Research · October 23, 2026
Are websites ready for AI? 180 sites in 6 languages
When someone asks ChatGPT, Claude or Perplexity about a company, the assistant reads that company's site. We checked 180 well-known sites, 30 in each of six languages, for the things that decide whether it can.
What we found
- 15% of the sites we could read tell ChatGPT, Claude or Perplexity to stay out in their robots.txt. Those assistants can't quote a page they are told not to read.
- Korean sites lean hardest on JavaScript: 68% have readable text before any script runs, against 100% of French sites. Most AI agents don't run scripts, so the rest look close to empty to them.
- Only 47% say who runs the site in Organization schema, the markup AI reads for a company's name, logo and profiles.
- llms.txt is still rare: 42% of English sites have one, 7% of Japanese sites.
भाषा के अनुसार
| Language | Sites read | Average score | Keep an AI agent out | Text without JavaScript | Organization schema | llms.txt |
|---|---|---|---|---|---|---|
| English | 27 of 30 | 79 | 22% | 89% | 41% | 42% |
| Korean | 28 of 30 | 74 | 18% | 68% | 43% | 22% |
| Japanese | 24 of 30 | 83 | 8% | 92% | 33% | 7% |
| German | 18 of 30 | 76 | 6% | 83% | 44% | 27% |
| French | 19 of 30 | 89 | 11% | 100% | 58% | 12% |
| Spanish | 22 of 30 | 85 | 23% | 91% | 68% | 32% |
Percents are of the sites we could read in that language. "Keep an AI agent out" means robots.txt tells ChatGPT, Claude or Perplexity not to read the home page.
How we checked
- 180 sites we picked by hand: well-known shops, banks, telecoms, news sites, travel and software companies in each language's home market. It is a sample of big names, not a random one.
- We used the same check as Tappy's free AI agent check (tappy.dog/compare) on October 23, 2026. Real Chrome read each home page, robots.txt and llms.txt.
- We ran it from a home connection in Seoul. A firewall can treat that differently from the data centers AI agents come from, so firewalls are left out: an agent counts as kept out only when robots.txt says so.
- The score out of 100 adds up robots.txt access for ChatGPT, Claude, Perplexity and Gemini, text before JavaScript runs, and the home page's title, description, main heading, language, Organization schema, contact link, canonical link and preview picture.
- 42 of 180 sites gave us no home page to read: a bot check, or no answer within two minutes. They are left out of the numbers (138 sites read).
Every site we checked
English
amazon.com, target.com, walmart.com, homedepot.com, nytimes.com, bestbuy.com, costco.com, bbc.co.uk, washingtonpost.com, cnn.com, theguardian.com, chase.com, wellsfargo.com, verizon.com, delta.com, t-mobile.com, bankofamerica.com, att.com, expedia.com, booking.com, airbnb.com, united.com, etsy.com, notion.so, stripe.com, shopify.com, slack.com, zillow.com, hubspot.com, mailchimp.com
Could not read: homedepot.com, delta.com, etsy.com
Korean
naver.com, daum.net, coupang.com, 11st.co.kr, kakaocorp.com, gmarket.co.kr, musinsa.com, shinhan.com, kbstar.com, kia.com, wooribank.com, lguplus.com, sktelecom.com, yes24.com, kyobobook.co.kr, yanolja.com, baemin.com, oliveyoung.co.kr, ssg.com, lotteon.com, danawa.com, donga.com, hani.co.kr, chosun.com, toss.im, kurly.com, lge.co.kr, saramin.co.kr, wanted.co.kr, kt.com
Could not read: lge.co.kr, kt.com
Japanese
nitori-net.jp, yahoo.co.jp, rakuten.co.jp, kakaku.com, yodobashi.com, jalan.net, tabelog.com, zozo.jp, jal.co.jp, nikkei.com, biccamera.com, mufg.jp, smbc.co.jp, asahi.com, yomiuri.co.jp, au.com, ana.co.jp, docomo.ne.jp, suumo.jp, hotpepper.jp, softbank.jp, cookpad.com, sony.jp, panasonic.jp, toyota.jp, moneyforward.com, cybozu.co.jp, muji.com, sansan.com, freee.co.jp
Could not read: nitori-net.jp, biccamera.com, mufg.jp, smbc.co.jp, docomo.ne.jp, muji.com
German
otto.de, zalando.de, spiegel.de, tagesschau.de, faz.net, rossmann.de, mediamarkt.de, lidl.de, bild.de, rewe.de, zeit.de, dm.de, deutsche-bank.de, aldi-sued.de, commerzbank.de, telekom.de, vodafone.de, idealo.de, bahn.de, chefkoch.de, sparkasse.de, mobile.de, n26.com, check24.de, immobilienscout24.de, personio.de, thalia.de, otto-group.com, aboutyou.de, tchibo.de
Could not read: otto.de, zalando.de, rossmann.de, mediamarkt.de, lidl.de, rewe.de, zeit.de, telekom.de, vodafone.de, idealo.de, bahn.de, otto-group.com
French
liberation.fr, lemonde.fr, lefigaro.fr, darty.com, boulanger.com, leboncoin.fr, fnac.com, cdiscount.com, orange.fr, free.fr, auchan.fr, sncf-connect.com, carrefour.fr, laredoute.fr, sfr.fr, galerieslafayette.com, credit-agricole.fr, decathlon.fr, sephora.fr, bouyguestelecom.fr, lcl.fr, blablacar.fr, doctolib.fr, marmiton.org, alan.com, qonto.com, societegenerale.fr, leroymerlin.fr, seloger.com, maif.fr
Could not read: liberation.fr, darty.com, boulanger.com, leboncoin.fr, fnac.com, orange.fr, free.fr, sncf-connect.com, galerieslafayette.com, societegenerale.fr, leroymerlin.fr
Spanish
mercadona.es, pccomponentes.com, carrefour.es, mediamarkt.es, abc.es, elcorteingles.es, elpais.com, elmundo.es, santander.es, as.com, bbva.es, bankinter.com, marca.com, movistar.es, vodafone.es, caixabank.es, orange.es, idealista.com, renfe.com, fotocasa.es, iberia.com, milanuncios.com, casadellibro.com, fnac.es, decathlon.es, cabify.com, holded.com, wallapop.com, glovoapp.com, lavanguardia.com
Could not read: carrefour.es, mediamarkt.es, caixabank.es, idealista.com, milanuncios.com, casadellibro.com, fnac.es, lavanguardia.com
Being read is step one
A site AI can read still has to be the one it names. Proxylang Answers asks the assistants your customers' questions every week and writes answer pages on answers.yoursite.com in each language you sell in.