A URL is read by people more often than you think
URLs are read out loud, pasted into messages, typed from memory and printed on packaging. That is the argument for making them readable, and it is a better argument than any of the ones about ranking.
The short answer
A readable URL uses lowercase letters, hyphens between words, no session or tracking parameters, and a path that describes the page. None of that is why a page ranks. It is why a URL survives being pasted into a message, read aloud, or shown truncated in a result, and the rule that matters most is to leave working URLs alone.
Five rules, and none of them is why a page ranks
The conventions are settled and short. Lowercase, because some servers treat case as significant and two casings of one path are two URLs. Hyphens between words, because underscores were historically not treated as separators and the habit stuck. Words rather than identifiers, so the address describes the page. Short enough to be pasted without wrapping. No file extension, since .php and .aspx tell a reader what you built the site with and commit you to it.
awkward better
----------------------------------------- -----------------------------
/Products/Knives/ITEM_2000.aspx?id=4471 /knives/carbon-steel-chef
/blog/index.php?p=8812 /blog/sharpening-carbon-steel
/en_GB/produkt/messer/kn%2D2000 /uk/knives/kn-2000
/guides/how_to_sharpen_a_knife /guides/sharpening
/c/47/s/12/p/8812 /knives/chef/kn-2000The right column is nicer in every respect that involves a human being: it survives being read aloud, it tells somebody what they clicked before the page loads, and it is guessable. What it does not do is rank better than the left column for reasons of its own. Google has said the words in a URL are a very small factor, and the honest reading of that is that this check is about people.
One genuine mechanical benefit worth knowing: a readable path is what appears in your breadcrumb in a search result, so it is visible to somebody deciding whether to click. That is a small effect on a real surface, which is more than most of the claims made for this check.
The rule nobody states: leave working URLs alone
Every guide about URL structure ends with a list of improvements and stops. The missing sentence is that changing an existing URL costs more than almost any benefit on that list.
What a rename actually involves. A redirect from the old address, which now has to live forever because you cannot know who linked to it. Every internal link updated, or you have a site linking to its own redirects. Your sitemap regenerated. Whatever accumulated behind the old URL passing through a hop, which mostly works and is not free. And a period afterwards where a search engine is working out that the two addresses are the same page.
Then the second rename, in two years, by somebody who reads a guide like this one, stacking a hop onto the first. That is where chains come from, and it is the strongest practical argument for restraint.
Three situations where changing a URL is genuinely worth it.
The path is actively misleading, because the page moved subject or the product was renamed and the address contradicts the content. The URL leaks something it should not, an internal identifier, a staging path, an email address. Or you are restructuring for a different reason already, and the URLs come along with a migration you were doing anyway.
Everything else, including capitals, underscores and an ugly numeric identifier, is a preference. Apply the rules to new URLs, where they are free, and let the old ones be.
Read the shape of one URL
This looks at the path of the URL you enter and reports the mechanical faults: capitals, underscores, encoded characters, a file extension, a numeric identifier, excessive length. All of them are patterns, which is exactly what a machine can see about a URL.
No signup required. Each free search audits one page, paste any URL to see it in action.
What it deliberately does not do is tell you whether the words describe the page. That is a judgement, it needs the content, and a tool that guessed at it would be inventing findings.
Does this affect AI search?
Slightly, and in one specific way: the URL is often the only thing shown when a source is cited. A chat answer with a link, a footnote, a citation list. In that position the address is doing the job a title tag does in a search result, and /c/47/s/12/p/8812 tells the reader nothing about whether it is worth following.
There is a second and smaller effect. A path made of words is a few more words describing the document, in a place that is always available even when the page is truncated. It is a weak signal and it exists.
What has no basis is the idea that keywords in a URL help you get retrieved. Retrieval works on the content of passages, and no amount of care in an address compensates for a page that does not answer anything. Treat this as being about the moment a person sees your link.
Why a readable path is the definition of a Refinement
Everything about this check argues for the bottom of the scale. The page works. It is crawled, indexed and ranked. The influence on ranking is acknowledged to be very small. And the recommended action on an existing site is frequently to do nothing at all, which is not a property shared by any finding that matters.
It stays in the catalogue because URLs are permanent in a way little else on a page is. A path chosen carelessly today is a path you are stuck with, or a redirect you maintain forever. The value of the check is at the moment of creation, and a report is a reasonable way to notice that your new templates are producing addresses you will regret.
The rating also protects you from the check. Rated higher, it would authorise exactly the migration the previous section argues against, and the traffic risk in that migration is larger than anything a tidy path would earn. A refinement is the honest rating for work whose main hazard is being taken too seriously.
Questions this check raises
- Do keywords in the URL help rankings?
- Marginally at best, and Google has described the effect as very small. The real benefit is human: a path that names the page tells someone what they are about to open before they click it, and it survives being shared without context. Rewriting a working URL to add a keyword costs more than the keyword is worth.
- Should I use hyphens or underscores in URLs?
- Hyphens. Google treats a hyphen as a word separator and an underscore as a joiner, so page_about_us reads as one token and page-about-us reads as three. The distinction is old and has not changed, but it only matters on new URLs: changing existing ones for this reason is not worth the redirect.
- Is it worth changing an existing URL to make it cleaner?
- Usually not. A rename means a redirect, a period of split signals, every external link pointing at the old address, and a real risk of breaking something. The exception is a URL that is actively harmful, such as one carrying a session ID or leaking internal identifiers. Cosmetic improvements to working URLs are the definition of a change that costs more than it returns.