URL Structure: Hyphens, Language, and When to Change

How Google Recommends Writing URLs

Google's URL structure best practices split into requirements for crawlable URLs and recommendations for clarity. Hyphens and readable words are recommendations — ignoring them doesn't keep you out of search. Switching content with #, on the other hand, means Google won't read those views as separate pages.

What Changed in the October 2026 Update

According to Google's documentation updates, on October 23, 2026 it clarified the guidance on separating words with hyphens versus underscores. The recommendation itself (hyphens) is unchanged; two things are new.

• It now explains why underscores aren't recommended: they're commonly used to keep concepts together, as in function names like format_date, whereas hyphens clearly signal a word boundary
• The examples now cover parameters as well as paths: /summer-clothing/filter?color-profile=dark-grey is recommended; the same URL with underscores, or joined words like /greendress, is not

Some older articles still ranking say Google doesn't care either way, but the current official guidance is hyphens. The Japanese version of the page has the same content (checked November 9, 2026).

Test: We Counted the URLs on Our Own Two Sites

On November 9, 2026, we pulled every URL from the XML sitemaps of two sites we run and counted paths containing underscores, uppercase letters, or Japanese. One is this site (Direbase, built with Next.js); the other is a WordPress media site for web creators.

• The Next.js site had none, since its URLs are defined in code from the start
• Of the two on the WordPress site, one was an author archive (/author/login-name/): WordPress builds it from the login name, so an underscore in the login name appears in the URL. The other was a post whose slug is a romanized personal name, with an underscore between family and given name
• Neither site has Japanese slugs. With post-name permalinks, WordPress can turn a Japanese title straight into the slug, so we set an English slug before publishing

Can You Use Japanese (Non-English) URLs?

Yes. Google recommends words in your audience's language and gives a Japanese example (/ペパーミント) for readers searching in Japanese. In links on your pages (href values), percent-encode the non-ASCII part.

The catch is length. Percent-encoded, one Japanese character becomes about nine characters: a slug meaning "how to check search volume" (11 characters) became 91. It looks like Japanese in the address bar but can arrive as a long string of % codes when pasted into chat or email.

Google doesn't say either choice is treated differently in search, so decide on operations: English words if URLs are often shared or sit alongside English pages; the readers' language if not — and keep one rule across the site.

Should You Change Published URLs?

Steps when you do change URLs

1. Make a mapping of old to new URLs (export every URL from the sitemap and list only the ones that change)
2. Set up server-side 301 (permanent) redirects from old to new. Google says permanent redirects such as 301s don't lose PageRank
3. Update internal links, canonicals, hreflang, and sitemaps to the new URLs instead of relying on redirects
4. Submit the new sitemap in Search Console and check with URL Inspection that the new URLs get indexed
5. Keep redirects as long as possible — Google says generally at least a year

For consolidating case duplicates with canonical tags, see the canonical tag guide; if new URLs don't get indexed after a change, see fixing "Discovered - currently not indexed".

Check URL Format with the SEO Checker

Direbase's SEO Checker looks at the URL of the page you check and shows "Notes on the URL format" for the three cases below. The notes aren't part of the score and appear on every plan.

• The path contains underscores (_) (parameters aren't checked)
• The path contains uppercase letters (non-ASCII characters aren't checked)
• The URL switches content with #/ or #! (in-page anchors like #section aren't flagged)

The notes are input for a decision, not bugs to fix. If you see the underscore note on a published page, use the table above to decide. For the case and # notes, first check whether the same page is served at several URLs and whether Google reads your views as separate pages.

Common Mistakes

• Switching every underscore URL to hyphens at once. Missed redirects or internal links cost more than the change gains
• Changing a URL and adding a 301, but leaving internal links, canonicals, and sitemaps on the old URL. That sends Google conflicting signals
• Missing URLs your CMS generates, such as WordPress author archives. If a login name has an underscore, it shows up in the URL
• Assuming a case-insensitive server makes case harmless. Even if the server returns the same page, Google treats them as different URLs
• Stuffing keywords into URLs. Words that describe the page are enough; longer URLs don't help, and Google's guidance stops at readable words and separators

Related Tools

• SEO Checker — 45 checks plus notes on the URL format
• Site Diagnosis — check the pages in your sitemap together
• Meta Tag Checker — check canonical and OGP URLs

Sources (checked November 9, 2026)

• Google Search Central — URL structure best practices for Google Search (last updated 2026-10-21)
• Google Search Central — Documentation updates (October 23, 2026: clarified word separation with hyphens vs. underscores)
• Google Search Central — Redirects and Google Search (permanent redirects show the new URL in results)
• Google Search Central — Site moves with URL changes (no PageRank loss from 301s; keep redirects for at least a year)