8 Common International SEO Mistakes (and How to Fix Each)
International SEO fails in remarkably repetitive ways. The eight most common mistakes are: broken hreflang pairs, shipping raw machine translation, translating a keyword list instead of researching one, picking the wrong site structure, forcing redirects by IP, ignoring the target market’s search engine, stuffing every language into one sitemap, and tracking Google Search Console at the site-wide average. Each has a clear symptom and a clear fix — here they are, one by one.
Mistake 1: hreflang with no return link or no self-reference
Symptom: language versions cannibalize each other in the rankings; your Simplified Chinese page shows up in Taiwanese results. Cause: hreflang is a two-way contract. The moment page A points to page B but page B doesn’t point back, Google ignores the whole annotation set — and skipping the self-reference breaks it the same way. This is the number-one cause of Traditional-vs-Simplified self-cannibalization. The fix: on every page, list all language versions plus a self-reference plus x-default, kept bidirectionally consistent. Crawl the whole site with a tool to verify, and cross-check the country × directory split in GSC.
Mistake 2: shipping raw machine translation as your localized version
Symptom: the foreign-language version has a low indexing rate and near-zero clicks; 90 days in, it’s a pile of zero-click pages. Cause: raw MT gets the wording subtly wrong, misses the local keyword, and reuses examples that mean nothing to a local reader — with zero appeal, while dragging down your site-wide quality signal. This is exactly how you manufacture content debt. The fix: treat MT as a first draft only. Before publishing, run it through a local editor: fix the wording, rewrite the title / meta / slug, and swap in local examples.
Mistake 3: translating a keyword list instead of doing native research
Symptom: you’ve written plenty of articles, but that language’s impressions won’t lift; the GSC query terms don’t match your article titles. Cause: every market phrases and asks things differently (in Chinese, 搜尋 vs 搜索; “health check” vs “diagnosis”). A translated list aims at keywords nobody actually searches. The fix: run keyword research and build an intent matrix separately for each language.
Mistake 4: the wrong (or messy) site structure
Symptom: multiple domains each struggling alone, with authority scattered; or the same language living on both a subdomain and a subdirectory, cannibalizing itself.
Cause: several structures grew up without a single deliberate decision.
The fix: in most cases, choose subdirectories (/en/, /zh-hant/, /zh-hans/) — they share the domain’s authority and cost the least to maintain. Reserve ccTLDs and subdomains for cases with a clear legal or organizational requirement.
Mistake 5: forced redirects based on IP
Symptom: users can’t freely switch language; Googlebot (which crawls mostly from US IPs) only ever sees the English version, so the other languages index abnormally. Cause: detecting IP and 302-redirecting locks the crawler into a single version. The fix: make every language version directly accessible. Suggest a language with a dismissible banner rather than a forced redirect, and leave the routing to hreflang.
Mistake 6: ignoring the target market’s search engine
Symptom: you built the Simplified Chinese version, then discovered your target audience can’t find it at all — because they’re in mainland China. Cause: Google has been unavailable in mainland China since 2010; that market is a different game led by Baidu (ICP filing, in-country hosting, different indexing logic). The fix: define the market before the language. Overseas Simplified Chinese runs on Google/Bing and can reuse your existing playbook; the mainland market needs its own separate project to evaluate Baidu SEO — don’t lump the two together.
Mistake 7: one sitemap for every language
Symptom: when the indexing rate drops, you can’t tell which language is the problem. The fix: one sitemap per language, submitted together via a sitemap index, so GSC can monitor indexing by language.
Mistake 8: reading only the site-wide GSC average
Symptom: the dashboard numbers look “fine,” but in reality one language version flatlined long ago — averaged out by the data from the others.
The fix: filter performance reports by directory (“Page contains /en/” and so on), and track each language as its own product line. Run a country × directory cross-check once a month.
Self-audit: a 5-minute quick screen
- Pull 3 random pages — does each hreflang have all languages + self-reference + x-default, kept bidirectionally consistent?
- Are the foreign-language titles written in local phrasing, or translated word-for-word from the original?
- Search your brand term in incognito + a VPN (or by changing your language setting) — does the correct language version come up?
- In GSC, are the top three impression countries for your Simplified Chinese directory reasonable?
- Does each language have its own sitemap?
If you can’t answer or get two of the five wrong, it’s worth scheduling a full audit.
FAQ
Q1: Of these 8 mistakes, which should I fix first? Fix hreflang first (mistake 1) — while it’s broken, wrong routing eats the payoff from everything else you do. Next comes content localization (mistakes 2 and 3), and only then the structure and tracking issues.
Q2: How long after fixing a mistake until I see improvement? Technical fixes (hreflang, sitemap, redirects) usually take effect gradually over 2–6 weeks as pages get re-crawled. Content-localization fixes need 2–3 months or more, because new content has to accumulate engagement signals.
Q3: My site is small and short on resources — can I just do one language first? Yes, and you should. Get your primary language deep and solid first, then copy the same process to a second language. A half-baked trilingual site is worse than a solid single-language site with a planned expansion path.
Q4: Do these mistakes affect getting cited by AI (GEO)? Yes. AI engines pick the “best-stated source” separately in each language, so raw MT and mismatched keywords knock you out of the citation race in that language entirely; messy hreflang can make the AI grab the wrong language version.
Of the eight mistakes, hreflang is the easiest to solve automatically — GeoSeoToday’s hreflang generator and GEO checker produces the correct “all languages + self-reference + x-default” tags in one click. For the systematic approach and the full checklist, see the pillar guide, the Complete Guide to Multilingual / International SEO.