 68 → 84 → 87 while its structured-data score held at 91 on every run, another gained 12 points in 21 hours when llms.txt and agents.md appeared, and an unchanged store reproduced 13/100 four times in six minutes. A readiness score is a snapshot — the three layers move at three speeds.
Read →Four AI labs shipped browser agents that complete purchases — Anthropic's computer use, OpenAI's Operator, Amazon's Nova Act, Perplexity's Comet. Most checkout flows were built for human hands, and the walls that stop fraud also stop the agent your shopper sent. Here is what breaks after the agent already chose you.
Read →We ran 120 live ecommerce stores through the full AI-readiness pipeline. Only 4 produced a product record complete enough for an AI shopping agent to compare confidently. Readiness turned out to be almost entirely a platform artifact, not a merchant decision.
Read →Three checks that tell you whether ChatGPT, Gemini, and Perplexity can read your product data and recommend your store. No tools required beyond a browser and a text editor.
Read →When an AI buyer like ChatGPT evaluates your store, it does not browse. It calls your product URL and reads the structured data it returns. If the response is empty, your product does not make the list — no matter how good the page looks to a human.
Read →Your Google Merchant Center feed powers Google Shopping — and nothing else. ChatGPT, Gemini, Perplexity, and Claude read the structured data on your product pages, a completely separate pipeline. Merchants who optimize only for GMC are invisible to the AI agents that now drive a growing share of purchase decisions.
Read →In our June–July scans, 37% of stores scored zero on structured data, averaging 23/100 overall. The 19 stores with at least some JSON-LD averaged 83. That 60-point gap is the difference between being recommended by an AI buyer and being invisible.
Read →AI crawlers check for a file at your domain root that most e-commerce stores do not have. Here is what llms.txt is, what goes in it, and why its absence leaves an AI buyer guessing about which pages matter.
Read →In our June scans, 24 of 30 stores skipped the AI preference dimension entirely, defaulting to a neutral score. The 6 that tested it saw results ranging from 40 to 100 — a 24-point readiness swing from one dimension. The heaviest-weighted part of the score is the one merchants measure least.
Read →We scanned 20 e-commerce stores in late June. Among those with solid product structured data, stores with weak crawlability scored 15 points lower on average — a gap that costs them AI buyer recommendations despite having the hardest part already done.
Read →We scanned 8 e-commerce stores that ship valid product structured data. Only 2 included return or shipping policy in a form an AI buyer can read. The rest — stores scoring up to 91 on product schema — scored zero on the dimension that decides the outcome when products are otherwise equal.
Read →We scanned 7 e-commerce stores for AI buyer readiness. Every Shopify store had complete product structured data by default. Only 1 of 4 custom-platform stores did — and the gap explains most of the readiness score.
Read →We scanned 8 e-commerce stores and found a clean split — half have solid structured data and score 60+, the other half have none at all. There is no middle ground on what matters most to AI buyers.
Read →AI buyers build their shortlist from your product page's data layer, not its design. Here are the eight JSON-LD fields that decide whether ChatGPT, Gemini, and Perplexity can recommend your products — and the three mistakes that get you dropped.
Read →AI buyers like ChatGPT, Gemini and Perplexity read your store before a human ever does. Here is what they look for, and why structured data decides whether your products get recommended.
Read →