Khanyo research · published 24 October 2026
More than 1 in 4 South African online shops don’t tell AI assistants what they sell or what it costs
Khanyo checked 70 South African online shops. 27% have no product data at all on their product pages, and 31% have at least one problem that stops AI assistants like ChatGPT recommending their products. On WooCommerce and custom-built shops it is more than 1 in 3.
More shoppers now start by asking ChatGPT, Claude, Perplexity or Google’s AI what to buy. Those assistants can only recommend what they can reach and read. We checked each shop the way AI crawlers do: its robots.txt rules, its pages without JavaScript, and its product data. We did not name or contact any shop.
What we found
| Finding | South African shops (n = 70) |
|---|---|
| A real problem (any of the below) | 22 (31%) |
| No product data at all on the product page | 19 (27%) |
| Product data missing price or stock | 1 (1%) |
| Product pages nearly empty without JavaScript | 3 (4%) |
| Partly visible (some AI agents blocked in robots.txt) | 4 (6%) |
| Not visible at all | 0 |
By platform
| Platform | Shops | With a real problem |
|---|---|---|
| WooCommerce | 28 | 10 (36%) |
| Shopify | 25 | 4 (16%) |
| Other or custom-built | 14 | 7 (50%) |
Magento (3 shops) left out: too few to quote. Small samples per platform: read these as “about”.
What it means
Blocking was rarer than expected. The bigger issue is that many shops never describe their products in a form AI can read: no machine-readable name, price or stock, so an assistant can’t recommend the product with confidence. Shopify writes this data automatically, which is why its shops did best; WooCommerce and custom-built shops are where it’s missing.
The fix is usually small: a product-data template or plugin setting, or a few lines in robots.txt. Check your own shop free, in about a minute, no sign-up.
Method
Shops came from the Tranco ranking of popular websites and the public Common Crawl index (.za domains with a product page), each counted once by its latest check, between 21 and 24 October 2026. Sites that didn’t load or had no product page are left out. Checks identified themselves as Khanyo-Research and obeyed each site’s robots.txt (how we research). Server and firewall blocks were not tested, so the true share of hidden shops is likely higher. Blocking AI training crawlers is a shop’s own choice and isn’t counted as a problem. Live, updated figures: the state of AI visibility. You may quote this report with a link to this page.