E-commerce statistics in articles are usually borrowed from US studies. So I made my own: on 6 August 2026 I measured 21 Czech e-shops with a single methodology — from smaller brand stores to big players, across sectors (fashion, cosmetics, food, outdoor, books, pharmacies, pets, musical instruments). Not to pillory anyone by name — results are reported in aggregate — but to have data about the state of Czech e-commerce instead of impressions.
Two numbers stand out: 7 of 21 e-shops already have llms.txt — the file a website uses to talk to AI search engines. And at the same time, 8 of 21 don't even have a correct H1 heading. Czech e-commerce is simultaneously sprinting into the AI era and tripping over 2010-era basics.
Of 21 measured Czech e-shops (8/2026): 33% have llms.txt for AI search engines (mostly auto-generated by platforms like Shopify), 67% have some structured data, but only 38% have an Organization/Store type and exactly one has FAQ schema. 38% have a broken H1 (none or several), a median of 28% of images lack alt text, and only 43% have complete OG tags for sharing. Speed is good on desktop (median LCP 0.95 s in lab conditions), but 4 e-shops have layout-shift problems (CLS > 0.1). Conclusion: the technical top is preparing for AI search, the average loses points on things that cost hours of work.
Methodology (so you can repeat it)
- Sample: 21 Czech e-shops of various sizes and sectors. Originally 22; one didn't finish measurement (timeout). Results are published in aggregate, without names — the goal is a picture of the market, not a pillory.
- When and how: 6 August 2026, headless Chromium, desktop 1440×900, no network throttling, each e-shop's homepage. LCP and CLS via PerformanceObserver, content checks from the DOM after full load, files (robots.txt, sitemap.xml, llms.txt) via a plain HTTP client with content verification.
- What follows from that: the speed numbers are lab desktop under ideal conditions — real mobile field data (CrUX) will be worse. Content checks (H1, alt, schema, OG) don't depend on conditions.
- I archive the complete anonymized dataset and will gladly document the methodology.
Finding 1: a third already talks to AI search engines
7 of 21 e-shops (33%) have a working llms.txt — a text file that hands AI crawlers (ChatGPT, Perplexity, Claude…) a structured description of the store. For some it's a file generated automatically by the platform (Shopify started issuing "Agent Instructions"), for others it's deliberate work.
Why that's interesting: llms.txt is an opt-in signal — nobody has to have it, and a year ago almost nothing in Czechia did. A third among established e-shops means preparing for AI search has stopped being exotic. If you run an e-shop on a custom build or a platform that doesn't generate this, your competitors may already be answering AI search engines — and you aren't. (Why AI traffic is worth caring about is covered in AI search isn't the future — it converts several times better than organic.)
Related: only 11 of 21 have a sitemap at the standard /sitemap.xml address, and 6 of 21 didn't serve robots.txt to a plain HTTP client (typically due to anti-bot protection). The latter cuts both ways — scraper protection can also throttle the crawlers you want to let in.
Finding 2: basics that should be a given, aren't
| Check | Result (of 21) | What it means |
|---|---|---|
| Exactly one H1 on the page | 13 ✓ / 8 ✗ | 38% have zero, two or three H1s |
| Meta description present | 20 ✓ / 1 ✗ | the market is good here |
| Image alt texts | median 28% of images without alt | 12 e-shops have >20% undescribed images |
| Complete OG tags (title + image) | 9 ✓ / 12 ✗ | 57% render wrong when shared |
| Structured data (JSON-LD) | 14 ✓ / 7 ✗ | but only 8 have an Organization/Store type |
| FAQ schema | 1 of 21 | a huge unused opportunity |
These are exactly the things that decide citability in AI answers as well as classic SEO — and they don't cost tens of thousands. Missing alt texts have also been a legal topic since June 2025 under the accessibility act, which applies directly to larger e-shops.
Finding 3: desktop speed is decent, layouts jump
- Median LCP 0.95 s, average 1.03 s — solid under ideal lab conditions. Only one e-shop crossed the 2.5 s threshold (3.3 s); the fastest managed 0.29 s. The gap between fastest and slowest: 11×.
- CLS: 4 e-shops above 0.1 (one above 0.25) — the layout jumps during load, which frustrates users and drags down Core Web Vitals.
- Median 115 HTTP requests per homepage, the record holder 403. Median 3,300 DOM elements, maximum 8,380.
A reminder: no throttling, fast connection. On an average mobile connection the values will be several times worse — desktop lab numbers are the "best possible case".
What to take from this if you run an e-shop
- Check your H1, alt texts and OG tags. Free check: F12 → Elements, search
<h1; test sharing at opengraph.xyz. Fixing is a matter of hours. - Add structured data — at minimum Organization/Store and Product; almost nobody has FAQ schema, yet it's the most easily citable format for AI answers.
- Consider llms.txt. A third of the market already has it. It's not magic — it's a readable summary of what you sell, for how much, and where you deliver.
- Measure CLS on mobile (PageSpeed Insights) — a jumping layout was the most common "invisible" problem we found.
- If all of this sounds like work you don't have capacity for — these are exactly the things I handle within SEO services and Shoptet support, AI visibility included.
Summary
21 Czech e-shops, one methodology, August 2026: 33% already have llms.txt and are preparing for AI search, 67% have at least some structured data. At the same time 38% have a broken H1, a median of 28% of images lack alt text, 57% have incomplete OG tags, and exactly one e-shop has FAQ schema. Desktop speed is decent (median LCP 0.95 s lab), CLS misbehaves at four. The market is splitting: the top invests in AI visibility, the average loses on basics whose fix costs hours, not budgets. The data is my own, methodology above — I'll repeat the same measurement later and we'll see the shift.
Measured 6 Aug 2026. Want to know how your e-shop would do in the same test? Ask for an audit — I'll measure it with the same methodology and add fix priorities.