Free on-page SEO checker

How the number is worked out

Each of the twenty-one checks scores good, worth a look, or needs fixing. Every check belongs to one of four groups, and each group carries a weight: can it be found counts seven, how it looks in results counts four, the page itself counts three, and technical health counts two. The group results are multiplied by those weights, added together, and expressed as a score out of 100.

Benchmarks worth comparing against

Two ways in, and exactly what each one does

Most free SEO checkers ask for a URL and say nothing about where it goes. This one gives you both options and tells you the difference, because the difference is real.

Give it an address and this site fetches the page for you. That step has to happen on a server: a web page is not allowed to fetch another site’s page and read it, because the browser’s same-origin policy forbids exactly that, which is what stops any site you visit from quietly reading your webmail. So the address is posted here, the page is fetched, the HTML comes back to your browser, and the twenty-one checks run there. The address is not logged, the page is not stored, and the report is never written down anywhere.

Paste the source and none of that happens. Nothing is sent, because there is nothing to send: the checks read the text in the box and the whole thing stays in the tab you already have open.

Pasting also does something no fetcher can. It works on a staging site behind a login, on a page that has not shipped yet, on a local build, and on any page that blocks automated requests. Those are the moments when finding a stray noindex tag is worth most.

What the fetcher will and will not open

A tool that fetches any address a stranger types is a well-known way to get a server used against itself, so this one is deliberately narrow, and it is worth publishing what it refuses rather than discovering it as an error message.

• Public http and https pages only, on ports 80 and 443. Not other schemes, not other ports.
• No addresses that resolve to private, loopback, link-local, carrier-grade NAT or reserved networks, and none of the hosting-provider metadata services.
• No addresses carrying a username or password, which are both a security problem and a common way to disguise where an address really points.
• Redirects are followed up to three times, and every hop is checked again — an address that redirects somewhere private is refused at the hop, not waved through because the first address looked fine.
• A limit on fetches per visitor per hour, and a second one for the site as a whole, so this stays available rather than becoming somebody’s crawler.

If a refusal gets in your way, paste the source. That mode has no limits and no guard list, because it makes no request.

How the score is put together

Twenty-one checks, four groups, four weights, and the same twenty-one either way you give it the page. Can it be found carries seven, because a page search engines have been told to ignore cannot be rescued by a better title. How it looks in results carries four. The page itself carries three. Technical health carries two.

Each check returns one of three states, good, worth a look, or needs fixing, and the weighted total is expressed out of 100. Read the groups in order rather than chasing the number.

A score of 71 with a clean first group is a page that needs an afternoon of tidying. A score of 71 with a failure in the first group is a page that is invisible, and every point earned in the other three groups is currently worth nothing.

What this check cannot tell you

• Anything JavaScript adds. The checks read the HTML as it arrives, whether fetched or pasted. Content that only appears after a script has run is not counted, which matters on single-page applications — on those, view the source in your browser and paste what you see there instead.
• Core Web Vitals. Those are field measurements taken in real browsers over real visits. Google’s good thresholds are 2.5 seconds for Largest Contentful Paint, 200 milliseconds for Interaction to Next Paint and 0.1 for Cumulative Layout Shift, all judged at the 75th percentile of page loads. No single copy of a page’s source can produce any of them.
• Backlinks, rankings, competitors or search volume. None of it is in the HTML, so none of it is in the report.
• The rest of the site. One page at a time. A broken link four levels deep will not appear here.
• Whether the page deserves to rank. A page can pass all twenty-one checks and still sit behind one that answers the question better. The checks are the floor, not the ceiling.

What to fix first

Work the groups in order. Anything in the first group is a same-day fix, because it is the difference between existing in the index and not. Title and meta description come next, since they decide the click you get from a position you already hold. Headings and alt text follow. The charset tag can wait until Friday.

There is no sign-up, no email wall and no stored report either way. If you paste the source, nothing leaves your machine at all; if you give it an address, that address is used once to fetch the page and then forgotten.

When to stop checking and talk to someone

Stop when the report is clean and the page still is not working. That is the point where the remaining problem is competitive rather than mechanical, and no checker of any size will find it for you. Stop, too, when the same issue keeps reappearing after every deploy, which is a publishing-process problem rather than an SEO one and belongs with whoever builds the site: ours is the website development team.

If the first group failed and you are not sure why, our guide to fixing Google Search Console errors covers the coverage and indexing reports where those problems usually surface first. For the wider picture, what SEO services actually include sets out the work that follows a clean audit, and our digital marketing service is where that work gets done.

If the page is mechanically fine but the site around it is not, the website ROI and payback calculator puts a number on what a rebuild would be worth before you commit to one.

Or run the check, then send us the score. A free consultation is a better use of an hour than a second opinion from a second tool.