We scanned 20 German stores. None is fully readable for AI.

AI assistants answer shopping questions by reading stores from the outside. What they can read decides whether a store appears in those answers. We wanted the number behind that sentence, so we checked 20 established German stores across categories, by hand, against the signals an agent actually requests.

What was checked

Eight public signals per store, grouped into the six criteria the scan works by. Whether the robots.txt lets AI crawlers in. Whether a machine-readable product catalog is reachable. Whether product pages carry valid JSON-LD schema with price, availability, GTIN and rating markup. Whether shipping and return data exist in machine-readable form. And whether an llms.txt with agent instructions is present. Every signal is public, every store owner can check them without any tool.

The results

• 0 of 20Fully readable for AI agentsNo store passes every checkable signal.
• 0 of 20Shipping or return data in the schemaThe most consistent single gap in the sample.
• 9 of 20No machine-readable catalog access7 passive defaults, 2 active blocks.
• 2 of 20Effectively closed to AI accessVia robots.txt rules or bot protection.
• 2 of 20Deliberate AI bot policy of any kindThe rest is untouched default state.
• 6 of 11Product pages without GTINOf the 11 checkable product pages.
• 8 of 11Product pages without rating markupOf the 11 checkable product pages.
• 2 of 11Product schema missing or brokenOf the 11 checkable product pages.

Two denominators matter here. The access signals were checkable for all 20 stores. The product-page signals (GTIN, rating, schema) were only checkable where an open catalog exposed a product URL, which leaves 11 stores. 6 of 11 without GTIN means 6 of 11, not 6 of 20.

How your own store scores on these signals is what the free scan measures, in around 30 seconds.

What stands out

The shipping gap is the loudest number. Not one of the 20 stores tells an agent what delivery costs or how returns work, in a form the agent can read. A buyer asking an AI assistant about shipping gets a guess, or nothing.

The second pattern is how quiet the field is. Only 2 of 20 stores show any deliberate AI bot policy at all. Everything else is the untouched default state of the platform. I read that as a sign that this has not reached most merchants' desks yet, and that is a reading, not a measurement. n = 20 cannot prove a market.

Methodology and limits

Live check per HTTP against the public stores on July 2, 2026, with a standard client. 20 established German stores across categories, selected by prominence, not randomly. Checked were the eight signals above, nothing else.

The limits, openly. First, participation in closed AI product feeds cannot be verified from the outside, so it was not counted. Second, actual presence inside AI shopping indexes was not queried, only the technical basis for it. Third, when a store is unreachable for a standard client, that can be generic bot protection rather than an AI decision. It affects AI crawlers all the same, which is why it counts as a gap, not as intent. And the sample is 20 stores on one day. A snapshot, not a market study.

Citing this study? Name it as: manual check of 20 German stores, July 2, 2026, n = 20, not representative. Press questions reach me via the contact page.