🛒 MENA AI Shopper Readiness Benchmark

Can an AI shopper actually buy from MENA retailers?

We tested 7 leading storefronts the way a delegated shopping agent (ChatGPT, Perplexity, Claude, WebMCP clients) would — reading the server response, without executing the site's JavaScript. Public pages only; no accounts, purchases, or bot-protection bypass. Two marketplaces returned a bot-protection challenge to the server-side fetch and are noted separately. Snapshot: , generated with the live PerfObserveAR engine.

The scoreboard

Where every retailer struggles

What this means

The product isn't in the page — it's in the JavaScript. Every site here renders fine in a browser, but agents read the server response. That's where product identity, variants and delivery go missing: shipped as an empty shell and painted in by JavaScript a plain agent fetch never runs. Product Identity and Availability are the lowest-scoring dimensions for exactly this reason.
Price is the most consequential gap. Even where a price exists, it's often a placeholder 0.00 in the server HTML, or a real number buried under stacked coupons, instalments and card cashback. A human resolves this at checkout; a delegated agent cannot, so it can't quote a real total — and a shopper it can't price, it can't sell to.
This is a first-mover window. Agent-driven shopping is arriving in MENA before local storefronts are ready for it — two of the biggest even block agent-style fetches outright. The retailer that becomes machine-readable first is the one agents can actually recommend and transact with.

See where your store lands

Run the same audit on any product page — free, ~10 seconds, evidence-backed scorecard.

Audit your store →

Method. One automated fetch per storefront over a representative product page, analysing schema.org structured data, machine-readability of price/variant/delivery/returns, AI-crawler access (robots/llms.txt) and server-render vs JS-shell. It deliberately does not execute JavaScript — because most shopping agents don't either. A large HTML payload with little visible text is treated as a JS shell, and any schema on it is flagged unverified.

Fair-use note. Scores reflect one product page at one moment and describe machine-readability, not site quality for humans. A low score means "an agent would struggle here", not "this site is broken". A full audit pairs this automated pass with rendered-journey mission testing. Allbirds is included as a global reference point, not a MENA retailer.

Built in Dubai 🇦🇪 · PerfObserve · the commerce tier of PerfObserveAR.