Check that AI crawlers see the right language version of your multilingual site

My site [site URL] is published in these languages: [languages and URL structure, e.g. en: /, de: /de/, fr: /fr/]. My Search Console property: [Search Console property]. The two most important pages per language: [homepage and one service page URL in each language]. One commercial question per market: [one question per language]. This is read-only. Most AI crawlers and Google's crawlers come from servers in the United States, and their browser language is usually English. If the site redirects visitors automatically by IP country or browser language, these crawlers may never see the German or French page. The point of this job is to measure that. 1. Tag alignment. Read the two pages in each language with site_read_url (default max_bytes; if the body is cut before `</head>`, use max_bytes=1000000). For each page extract the `<html lang>` value, the canonical URL, every `rel="alternate" hreflang` link and x-default. Check: does the canonical point to the page itself (if a language version's canonical points to another language, that version drops out of the index); are the hreflang links reciprocal (if A points to B, B must point to A); does every hreflang URL return 200 (site_read_url with max_bytes=0); are the language codes valid (e.g. "en", "de", "en-GB"; do not mix "de-DE" and "de"); do html lang and hreflang name the same language. 2. Sitemap. Read sitemap.xml and any child sitemaps with site_read_url. Compare the xhtml:link alternates with the on-page hreflang for two pages. 3. Redirect behaviour. Read each language's homepage with site_read_url and max_bytes=0 and note the redirects list. This tool sends requests as a browser from one fixed server, so it does not show what a visitor from another country sees. For that, also measure each language's homepage with seo_technical_audit, which renders the page with JavaScript from servers in another location. If the two tools show a title or H1 in different languages, the site is redirecting visitors from that location to another language version. Flag this at the highest severity and ask me: does the redirect rule exempt search and AI crawlers (Googlebot, Bingbot, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot), and does it use 302 rather than 301. You cannot see the code or server settings, so do not guess the answer. 4. Google's choice. Inspect the two pages per language in one gsc_inspect_urls call. If Google's chosen canonical differs from the page itself (for example the English page is chosen for the German page), say so. Use get_report with platform='search_console' and breakdown='country' for the last 90 days, then breakdown='page by impressions', and add up impressions and clicks per language folder. Flag any language version that gets almost no impressions in its own market. 5. The language AI sees. Run each market's commercial question through seo_ai_overview with that market's location and language values and target=[domain]. If one of my pages is cited, note which language version it is. If references is empty, use the links inside ai_answer. My English page being cited for a German question is a sign the German page is not being seen enough. Format the output like this. First a four-sentence summary. Then a page table: URL, html lang, canonical (self or not), hreflang count and reciprocity, x-default, Google's chosen canonical, status. Then a redirect table: language homepage, title in site_read_url, title in seo_technical_audit, difference. Then a market table: language, impressions and clicks in its main market, version cited in the AI answer. End with a fix list in order of severity. If there is an automatic language redirect, add this general recommendation: exempting crawlers from the redirect, or showing visitors a banner that suggests a language instead of redirecting them, is the approach Google recommends for multilingual sites.