On-page SEO audit: read ten checks of a page in one report

How to use the audit

Paste the full address of a page and click “Audit”. The server downloads the page as a browser would, follows redirects for up to ten hops like Googlebot, then reads the robots.txt of the final host. A single request is counted for the whole. The browser receives the useful signals and runs the checks on your side: the audit does not send you back the whole page.

The report opens with a summary that ranks findings by severity, blocking points first, then those to fix. Each line of the summary leads to its section further down. Each section has its verdict, the list of its findings and a link to the detailed tool, where you can dig in, paste code or test another address.

What the audit checks

Ten sections, built on the site’s page tools. The title and meta description tester supplies the title’s width in pixels and the description’s length in characters. The redirect checker gives the chain of HTTP codes, its loops, its temporary redirects and any move from https to http. The header analyzer reads X-Robots-Tag, compression, caching and security headers.

The indexability checker weighs the page’s status code, the robots.txt, the meta robots tags, the X-Robots-Tag header and the canonical, and concludes: indexable, to watch or not indexable. AI crawler access reads the same robots.txt and says which search, training or on-demand robots are blocked on this page. The hreflang checker validates language codes and whether the page declares itself.

The heading structure checker spots missing or multiple H1s, skipped levels and empty headings. The alt text checker counts images with no alt attribute, linked images with no text at all and suspect alt text. The link extractor counts internal, external and nofollow links, and flags empty anchors and links Google cannot follow. The structured data checker reads the JSON-LD, compares it with the properties Google lists and flags invalid blocks.

How to read the report

Four levels. Blocking: the point stops Google from indexing the page or breaks markup, such as a noindex, a 404 error or invalid JSON-LD. To fix: the point degrades the page without excluding it, such as a redirect chain, an image with no alt or a missing required property. To note: advice or information, such as a short title or a blocked AI robot. OK: the check finds nothing.

The score out of 100 sums up the page in one figure, and the summary by severity says what to handle first. A page can pile up twenty minor remarks and remain perfectly indexable, or have a single noindex and never appear: that is why a page excluded from the index is capped at 15, whatever the rest. The method is detailed below.

A blocked AI robot is classed as a simple remark. Blocking an assistant’s training or search crawler is an editorial choice. Only blocking Googlebot, handled by the Indexability section, becomes blocking.

How the score is calculated

The score out of 100 is the tool’s own scale: Google publishes no weighting between these checks. It follows the order of priority of its documentation: first the technical requirements, without which a page cannot be indexed, then what hinders crawling, then presentation best practices.

The page starts at 100 and each finding removes a number of points specific to its type. A missing title removes 14 points, a title that is too long 4, a missing meta description 3. A page that sends almost no text in its HTML loses 25 points, a canonical pointing to another address 12, a move from https to http 15, a temporary redirect 4. When the finding covers a number of elements, such as images without alt or links without anchor text, the deduction ranges from 40% of the weight for an isolated case to the full weight when every element is affected.

Each section has a cap, so that one repeated defect does not hide the rest: 45 points for indexability, 35 for HTTP codes, 18 for the title and description, 12 for links and for robots, 10 for structured data, 8 for hreflang and for alt text, 7 for headings, 6 for headers. A page excluded from the index, by a noindex, a 4xx or 5xx error, robots.txt or a redirect loop, is capped at 15: as long as it is excluded, the rest of the optimisation brings it nothing.

Below the score, the report details where the points went, finding by finding. Reading guide: 90 and above, very good; 75 to 89, good; 50 to 74, needs work; below 50, poor.

What Google asks for at minimum

Google Search Essentials are split into three parts: technical requirements, spam policies and best practices.¹ The technical requirements come down to three conditions for a page to be indexable: do not block Googlebot, return an HTTP 200 status code and offer indexable content.² Meeting them does not guarantee indexing, but failing them rules it out. That is why the audit puts the response code, the robots.txt and noindex directives at the top of the summary when they are a problem.

On best practices, the SEO Starter Guide recommends a title that is unique to each page, clear and concise, and a short meta description that works as a summary, while noting that the displayed snippet most often comes from the page content.³ It advises describing each image with alt text, choosing anchor text that tells readers about the target page and adding structured data to make the page eligible for rich results. On headings, it states that heading order does not matter for Google Search and that there is no ideal number of headings. The audit applies this nuance: a skipped level stays a simple remark.

On redirects, Googlebot follows up to 10 hops by default. A 301 or 308 redirect is a strong signal for which page to keep, a 302 or 307 a weaker one, and a 404 or 410 error gradually removes the page from the index.⁴ These are the thresholds behind the HTTP codes and redirects section.

In what order to fix

Start with the blocking points: a noindex, a 4xx or 5xx error, a robots.txt that forbids the page, invalid JSON-LD. As long as the page is not indexable, the rest of the optimization produces nothing, since Google’s technical requirements come before best practices.

Next come the points that hinder discovery and crawling: links Google cannot follow, redirect chains, a temporary redirect where a permanent one belongs. Google states that the vast majority of the new pages it finds each day are found through links,³ hence the value of internal linking made of real links with an href.

Then the care of the page itself: title and meta description, headings, alt text, required properties of structured data. Remarks at the “to note” level come last. They deserve a glance, rarely an emergency.

What the server does with the address

The script reads public pages only. It refuses an address that targets a private or local network, checks the IP address of each redirect before following it and gives up after eight seconds. It accepts only requests coming from the site’s pages, limits each visitor to 20 requests per 10-minute window and keeps neither the address nor the content. The value of cookies sent by the audited site is never returned: only the cookie name and its attributes feed the header analysis.

The limits of the audit

The audit reads the HTML as the server sends it, without running JavaScript. Headings, links or markup added by a script do not appear, and a page rendered in the browser shows up as a too-short served text alert in indexability. Broken link testing is not part of it: it takes one request per link, which the link extractor does for ten addresses of your choice. Checking that hreflang versions point back to each other requires reading every alternate version, which the hreflang checker does.

Lists are capped to stay light: 500 headings, 300 images, 3,000 links and 60 structured data blocks. The page is read up to 1.5 MB. The quota is 20 requests per 10-minute window and per visitor, shared with the site’s other tools.

Finally, an automatic audit reads signals and does not judge a strategy. It does not say whether the content answers search intent, whether inbound links are enough, or whether the page should exist. That work happens in the full SEO audit, which ties these findings to the pages that matter for your business.

After the audit

Take the section with a problem into its detailed tool: you can paste code there to test a fix before publishing it. The SEO tools hub gathers the other checks, from the robots.txt generator to the list comparer.

For a first look at the whole site, the free SEO audit by email applies these ten checks to the home page and ten other pages, adds the sitemap, the domain name and the hosting, and counts the pages affected by each problem.

Frequently asked questions