Google Updates 2 Help Pages on 23 October: Hyphens Over Underscores in URLs, Percent-Encoding in Robots.txt

Ras Al Khaimah, United Arab Emirates, 24 October 2026.

Google updated two of its core help pages on 23 October 2026, writing advice that its staff had given in recent videos and social posts into the official documentation. The URL structure best practices page now spells out why hyphens are preferred to underscores as word separators, and the robots.txt specification page now recommends percent-encoding non-ASCII characters in rule paths. Both changes are logged in Google's Search documentation updates and crawling documentation changelog, which give the same reason for each: to bring the pages in line with episodes of Google's "The Non-Canonical Cabinet" video series.

Neither change alters how Google crawls, indexes or ranks pages. Google's changelogs describe one as a clarification and the other as an added recommendation. What changes is that advice SEO teams previously had to cite from a video or a social post is now in the reference pages that developers and auditors check first.

What changed

Page	Changelog date	What the page now says
URL structure best practices (Search Central)	23 October 2026	Recommends hyphens instead of underscores to separate words, and explains that underscores are commonly used to keep concepts together, while hyphens signal a word boundary
How Google interprets the robots.txt specification (Crawling docs)	23 October 2026	Rule paths may be raw UTF-8 or percent-encoded; Google canonicalises raw UTF-8 into percent-encoded form, and percent-encoding is recommended because it is easier for third-party tools to process

Hyphens versus underscores

The URL structure page now has a section headed "Use hyphens to separate words". It recommends separating words in URLs where possible and using hyphens rather than underscores, because this helps users and search engines identify concepts in the URL. The page gives a historical reason: underscores are commonly used in programming languages to name functions, such as format_date, so they tend to be read as joining words together. Hyphens, the page says, clearly signal a word boundary.

The page's examples apply the rule to paths and to parameters alike. It lists summer-clothing/filter?color-profile=dark-grey as recommended, and both an underscore version of the same URL and a version with words joined together (greendress) as not recommended.

The same page already recommends percent-encoding non-ASCII characters in URLs and shows encoded and unencoded examples in Arabic, Chinese, German and emoji. Google's changelog for the search documentation lists the hyphen clarification as the 23 October change.

Percent-encoding in robots.txt

On the robots.txt specification page, the section on URL matching now states that rule paths can be written as raw UTF-8 characters or as percent-encoded UTF-8 strings. It explains that, as defined in section 2.2.2 of RFC 9309, Google's crawlers compare rules against URLs in their percent-encoded form and canonicalise raw UTF-8 rule paths into that form automatically. The page gives an example: a disallow rule for /foo/bar/ツ and one for /foo/bar/%E3%83%84 are treated identically by Google's parser.

The new recommendation is practical rather than technical. Google says its own systems parse raw UTF-8 properly, but that percent-encoding rule paths makes the file easier to process for third-party tools and services. In other words, a robots.txt file that works for Googlebot may still be misread by an SEO crawler, a CDN rule engine or another bot if it relies on raw non-ASCII characters.

Context: from video series to documentation

Both changes trace back to recent public comments by Google staff, which this newsroom has covered as they happened:

• On 16 October we reported Gary Illyes's comments that keywords in URLs are a tiny signal and hyphens are the better separator.
• On 21 October we reported John Mueller's explanation that bracketed and percent-encoded URLs are treated as different URLs, so sites should pick one form.
• On 22 October we covered Google's advice that emoji user-agents, decorative rules and custom AI directives do not work in robots.txt.

The 23 October changes are the documentation catching up with those statements. They also continue a busy period for Google's help pages. The search documentation log shows changes on 1, 8, 16, 21 and 23 October, and the crawling changelog shows changes on 6 and 23 October. Our earlier analysis of Google's 2026 documentation changes found that edits of this kind are frequent and rarely announced outside the changelogs.

The two changes are separate from the robots.txt and URL matching rules themselves, which have not changed. Google's robots.txt page still says, for example, that Google caches robots.txt for up to 24 hours in general and enforces a 500 KiB size limit.

What this means for SEO teams

• Do not rush to rewrite underscore URLs. The documentation states a preference for new URLs. Changing existing URLs means redirects and a period of reprocessing, and Google's statements on URL keywords suggest the ranking benefit is small. Apply hyphens to new sections, new templates and new parameters.
• Audit robots.txt files on non-English sites. If rule paths contain raw non-ASCII characters, convert them to percent-encoded UTF-8. Google treats both forms the same, but other crawlers and testing tools may not.
• Keep one URL form across the site. Combined with Mueller's guidance on encoded and raw brackets, the safest pattern is to emit a single, encoded form in internal links, canonicals and sitemaps.
• Test with more than one parser. After editing robots.txt, check the rules in Search Console's robots.txt report and in at least one third-party crawler, since the point of Google's recommendation is cross-tool consistency.
• Watch the changelogs. Both changes appeared only in the changelogs. Subscribing to the search documentation RSS feed is the simplest way to catch the next one.

Conclusion

Google's 23 October edits do not change any ranking or crawling behaviour, but they turn two recent pieces of advice into documented recommendations: hyphens over underscores in URLs, and percent-encoded paths in robots.txt. For SEO teams, the practical work is limited to new URL patterns and to robots.txt files on sites that use non-ASCII paths.

Google has said both edits were made to synchronise its documentation with "The Non-Canonical Cabinet" video series, so further changes are likely as new episodes air.

Questions and answers

What did Google change on 23 October 2026?

Google updated its URL structure best practices page to clarify that hyphens, not underscores, should separate words in URLs, and updated its robots.txt specification page to recommend percent-encoding in rule paths. Both changes are logged in Google's documentation changelogs.

Does Google treat raw UTF-8 and percent-encoded robots.txt rules differently?

No. Google's robots.txt specification says its crawlers canonicalise raw UTF-8 rule paths into percent-encoded form, so a rule for /foo/bar/ツ and one for /foo/bar/%E3%83%84 are treated identically. Percent-encoding is recommended because it is easier for third-party tools to process.

Why does Google prefer hyphens to underscores in URLs?

Google's URL structure page says hyphens help users and search engines identify concepts in a URL, while underscores are commonly used in programming to keep concepts together, as in format_date.

Why did Google make these documentation changes?

Google's changelogs say both edits were made to synchronise the documentation with episodes of its The Non-Canonical Cabinet video series.

###