Blog Post URL Slug Best Practices: Pick Once, Keep Forever
The best blog post URL slug is one you never have to change: lowercase, hyphenated, a few words that name the topic, no year on an evergreen post, and pinned in front matter so a title edit can't move it. Google's own SEO Starter Guide says keywords in the URL path "alone have hardly any effect beyond appearing in breadcrumbs." So the slug matters less for ranking than for permanence, and permanence is harder to get than it looks. Ahrefs found that AI assistants send visitors to 404 pages 2.87x more often than Google Search, and one suspected reason is pages that were "deleted or moved (without redirecting the original page)."
If your blog lives in a repo on Jekyll, Hugo, Astro, Next.js or Eleventy, you get to set this convention once, before post #1. After that, every rule is about not having to rename anything later.
Why the slug still matters when it barely ranks
Start with the uncomfortable part: the slug is a weak ranking lever. John Mueller has said words in URLs "only play a tiny role for Google Search" (Search Engine Journal, 2022). Back in 2016 he said keywords in URLs are "a very small ranking factor", not something he'd "really try to force" (Search Engine Roundtable).
The slug is also less visible than it used to be. Since 23 January 2025, Google no longer shows breadcrumbs on mobile results. The mobile result shows just the domain. Desktop still shows domain plus breadcrumb.
So who still reads your slug? Desktop searchers. Anyone looking at your link in Slack, a newsletter or a forum thread. AI assistants that cite your URL. And you, eighteen months from now, scanning your content/ folder. None of them care whether the exact keyword is in it. All of them care whether it still works. Mueller's 2022 advice says the same: "Use the URLs that can last over time, avoid changing them too often and try to make them useful for users."
The slug convention, decided once
Google's URL structure documentation covers the mechanics:
- Readable words, not IDs. It recommends "readable words rather than long ID numbers."
- Hyphens, not underscores. Run-together words like
greendressare listed as not recommended either. - One case. "Google Search's URL handling is case sensitive," so
/Blog/My-Postand/blog/my-postare different URLs. Lowercase everything. - Your audience's language, with non-ASCII characters percent-encoded.
Beyond those, three decisions are yours.
How many words, and which ones
A workable house rule is three to five meaningful words built around the phrase the post answers. For a post titled "How to Fix 'Crawled — Currently Not Indexed' on a Small Blog", crawled-not-indexed does the job. how-to-fix-crawled-currently-not-indexed-on-a-small-blog is the title with hyphens, and it looks wrong the moment the title changes. This is a convention, not a measured optimum. Google's starter guide contrasts a helpful example.com/pets/cats.html with random identifiers, and nothing in its documentation rewards a longer slug.
Stop words: optional, so decide by meaning
There's no Google rule here. Mueller: "Whether you include stop words in them or not, decide to use numeric IDs, that's totally up to you" (SEJ).
So drop "the", "a" and "of" when the slug still reads cleanly, and keep them when removing them changes the meaning. Our post on refreshing old posts versus publishing new ones lives at refresh-or-publish-small-blog. Without the "or", the slug means nothing. The real failure is the other extreme, stacking variants like url-slug-slugs-seo-url-best-url. Google's starter guide calls keyword stuffing ("repeating the same words excessively, even in variations") a spam-policy violation.
Length: don't chase a number
You've probably seen the claim that shorter URLs rank higher. Backlinko's analysis of 11.8 million Google results found that #1 URLs are on average 9.2 characters shorter than #10 URLs. That's a correlation, and short URLs tend to belong to strong pages on strong sites.
A 2026 controlled study by seo-kreativ.de tried to separate those factors across 22,250 top-10 results for 2,862 keywords on google.de. Controlling for page type and other confounders shrank the length effect by 35.9%, to roughly 0.18 positions per 50 extra characters in the main model. The full model explained about 1.2% of position variance. The author suggests 31–60 characters for new URLs and explicitly advises against renaming existing URLs for length alone. It's one study of German SERPs, but the takeaway holds: length is a tidiness preference for new posts and never a reason to touch a live URL.
Trailing slash: pick one
Google treats /fish and /fish/ as separate URLs. Pick one, use it in every internal link and the sitemap, and redirect the other. Generators and hosts disagree on the default, so check what your build actually outputs.
That covers words, length and punctuation. The decision that most often forces a rename later is about time.
Dates and years: the rule that saves you a migration
If you started on GitHub Pages with Jekyll and never touched the config, you likely have dates in every URL already. Jekyll's default configuration sets permalink: date, which expands to /:categories/:year/:month/:day/:title:output_ext. The built-in none style, /:categories/:title:output_ext, drops the date. Setting it in _config.yml before your first post costs one line. Changing it after fifty posts is a migration. Most other generators leave dates out unless you add them. Hugo "does not add dates to URLs by default" (docs), and neither does Eleventy.
To be fair, dates don't hurt rankings. Mueller told a user in 2019, "Having a date in the URL is fine" (as reported by FlipWeb; the original tweet wasn't located). And Tim Berners-Lee's classic "Cool URIs don't change" actually recommends starting URIs with the creation date, since a date "is one thing which will not change."
So the case against dates is about maintenance and how the post looks, not a penalty:
- A year in an evergreen slug invites a rename.
/blog/best-crm-for-startups-2025/is wrong the day you refresh it for 2026. You either live with the mismatch or change the URL. - A full date makes a refreshed post look stale wherever the raw URL shows: desktop breadcrumbs, shared links, AI citations. You rewrote it last month, but the URL says 2021.
Berners-Lee's logic actually supports this. A date you never intend to change is permanent and harmless. The problem is a date that stops being true after a refresh.
The pattern big publishers use is to put the year in the title and leave it out of the URL. Ahrefs' "107 SEO Statistics for 2026" lives at ahrefs.com/blog/seo-statistics/. The title gets bumped every refresh, and the URL keeps every link it has earned.
Here's a confession. Of the twelve slugs on the first page of our own blog, two end in -2026 (best-llm-for-blog-writing-2026 and programmatic-seo-scaled-content-2026). Meanwhile "Do You Have to Disclose AI-Written Content? (2026)" uses the year-free disclose-ai-written-content. That's two conventions on one page, and it's exactly why the rule belongs in writing. Going forward, the year goes in the title.
Dated slugs are still right for content tied to a moment: changelogs, release notes, news, and annual reports you won't update next year. Ask whether you'll ever refresh the post in place. If you will, keep the date out of the slug.
Pin the slug so a title edit can't move it
Static generators create a quieter risk: the URL can change without anyone deciding to change it. Eleventy's docs show a permalink like "subdir/{{ title | slugify }}/index.html", which builds the URL from the title, so fixing a typo in the title moves the page. In Hugo, the URL follows the file path unless slug: in front matter overrides the last segment, so tidying up a filename renames a live page.
The fix is an explicit slug, written once:
---
title: "How to Fix 'Crawled — Currently Not Indexed' (2026)"
slug: "crawled-not-indexed"
---
Then add a check that stops a bad slug from merging. Tamás Sallai at Advanced Web Machinery added build hooks after his migration. One checks that each slug matches its slugified title, and another rejects duplicates. A CI step that checks for lowercase and hyphens, and rejects a four-digit year unless the post is tagged as dated, does the same job.
The cheapest moment to enforce this is while the post is still a draft. When posts reach your repo as pull requests, the slug is right there in the diff, which is the last point it can change for free. That's how Magic Share is built. Every draft passes schema checks on slug, title and description length, tags and target keyword before it's saved, then arrives as a PR. Merging publishes it, so you approve the slug before anyone can link to it.
Still, someday a slug will have to change: a product rename, a real typo, a post that changed scope. That's when the redirect plan matters.
If you change one anyway: the redirect plan for your host
Google's redirect documentation is clear. Server-side 301 and 308 responses mean a page "has permanently moved." An instant meta refresh is also treated as permanent, but server-side redirects are preferred, and JavaScript redirects are a last resort because "rendering may fail." What you actually get depends on your stack:
| Stack | How you declare it | What it serves | Gotcha |
|---|---|---|---|
| GitHub Pages + Jekyll | redirect_from: (jekyll-redirect-from) |
Meta-refresh HTML stub | No server config is generated, so no true 301 |
| Hugo | aliases: in front matter |
Meta-refresh stub by default | For real redirects, generate _redirects or .htaccess and set disableAliases = true |
| Astro | redirects in astro.config.mjs |
301 by default | A static HTML build outputs meta-refresh files instead |
| Next.js | redirects() (docs) |
permanent: true → 308 |
permanent: false gives a temporary 307 |
| Netlify | _redirects (docs) |
301 by default | Case-sensitive, and a rule won't override a file that still exists unless forced with ! |
| Cloudflare Pages | _redirects (docs) |
302 if you omit the code | Always write 301; cap of 2,000 static + 100 dynamic rules |
| Vercel | vercel.json (docs) |
308 permanent / 307 temporary | Up to 2,048 redirects |
Two rows deserve a second look. On Cloudflare Pages, /blog/old-slug /blog/new-slug with no code is a temporary 302, so your permanent rename depends on three characters you didn't type. And on Netlify, if a Hugo alias stub still sits at the old path, your _redirects rule won't fire. That's why Hugo pairs generated _redirects with disableAliases = true.
The redirect is only part of the job. This checklist draws on Google's URL-change migration guide and canonicalization doc, plus one static-blog lesson from the case study below:
- Map old to new, one line per URL.
- Update internal links. Google says to "link to the canonical URL rather than a duplicate URL." If you've built a pillar-and-cluster internal linking structure, check every spoke that points at the renamed post.
- Update the sitemap and
rel="canonical". Google rates redirects and canonicals as strong signals and sitemap inclusion as weak. - Keep RSS GUIDs stable. Jekyll's
post.idchanges with the slug, so feed readers would see the renamed post as new. - Keep the redirect. Google says "generally at least 1 year." Gary Illyes added: "(but try keeping them indefinitely if you can for your users)" (SEJ).
Expect some turbulence. Google warns of "ranking fluctuations while Google recrawls and reindexes", and says a small-to-medium site can take "a few weeks for most pages to move."
Why keep redirects forever? In 2023 Patrick Stox removed redirects from four Ahrefs posts. Referring domains to those posts visibly dropped while traffic barely moved, and he wasn't "willing to conclusively say that permanent redirects pass value even after one year." On a static site a redirect is one line of text, so there's no reason to delete it. Keeping it also protects everyone who cited you. It's the same link rot you'd catch from the other side when auditing your archive for dead outbound links.
A Jekyll blog that removed its dates and lost nothing
Advanced Web Machinery moved off Jekyll's dated permalinks on 15 December 2019. /2019/07/17/upload_signed_urls_differences/ became /differences-between-put-and-post-s3-signed-urls/. GitHub Pages can't send 301s, so the redirects were jekyll-redirect-from stubs with a meta refresh, a JavaScript fallback, a link, a canonical tag and noindex, all kept out of the sitemap.
Traffic dipped only over the Christmas holidays and recovered "like nothing happened." Sallai's verdict: "No traffic was lost because of the migration." The write-up shows a chart, not raw numbers, so treat it as one good outcome, not a guarantee. The lasting value is in what he built afterwards: the slug hooks and a separate rss_guid field. As he put it, "Changing URLs is risky and requires careful planning." That planning is far cheaper before the archive exists.
FAQ
Should I put the target keyword in my blog post URL slug?
Yes, if it's the natural description of the post. It helps readers and desktop breadcrumbs. Don't expect a ranking gain, and don't rename a live URL to add one.
Is it bad to have dates in blog URLs?
Not for ranking. Google has said dates in URLs are fine. The cost is maintenance: a year in an evergreen slug goes stale at the next refresh. Keep dates for changelogs, news and one-off reports.
Do I need to remove stop words from URL slugs?
No. Google calls it "totally up to you." Remove them when the slug stays clear, and keep them when they carry meaning.
How long should I keep a 301 redirect after changing a slug?
Google says at least a year, and Illyes suggests keeping redirects indefinitely. On a static host it's one line of config, so keep it forever.
Does GitHub Pages support 301 redirects?
No. jekyll-redirect-from serves meta-refresh stubs, which Google treats as permanent when instant. If you expect to rename often, a host with real 301s fits better.
The short version
Lowercase, hyphens, a few meaningful words, no year on anything you'll refresh, one trailing-slash convention, and an explicit slug: field enforced in CI. If you do rename, use a real 301 or 308 where your host allows it, write the status code explicitly, update internal links and canonicals, and never delete the redirect.
If you'd rather have that convention checked before a draft ever reaches your repo, Magic Share validates the slug and the rest of the frontmatter on every post, verifies every cited link, and opens a pull request you approve before anything goes live. Your first three posts are free. Plant your first post on the free plan.