Kiwop Labs · tool

Is your store ready for AI?

Type the domain of an online store and in a minute you get a report with what an AI assistant sees when it arrives: whether it can get in, whether there is a real llms.txt, whether the product page exposes a price and stock it can read. Same method as our study of 300 Spanish stores.

We only read what is public: robots.txt, llms.txt, the home page and one product page. No login, no purchase. The report is public and has its own URL.

How it works

  1. An identified bot (KiwopResearchBot) requests robots.txt, llms.txt, the home page and one product page of your store. If the firewall throws it out, it retries with a sessionless Chromium browser, the way ChatGPT does when it browses.
  2. We count the same things as in the study: which AI crawlers are locked out, whether the llms.txt was written by the store or served by the platform, what structured data the home page carries and whether the product page has Product and Offer.
  3. The score comes from a public, versioned weight table. Changing a weight means changing the version: each report records which one it was measured with.
  4. The report is published with its own URL and a downloadable JSON. After seven days it can be measured again.

What we measure and how much it weighs

Seven checks, one hundred points. The weights reflect what changes an assistant’s answer the most: being able to get in and finding a readable product.

CheckPointsWhat we look at
AI crawlers25robots.txt against 11 AI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended…). A full block subtracts in full; an explicit rule disallowing paths only for that bot subtracts half; disallowing the cart for everyone is not closing the door.
Access for an identified bot10Whether the home page answers a bot that identifies itself. If it only answers a browser, an assistant crawling without a session stays out.
llms.txt15Whether it exists and who wrote it: the store’s own text scores fully; the platform’s automatic template scores half.
Structured data on the home page10JSON-LD on the home page with Organization, WebSite or OnlineStore. A broken block subtracts points.
Readable product page25We locate a product page via the sitemap or the home page links and look for JSON-LD Product with Offer (price and availability). It weighs the most: without it, the agent does not know what you sell.
Declared sitemap10Sitemap declared in robots.txt. It is the map crawlers use to discover the catalogue.
HTTPS5The final URL is served over HTTPS.
Total100

The three verdicts

Ready 80 points or more. An assistant gets in, understands who you are and finds products with prices.
Halfway Between 55 and 79. It gets in, but an important piece is missing: almost always the readable product page or the llms.txt.
Invisible to AI Under 55. Either it cannot get in, or it gets in and finds nothing it can use.

Latest reports

StorePointsVerdictMeasured on
tradeinn.com 74 Halfway 13 Sept 2026
mediamarkt.es 93 Ready 13 Sept 2026
sprintersports.com 97 Ready 13 Sept 2026
parfois.com 91 Ready 13 Sept 2026
generaloptica.es 95 Ready 13 Sept 2026
zalando.es 29 Invisible to AI 13 Sept 2026
yesstyle.com 70 Halfway 13 Sept 2026
zalando-prive.es 44 Invisible to AI 13 Sept 2026
womensecret.com 75 Halfway 13 Sept 2026
webuy.com 59 Halfway 13 Sept 2026
sklum.com 78 Halfway 13 Sept 2026
veepee.es 48 Invisible to AI 13 Sept 2026
singularu.com 78 Halfway 13 Sept 2026
samsung.com 87 Ready 13 Sept 2026
puma.com 58 Halfway 13 Sept 2026
playstation.com 82 Ready 13 Sept 2026
panini.es 57 Halfway 13 Sept 2026
pedrodelhierro.com 75 Halfway 13 Sept 2026
outlet-pc.es 81 Ready 13 Sept 2026
myspringfield.com 75 Halfway 13 Sept 2026

Where the method comes from

It is the same measurer as the study “AI in Spanish ecommerce” (300 stores): there, 10.7% block at least one AI crawler, 26% have a verified llms.txt and only 27.8% of product pages expose Product schema. Each report compares your store against that sample.

Read the full study

Method

  • Two passes per signal: HTTP with an identified User-Agent and, if it does not answer, headless Chromium with a browser UA. robots.txt and llms.txt always over plain HTTP.
  • We never impersonate GPTBot or any third-party crawler. If robots.txt says one thing and the firewall does another, the report says so.
  • One product page per store, the first we find via the sitemap or the home page links. If your catalogue has pages with and without schema, the report sees one of them.
  • No login, no cart, no purchases. Only what anyone entering through the front door can see.

Limits

  • Detecting a chatbot or search engine by HTML signatures is indicative: it does not tell an AI chatbot from a rules-based one.
  • The platform is inferred from technical signatures. A heavily customised theme can hide it from us.
  • A result is kept for seven days. If you fix something today, tomorrow the report stays the same until you measure again.
  • It is an outside snapshot. It does not measure content quality or an agent’s buying experience, which is the agentic part of the study.

Public reports

Everything we measure is public on your website and so is the report: it has its own URL and appears in the list. If you own the store and do not want it listed, write to [email protected] and we will remove it.

← Back to Kiwop Labs

Let's talk.

Initial technical consultation

AI, security and performance. Diagnosis with phased proposal.

  • NDA available
  • Response <24h
  • Phased proposal

Your first meeting is with a Solutions Architect, not a salesperson.

Request diagnosis