Frequently asked questions

The questions we receive most often about the validator and about markup validation in general. If yours is not here, the contact page is the place to send it.

Is this validator free?

Yes, entirely. No account, no subscription, no email address to hand over. Two technical limits apply: a single run stops at 1500 pages, and each visitor may start 80 analyses per 24 hours. Both exist to keep the service responsive, not to push you towards a paid version, because there isn't one. Advertising pays for the hosting.

Why does the page keep refreshing while it works?

Because shared hosting caps how long a single request may run, usually at thirty seconds or so. Checking a hundred pages takes far longer than that, so the work is cut into short steps: each refresh handles the next pages and saves its progress. That is also why the tab has to stay open. Closing it simply ends the run, with no consequence other than losing the results collected so far.

Why were some of my pages not analysed?

The crawler starts at the address you type and follows the links it finds in the HTML, staying inside the same domain. A page is therefore invisible to it if nothing links to it, if the only route is a form or a menu built by JavaScript after loading, or if it sits behind a login. Files that are not web pages, images, PDFs, archives, are deliberately skipped. If an important section is missing from the results, start a second run directly from that section's address.

What is the difference between an error and a warning?

An error means the markup breaks a rule of the specification: an element left unclosed, an attribute that does not exist on that element, two identical identifiers, a table whose columns do not add up. A warning means the markup is legal but suspicious or outdated, such as an empty heading or a missing language attribute. Fix the errors, then look at the warnings one by one and decide. Both categories are detailed on our common errors page.

Does valid HTML improve my ranking?

Not on its own, and anyone claiming otherwise is overselling. The error count is not a ranking signal. What can cost you is the consequence of certain errors: a broken structure changes how a crawler reads your content, and malformed structured data loses you rich results entirely. The full argument, with the list of what matters and what does not, is on HTML validation and SEO.

I fixed one thing and two hundred errors disappeared. Is that normal?

Entirely normal, and it is the usual shape of a first clean up. Sites are built from shared templates, so one forgotten closing tag in a header appears on every single page. This is precisely why checking a whole site at once is more useful than checking one page: seeing the same message repeated across the list tells you the problem lives in a template, not in the content.

Can I check pages that require a login?

No. The crawler arrives as an anonymous visitor with no session and no cookies, so it sees exactly what a logged out user sees. For a private area, use a validation extension running in your own browser, where your session already exists.

Does the tool correct the errors?

It finds them and tells you where they are. Correcting markup automatically is a bad idea in practice: an unclosed div can be fixed by closing it in three different places, and only the author knows which one was meant. Each result links to a detailed report giving the line number and the reason, which is usually enough to make the fix obvious.

Why do some pages return an API connection error?

The markup checking engine is a shared public service. When it is busy, or when a page is unusually large, the request exceeds our timeout and comes back empty. It says nothing about your page. Re run the analysis later and those pages will normally go through.

Do you keep my results or my URLs?

No. The progress of a run is written to a temporary file so it can survive the refreshes, and that file is deleted as soon as the run finishes, or automatically after two hours if you abandon it. Nothing is archived, no list of addresses is kept, and no report is stored. What the site does collect, like any site with advertising and audience measurement, is described on the cookies page.

Can I validate a site that is not online yet?

Not with this tool, since it fetches your pages over the public web like any visitor. A local address or a private network address is refused for security reasons. Either publish to a staging address that is publicly reachable, or use a validator installed on your own machine.

Should I aim for zero errors everywhere?

On your templates, yes, because those errors are multiplied by your number of pages and the fix is made once. On individual pages, use judgement. A warning you have read and understood, and decided does not apply to your situation, is a closed subject. Chasing a perfect report on a large site can consume a week that would be better spent on content or loading speed.

Will the crawler slow my server down?

Two pages per step, with a pause before each validation request, which is considerably gentler than a search engine crawler passing through. Any normal hosting handles it without noticing. If your server is rate limiting aggressively, you may see a few connection errors, which is a sign of your protection working rather than a problem with the tool.

Does it check CSS and accessibility as well?

It checks HTML markup. Stylesheets are validated by a separate service, described on our CSS validation page. Accessibility overlaps with markup validity but goes well beyond it, and a large part of it can only be checked by hand, which is the subject of validation and accessibility.