Learning Center

Agent readiness statistics

Figures from our own research into how well websites work for AI agents. Each one links to its own line, so you can cite exactly the figure you use, with the method behind it.

Online stores: 255 sites, October 2026

600 websites drawn at random from the online shops listed in Wikidata, so the sample leans toward established retailers from many countries. 467 could be scanned, and 255 were stores, with products or a cart. Scored with AgentScore 2026.10-v8. Read the study or see the benchmarks.

  • 80 out of 100 was the median AgentScore of 255 online stores.

    #
  • 51% of online stores scored 80 or more out of 100.

    #
  • 21% of online stores had a failure serious enough to cap their AgentScore, such as blocked agents, prices that disagree or a buy button agents can't press.

    #
  • 92% was the median store's share of the Access points.

    #
  • 89% was the median store's share of the Readability points.

    #
  • 72% was the median store's share of the Navigability points.

    #
  • 70% of online stores fell short on "Cart and checkout fields are labelled for agents" where it was checked. How to fix it

    #
  • 67% of online stores fell short on "Clickable things are real buttons and links" where it was checked. How to fix it

    #
  • 58% of online stores fell short on "Buttons and links have names" where it was checked. How to fix it

    #
  • 58% of online stores fell short on "Form fields are labelled" where it was checked. How to fix it

    #
  • 57% of online stores fell short on "Clear page title, description, and headings" where it was checked. How to fix it

    #
  • 51% of online stores fell short on "llms.txt guide for AI" where it was checked. How to fix it

    #
  • 41% of online stores fell short on "Page has main and navigation landmarks" where it was checked. How to fix it

    #
  • 38% of online stores fell short on "Pop-ups and banners can be dismissed by agents" where it was checked. How to fix it

    #
  • 37% of online stores fell short on "Product pages give price and stock in a form agents can read" where it was checked. How to fix it

    #
  • 28% of online stores fell short on "Price, stock and name agree everywhere agents read them" where it was checked. How to fix it

    #
  • 77% vs 18% of the top and bottom quarter of online stores passed "Form fields are labelled". How to fix it

    #
  • 71% vs 16% of the top and bottom quarter of online stores passed "Buttons and links have names". How to fix it

    #
  • 78% vs 27% of the top and bottom quarter of online stores passed "llms.txt guide for AI". How to fix it

    #
  • 81% vs 41% of the top and bottom quarter of online stores passed "Page has main and navigation landmarks". How to fix it

    #
  • 57% vs 19% of the top and bottom quarter of online stores passed "Clickable things are real buttons and links". How to fix it

    #
  • 84% vs 51% of the top and bottom quarter of online stores passed "Product pages give price and stock in a form agents can read". How to fix it

    #

How to cite these figures

Ghost Agent Lab, AgentScore benchmark of 255 online stores, scanned October 20, 2026 (scoring 2026.10-v8). https://ghostagentlab.com/blog/agentscore-store-study-2026/

How we measure

Every site gets the same automated AgentScore scan, which reads public pages as a browser and as AI assistants do. It never fills in a form or places an order. We publish results only in aggregate and never name a site. A figure that says "where it was checked" counts only the sites where that check could run.

Scores from different scoring versions aren't directly comparable. See how AgentScore is scored. To work with the numbers yourself, download them from open data.