Google says keywords in URLs are a tiny ranking signal

Google has put on record that keywords in a URL are "an incredibly small signal" for ranking, in the first episode of a new Search Central video series released on 14 October 2026. Google's Gary Illyes and Cherry Prommawin also spelled out what can and cannot go in a domain name, how emoji in a URL path are handled, and why hyphens still beat underscores.

What Google published

The series is called The Non-Canonical Cabinet. Google teased it on 6 October, asked viewers to post their own edge-case questions in the comments, and released the first full episode eight days later. The episode description says the questions "don't belong on a standard FAQ page but are still useful for Search and generative AI features."

Illyes previously fronted How Search Works, a 2024 Search Central series that walked through crawling, indexing and serving. The new show runs the other way round. It starts from odd questions site owners actually ask and works back to how Google handles them. Episode one covers domains, URL paths, emoji and keywords. The next episode is about robots.txt, and Google has not given a release schedule.

Domains follow stricter rules than paths

Domain names are governed by standards from ICANN and the Internet Engineering Task Force. Those rules, RFC 1035 and RFC 1123, allow letters, numbers and hyphens, and Prommawin said Google follows the same standards.

Domains in scripts such as Chinese or Arabic are possible through Internationalised Domain Names. Behind the scenes they are converted to Punycode, a string that starts with "xn--". Browsers usually show the native script, but Illyes pointed out that some scripts contain characters that look almost identical to Latin ones, such as a Cyrillic letter that resembles an English "p", which can confuse both people and browsers.

Emoji are a different matter. Illyes said the current standards do not allow emoji in a domain name. ICANN's board took the same position in a 2017 resolution, finding emoji incompatible with the current rules for Internationalised Domain Names, although it acknowledged that some country-code registries still let people register them at the second level. The episode uses a chicken and fire emoji dot com as its example of a domain that cannot exist.

Emoji in a path get crawled, then percent-encoded

The URL path, meaning everything after the domain, has fewer restrictions. Illyes said Googlebot will crawl a path containing an emoji and Google's systems will process it. What actually travels over the wire is the percent-encoded version, so the chicken emoji in the episode becomes %F0%9F%90%94.

The catch is display. Illyes noted that some browsers and social apps will not render the emoji at all and show a jumble of encoded characters instead. None of this is new ground for Google's written guidance. Its URL structure page already tells site owners to percent-encode non-ASCII characters, and one of its own worked examples is a pair of emoji. If your URLs follow that page today, nothing changes.

Keywords in URLs count for almost nothing

The part most business owners will care about is the ranking question. Illyes called keywords in a URL "an incredibly small signal". He said it "really only matters when we have absolutely nothing else to go on", and the example he gave was a site that is seconds old with no content. Once Google has seen the content, keywords in the domain and URL "become basically irrelevant".

This matches what Google has said for a decade. John Mueller described URL keywords as a "very small ranking factor" in 2016 and a "very very lightweight factor" in 2021. The new episode restates the mechanism Mueller gave in 2021. The URL only carries weight in the brief window before Google has anything better to go on.

The episode's closing recap tells viewers to stop trying to game results by packing keywords into URLs. The example, best grilled chicken in London, cheap, delicious, amaze, amaze, amaze, is used to show what not to do.

Hyphens, settled again

On word separators, Illyes said Google reads a phrase joined by underscores in a path as a single word. That is not always a problem, since Google can sometimes still pick out the individual words, but hyphens are the separator Google recommends and the one most of the web already uses.

Google's documentation gives the same guidance and explains the history. Underscores have long been used to keep concepts together, for example in programming function names like format_date, which is why Google does not recommend them as word separators.

What the URL documentation already asks for

The episode does not change the written guidance, so it helps to know what that guidance says. Google's URL structure page asks for:

• Readable words instead of long ID numbers, so /wiki/Aviation rather than index.php followed by a topic number and a hash.
• Words in the language your audience searches in, including transliterated words where that applies.
• Percent encoding for non-ASCII characters in the href of your internal links.
• Hyphens between words, and as few URL parameters as possible.
• Consistent casing, because Google treats /APPLE and /apple as two different URLs.
• For multi-regional sites, a country domain such as example.de or a subdirectory such as example.com/de/.

The same page warns against using fragments, the part of a URL after a hash, to change page content, and points single-page apps to the History API instead. That advice sits alongside the soft navigation Core Web Vitals metrics Chrome published this week, which start to measure route changes on single-page apps.

Who this affects in Australia

If you have ever been told your plumbing site needs /emergency-plumber-melbourne-24-7-cheap-best/ to rank, Google has now said on record that it will not help once the page has content. The content, the links and the overall quality of the site do the work. The URL is a label.

The reverse matters just as much. Renaming URLs on a live site to add or remove keywords means redirects, a recrawl, and a period where rankings can wobble, all for a signal Google describes as basically irrelevant once content is indexed. Leave working URLs alone.

Where the advice does apply is new pages and new sites. Pick a readable, hyphenated path in plain English the first time, keep it short, and move on. For the short period before Google has indexed the content, a sensible URL is a small help. After that it stops mattering.

The episode follows Google's recent walkthrough of PDF indexing and ranking, another case of Google staff explaining what Search does with a technical choice rather than leaving site owners to guess.

What to do this week

• Do not rename existing URLs for keyword reasons. If an agency proposes it as an SEO win, ask what else the change will do, because the ranking signal is close to zero.
• For new pages, use short, descriptive, hyphenated paths in the language your customers search in.
• Check your CMS is not generating underscores or mixed-case duplicates of the same page. Pick one case and redirect the other.
• Skip emoji in paths on a business site. Google can crawl them, but a shared link that shows up as %F0%9F%90%94 in a customer's messaging app is a poor first impression.
• If you changed a batch of URLs recently, watch the performance report over the next few weeks, and remember that Search Console data delays can make a redirect look worse than it is for a few days.

Google is taking questions for future episodes in the YouTube comments under each video. If a URL edge case on your own site is not answered by the documentation, that is now the place to ask.