On-page SEO tools are the software that reads a single page and checks the elements that are up to you: title, meta description, headings, images, canonical, internal links and structured data. The basic kit costs nothing (Search Console's URL Inspection, the Rich Results Test, Lighthouse, a browser extension, Screaming Frog up to 500 URLs); the paid layer adds scoring checkers and suites such as Semrush or Ahrefs.
In the UK people mostly look for these tools as "checkers": "seo checker" gets 3,600 searches a month, "website seo checker" 2,400, "seo test" 480, "on page seo tools" 260 and "on page seo checker" 170 (average of September 2025 to August 2026, Google Ads data via DataForSEO, checked on 10 October 2026).
Most people use them the same way: paste the URL, look at the red warnings and try to clear every one of them.
What's wrong with that?
Plenty of those warnings also show up on pages that rank first on Google. On 10 October 2026 we ran 47 page-one results from Google UK through the on-page rules Semrush publishes: 16 of the 32 pages we could read would get at least one warning. That includes all four of the first results we could read.
We see it on our own projects too. Macropix, which makes custom LED screens in Milan, went from position 88 to 2 for "monitor pubblicitario" (advertising monitor) on Google Italy between 2020 and 2025. The work that moved the rankings was on-page, but no checker would have flagged it: getting the technical content out of the catalogue PDFs and giving each product line a page of its own.
Sunscape, which develops solar farms in Italy, went from 1 to 38 page-one keywords out of 143 in the six months from March 2026, with an architecture of pages by region. There too, the lever was the structure of the pages, not a checker warning.
In this guide you'll find:
- the seven elements an on-page tool checks, with the free and the paid tool for each;
- the test on 47 page-one results: which checker rules hold up and which don't;
- the real error we found on a first result, which a checker spots in two seconds;
- the order in which we use the tools, from free to paid.
Let's start.
What an on-page SEO tool checks (seven elements, and which ones you can check at no cost)
An on-page SEO tool compares a page's code with a list of rules: whether the title is missing, whether the meta description goes over a certain length, whether the images have alt text. It doesn't look at the links coming from other sites (for those you need off-page SEO tools) and it doesn't know which searches the page appears for.
The lists on page one of Google UK put all these tools together, from Search Console to Surfer, as if they did the same job. We've split them by element.
| Element | What to actually check | Free tool | Paid tool |
|---|---|---|---|
| Indexability | noindex, robots.txt, status code | Search Console's URL Inspection | Screaming Frog with a licence, Semrush Site Audit |
| Canonical | It points to the right page and that page returns 200 | URL Inspection (Google-selected canonical), Detailed extension | Paid crawlers |
| Title and meta description | They exist, they're unique, they describe the page | Extensions such as Detailed or META SEO inspector | Seobility and Semrush checkers |
| Headings | The page has one clear main heading | Browser extensions | Suites and checkers |
| Images | Alt text on images that carry information | Lighthouse | Crawlers and suites |
| Structured data | The markup is valid and matches the visible text | Rich Results Test | SEO plugins, crawlers with custom extraction |
| Content | It answers the search and can be read without JavaScript | URL Inspection (rendered page), comparison with the SERP | Content optimisers such as Surfer or Clearscope |
So the free tools cover everything?
Almost. You can check the first six rows without spending a penny, one page at a time. You pay when the pages run into the hundreds, because you need a crawler, or when you want an automatic comparison with competitors. The full comparison of the suites, with price lists, is in our roundup of the best SEO tools.
The test: 47 Google results run through a checker's rules (16 of 32 pages get a warning)
Of the 32 page-one pages on Google UK that we managed to read, 16 break at least one of Semrush's on-page rules. We'd already measured the overall score in our guide to the SEO score; here we went one level down, rule by rule.
The method. On 10 October 2026 we pulled the mobile Google UK results from DataForSEO for six different searches: "first time buyer mortgage", "running shoes", "tiramisu recipe", "car leasing", "robot vacuum" and "resignation letter". We kept the organic results that are HTML pages (leaving out videos, social posts and PDFs): 47. Then we downloaded each page from our server, without running JavaScript, the way an online checker does.
We couldn't read 15 of the 47 pages. Seven answered 403 (including MoneySavingExpert, MoneyHelper, adidas, Currys and Broma Bakery, first for "tiramisu recipe"), Start Fitness answered 429, GoCompare served a bot check, Lex Autolease served a "Page not found" even for its home page, and Sports Direct didn't respond within 25 seconds. The other four either failed on the connection (Runner's World, Good Housekeeping, Windsor Hill Mortgages) or, in Amazon's case, returned a stripped-down page with no title. These are anti-bot protections: Google sees the pages, an outside tool doesn't.
On the remaining 32 we applied the thresholds Semrush publishes in its list of Site Audit checks, plus two very common thresholds.
| Checker rule | Who uses it | Pages that break it (out of 32) | Of which first results (out of 4) |
|---|---|---|---|
| Title over 70 characters | Semrush | 3 | 0 |
| Title over 60 characters | Ahrefs gives it as a safe limit | 11 | 2 |
| Missing meta description | Semrush, Lighthouse | 0 | 0 |
| Meta description over 156 characters | Yoast gives it as the typical length | 8 | 0 |
| Missing H1 | Semrush | 3 | 1 |
| More than one H1 | Semrush | 2 | 1 |
| Images without an alt attribute | Semrush | 12 | 2 |
| Fewer than 200 words | Semrush | 1 | 1 |
| At least one Semrush warning | 16 | 4 |
The test tells us four things.
One: images without alt are the most common warning, on 12 pages out of 32. On Select Car Leasing's guide, fifth for "car leasing", 17 of 88 images have none. On Runners Need, first for "running shoes", it's 3 out of 1,128.
Two: only 2 pages have more than one H1, but one of them is a first result. HomeOwners Alliance, first for "first time buyer mortgage", has four.
Three: the H1 is missing on three pages, and one is a first result. Adobe Express is first for "resignation letter" without an H1. The iRobot UK category page, ranking for "robot vacuum", and a 2003 BBC News article on how to draft a resignation letter are in the same position.
Four: only one page falls below 200 words, and it's the Adobe page, which has no text at all in its HTML. The next shortest, FormTemplate, has 476 counting menus and footer.
So the checker rules are wrong?
No. They're generic: they apply to millions of sites, and on a page that already ranks they say little. The problem is treating them all as equally urgent.
The warnings you can leave at the bottom of the list (title length, multiple H1s, decorative alt text)
Three of the most common warnings concern things on which Google's documentation is explicit.
Title length. Google's guide to title links says there's no limit on how long a title element can be: in results it's truncated as needed, typically to fit the device width. The checkers' limit is an estimate of how much gets shown, not a ranking rule.
Our sample bears this out. Leaving out the Adobe page, which Google shows without a title (more on that below), Google rewrote the title of 7 pages out of 31, and 3 of those 7 had a title under 60 characters. Of the 11 titles over 60 characters, 7 appeared in full, truncated or with just the brand name removed. Why Google rewrites titles, and how often, is in our guide on Google ignoring the meta title.
Multiple H1s. Google's SEO starter guide says that having headings in semantic order is fantastic for screen readers, but from Google Search's perspective the order doesn't matter, and there's no ideal number of headings. Semrush itself admits, in its list, that multiple H1 tags are allowed in HTML5. How to structure them well for readers is in our guide to heading tags.
Alt text on decorative images. An icon or a graphic divider should have an empty alt, so screen readers skip it. Many checkers count any image without alternative text as a problem, and that warning goes to the bottom of our list. The alt text that matters is on images that carry information, like a product photo: we cover it in our guide to image optimisation.
And a meta description that's too long?
Same story. Google says there's no limit on how long a meta description can be and that the snippet is truncated to fit the screen. Yoast's threshold, about 155 characters, is a guide to what people can read in the results. Useful when you write, not a reason to worry.
The errors that do cost (Adobe's first result for "resignation letter" has noindex and a robots.txt block)
The warnings to fix straight away are the ones that can take a page out of Google or shift its signals to another address: noindex, a wrong canonical, a robots.txt block, content the server doesn't return.
In the test we found one on a first result.
An Adobe Express template page is the first organic result on Google UK for "resignation letter". Its HTML carries a meta robots noindex tag. And the robots.txt of new.express.adobe.com, on 10 October 2026, blocked every crawler from almost the whole subdomain ("Disallow: /", with exceptions only for embed links).
The two instructions cancel each other out. Google's page on blocking indexing explains that if robots.txt stops the crawler, it never sees the noindex rule, and the page can still appear in search results. That's exactly what happens: the result shows the bare URL instead of a title and the snippet reads "No information is available for this page". Whatever Adobe intended, it holds first place with a result nobody is tempted to click. The other classic costly error, a canonical pointing to the wrong address, is covered in our guide to the canonical URL.
Would a checker have spotted it?
Yes, in two seconds. It's the kind of error these tools are genuinely good for: binary, verifiable, on a page that brings in traffic.
The second costly error is the one no checker reports as an error: content Google can't use. On Macropix there wasn't a single red warning. The problem was that the technical specifications sat in the catalogue PDFs and not on the pages. Moving them onto the pages, and giving each product line its own page, is the work that took "monitor pubblicitario" and "schermo pubblicitario" (advertising screen) to second place, as we describe in our manufacturing case study.
Your audit has 200 warnings and you don't know which one matters?
A page that tells Google two contradictory things is worth more than a hundred missing alt attributes. We cross-reference your site's errors with the pages that bring in clicks and tell you which to fix this week and which to leave where they are.
Book a strategy consultationGoogle's own tools (the only ones that see the page the way Googlebot does)
Google's own tools, which cost nothing, are the only ones that show the page as Google processes it, JavaScript included. All the others are estimating.
Search Console's URL Inspection. According to the official help page, it shows the indexed version of the page, the canonical Google has selected and, in the live test, a screenshot of how Google-InspectionTool sees the page, along with the HTML it received. If a checker says the title is missing and URL Inspection can see it, the second one is right. How to read the rest of the reports is in our guide to Google Search Console.
Rich Results Test. At search.google.com/test/rich-results it checks the markup and tells you which rich results the page is eligible for. What structured data still does in 2026 is in our guide to schema markup.
Lighthouse and PageSpeed Insights. Lighthouse's SEO category runs nine checks, including meta description, canonical, HTTP status code and robots.txt; alt text is checked in the accessibility category. It has a limit worth knowing: it measures one page at a time, and speed doesn't count towards the SEO score.
Isn't Search Console enough on its own?
For one page, yes. To find which page out of three hundred has the wrong canonical, you need a crawler.
Extensions and crawlers: one page at a time or the whole site (Screaming Frog costs nothing up to 500 URLs)
An extension reads the page you have open in your browser; a crawler visits every page on the site by following links, as Googlebot would.
Extensions run in your own browser, and that solves the problem of the 15 blocked pages. An online checker downloads the page from its server and can get a 403, as we did; an extension reads what you see. In our check on the Chrome Web Store on 29 September 2026, Ahrefs' Detailed SEO Extension had 600,000 users, META SEO inspector 200,000 and SEO Minion 200,000. The full list, with ratings and update dates, is in our guide to Chrome extensions for marketing.
Screaming Frog SEO Spider is the most widely used crawler: the free version crawls up to 500 URLs, then a licence costs £199 a year per user (price list checked on 10 October 2026). It runs on your computer and pulls the titles, H1s, canonicals and status codes of the whole site into one table. How to set it up is in our guide to Screaming Frog.
The rule we use: under 500 pages, an extension plus the free crawler are enough. Above that, the licence pays for itself at the first migration.
Online checkers and suites: when to pay (and how much)
Online checkers give one page a score; suites do the same across the whole site and add keywords and competitors.
- Seobility: the SEO Checker allows 5 checks a day with a Basic account, which costs nothing; the Premium plan is €59.90 a month plus VAT (Seobility lists its prices in euros; checked on 10 October 2026).
- Semrush: without a subscription, Site Audit crawls 100 pages a month; the SEO plan costs $117.33 a month billed annually, $139 billed monthly.
- Ahrefs: Ahrefs Free runs Site Audit on sites you can verify as your own; the Lite plan costs $129 a month before VAT.
Semrush and Ahrefs prices checked on 10 October 2026 on their official pages; both list them in US dollars (at the Bank of England rate for 8 October 2026, $1.3226 to the pound, that's about £88.71 and £97.54 a month).
Then there are content optimisers such as Surfer and Clearscope, which on Google UK appear both in the AI Overview for "on page seo tools" and in SEOptimer's list on page one. They compare your text with the pages on page one and suggest terms to add. They're for people who publish a lot; prices and alternatives are in our guide to Surfer SEO alternatives.
What about the WordPress SEO plugin?
Yoast or Rank Math act as a checker while you write, with the traffic lights. Useful so you don't forget the meta description; less so for deciding what matters, as we saw in our test of SEO plugins for WordPress.
How we use them (five steps, in this order)
The order matters more than the tool. This is the one we follow when a page isn't performing as it should.
- Search Console's URL Inspection: is the page indexed? Is the canonical Google selected the one you declared?
- Browser extension: title, H1, canonical and noindex in ten seconds, on the real page.
- Comparison with the SERP: we open the top three results for the search we care about and measure the same elements. That's the threshold that counts, not the tool's.
- Crawler: if the error comes from a template, we look for every page that repeats it.
- Scoring checker: last, to follow the trend over time.
On Sunscape, the lever that took it from 1 to 38 page-one keywords was an architecture of pages by region: a choice that came from comparing with the SERP, not from a score. It's the kind of diagnosis we do in our SEO service.
Frequently asked questions
Yes. For one page at a time, Search Console's URL Inspection, the Rich Results Test, Lighthouse and an extension such as Detailed SEO Extension are enough. Screaming Frog costs nothing up to 500 URLs: above that, a licence is £199 a year.
No. Google doesn't use third-party tools' scores. In our test of 10 October 2026 on Google UK, all four of the first results we could read broke at least one of Semrush's on-page rules.
No. Google says there's no limit on how long a title can be and that in results it's truncated to fit the screen. In our sample, 3 of the 7 titles Google rewrote were under 60 characters.
Not for Google. Its SEO starter guide says there's no ideal number of headings and that their order matters for screen readers, not for Search. In our test, 2 of 32 page-one pages had more than one H1, and one of them ranked first.
Usually because of anti-bot protection that blocks requests coming from the tools' servers. In our test, 15 of 47 pages returned an error, a bot check or no usable page. In that case, use a browser extension or Search Console's URL Inspection.
It depends on how much traffic the page gets, but the order stays the same. First noindex, robots.txt blocks and wrong canonicals, then status codes and content Google can't see, and last title and meta description length and alt text on decorative images.
On-page SEO tools in five lines (and the threshold no tool knows)
The kit that costs nothing covers almost everything: Search Console, the Rich Results Test, Lighthouse, an extension, Screaming Frog up to 500 URLs.
16 of 32 page-one pages break at least one Semrush rule.
Title length, multiple H1s and decorative alt text can wait.
Canonical, noindex, robots.txt and content kept off the page can't.
An online checker can't read about one result in three.
The test left us with one last point. The checkers' thresholds are the same for a recipe and for car leasing, but the pages that win aren't: for "tiramisu recipe" none of the five readable pages has an image without alt and no title goes past 56 characters; for "car leasing" four of the seven have images without alt and three titles go past 60. The right threshold for your page isn't in the tool: it's in the pages ranking above you today. Measuring them takes half an hour with an extension, and it's the only comparison that tells you anything about Google.
Until next time!
