How we score AI-readiness
Two complementary scores. Both are observable and reproducible — anyone can re-run them on their own store.
What we measure (23 checks)
Each check contributes points to the synthetic Agent-Readiness Score (0–100). Total possible: 32 raw points, normalized to 100.
Reachability & transport
Structured data presence
Product schema completeness
Product identifier
Machine-readable policies
Proprietary MerchantStamp protocol (weight 0 per agent)
How per-agent visibility is computed
For each of the 5 AI shopping agents, we apply a weight in [0..1] to each check. Per-agent score = weighted average of passing checks ÷ sum of weights. Verdict: Readable (≥75%), Partial (≥40%), Invisible (<40%).
Proprietary MerchantStamp checks (manifest, Verifiable Credential) carry weight 0 in per-agent scoring — no public AI agent reads them today. They only contribute to the synthetic Agent-Readiness Score.
The 5 AI shopping agents we report on
- 🟢ChatGPT Shopping — heavy weights on GTIN, price, currency, JSON-LD per OpenAI Product Feed Spec
- 🔵Perplexity Shopping — Schema.org Product + Offer + Review + AggregateRating flagged as must-haves; GTIN required
- 🔴Google AI Overview — structured data correlates with +73% selection rate; multi-modal +156%
- 🟠Claude (Anthropic) — per-bot robots.txt directives are the documented control surface; no commerce-specific spec yet
- 🟣Microsoft Copilot — rides on Bing index; Schema.org and TLS table-stakes
Sources
Per-check weights are anchored to public documentation from each agent owner where available, and explicitly marked as heuristic where not.
- OpenAI — Product Feed Spec (Agentic Commerce)
- Perplexity Shopping — Merchant Setup Guide (Alhena)
- Perplexity Shopping — Optimization Guide (Shopify)
- Google AI Overview — Ranking Signals 2026
- Google AI Overviews — Ranking Factors (Wellows)
- Anthropic — Crawler Documentation (ClaudeBot, Claude-User, Claude-SearchBot)
- Search Engine Land — Anthropic Crawler Update