AI e-commerce readiness
When it is no longer only people reading your store
AI answer engines cite sources they can read cleanly. Most stores are not ready for that: content appears only after JavaScript, product data is unstructured, and there is no machine-readable version of the page. This site is the exception, and you can check that right now.
Check it against this page
Four things this site ships that almost no store has
Not a concept paper: every item here runs on shoplab.cc and is verifiable with one request. We only recommend what we operate ourselves.
A markdown version of every page
Append .md to any URL, or request the same URL with the header Accept: text/markdown. You get the same content without layout, navigation or scripts. That is the form in which a language model reliably understands a page instead of reconstructing it from HTML.
A table of contents for machines
/llms.txt and /llms-full.txt describe what this site offers and where it lives, alongside a skills index at /.well-known/agent-skills/index.json and a catalog at /.well-known/api-catalog. An agent does not have to guess what it will find here.
An MCP server, not just a file
A real Model Context Protocol server runs at /mcp, read-only and without authentication, so an agent can query the site instead of scraping it. We are not aware of another German-market Shopify agency shipping one.
Content that is there without JavaScript
Every line of copy is in the initial HTML, including inside sliders and accordions, with structured data for the organisation, pages, FAQs and articles. This is the unglamorous foundation without which the rest achieves nothing, and the most common reason a store never gets cited.
How we do it for you
From audit to a page worth citing
The effort is almost never in shipping the files. It is in ordering the content so a machine understands it without interpreting.
See what crawlers get today
We fetch your store the way an AI crawler does: no JavaScript, no cookies. Usually more is missing than expected, and sometimes the conclusion is that rendering and structured data come first and the rest can wait.
Structure the content
Write and mark up product and category copy so a single passage stands on its own. AI answers cite paragraphs, not pages, and a paragraph that needs three paragraphs of context does not get cited.
Ship the machine-readable layer
Markdown mirrors, llms.txt, structured data, clean feeds, and an MCP endpoint where it earns its place. Technically modest once the store is soundly built — which is exactly why this step is third and not first.
Measure what is measurable
And say plainly what is not. AI crawler hits and referrals from AI surfaces are traceable in logs and analytics. Whether a model mentions you inside an answer is only checkable by sampling. Anyone selling you a ranking guarantee here is selling something nobody can deliver.
Trusted by brands worldwide
Brands whose stores we build and look after
On this topic the best proof is not the logo strip, it is the website you are on. Append .md to this URL.
AI readiness FAQ
The questions this topic deserves
Including the sceptical ones: what it measurably returns, what Google reads of it, and when it is too early.
Sceptical?
Fair. Ask us.
Honest answer: rarely in a directly measurable way. We treat it as preparation, not as a channel. What already pays today is the foundation underneath, because server-rendered content, structured data and clean feeds help classic search just as much. If somebody promises you revenue from AI search, ask how they measure it.
No, and we do not claim otherwise. Google has said repeatedly that it does not use the file. The benefit lies with other systems and with agents fetching a page deliberately. That is why llms.txt is second on our list rather than first, and why the most important item above is the dullest one: content that exists without JavaScript.
Largely yes, and that is good news. The overlap is big: rendering, structure, clarity, speed. Three things are genuinely new to it — a machine-readable second form of the same page, passages that stand alone, and a query route like MCP. Anyone telling you it is entirely new is selling a package.
When your store does not server-render yet, when your product data is incomplete, or when you still have unfinished basics in classic search. Then those are the job, and we will tell you so. Putting AI readiness on an unfinished foundation is the most expensive possible order of work.
Part of it, yes — there are good apps for structured data. What apps do not deliver is a second clean text form of your pages, an endpoint like MCP, and the editorial work of making passages understandable alone. That last one returns the most and is the one no app can take over.
You do not have to take our word for it. Append .md to any URL on this site, open /llms.txt, look at /mcp. This website ships everything described on this page, in production, and has for months. That is a different kind of evidence than a reference list.
Related
What this is built on
The work AI readiness sits on top of.
Martin Winkler
Co-Founder & CEO
Mark Chang
Founder & CTO