SEO on page analysis is the systematic review of everything on a single URL that influences how search engines understand and rank it — content, intent match, headings, title and meta, internal links, structured data, and page experience. Done properly, SEO on page analysis tells you in about 20 minutes per page whether a URL is losing on substance or on mechanics, which are two very different fixes.
Here is the ranked version of the answer, because most checklists present 40 items as equally important and they are not. Intent match and content depth account for the majority of on-page ranking outcomes. Title tags and internal links account for most of the rest. Everything else is hygiene that prevents losses rather than creating gains.
Start with intent match, because nothing else compensates
Before checking a single tag, ask whether the page format matches what the SERP rewards. Search your target keyword and look at the top ten. If nine of them are comparison pages and yours is a how-to guide, no amount of tag optimization will fix that. Rewrite the format or change the target.
Three mismatches show up constantly: a product page targeting an informational query, a blog post targeting a transactional one, and a broad overview targeting a query that clearly wants a specific answer. Each takes a rewrite, not an edit.
Audit content depth against the weakest page-one result
Open positions eight, nine, and ten — the least impressive pages currently on page one. Compare your page against those, not against the leader. Count subtopics covered, check whether they include original data or examples, and note their approximate word count.
If your page covers fewer subtopics than the weakest result, that is your gap and it outweighs everything below. If it covers more and still ranks worse, the problem is usually links or intent, not on-page.
Depth is not word count. A 900-word page that answers the question completely beats a 2,500-word page that circles it. Check for the specific things thin pages lack: concrete numbers, named examples, a stated process, and answers to the obvious follow-up questions.
The 12-point mechanical checklist
Once substance is settled, run these in order. Each is fast and each has a clear pass or fail:
- Title tag — under 60 characters, primary keyword near the front, written to earn a click.
- Meta description — 140 to 155 characters, keyword used naturally, describes the value not the contents.
- One H1, matching the topic. Multiple H1s or none is a common CMS-template failure.
- Heading hierarchy — H2s for sections, H3s only where genuinely nested. No heading tags used for styling.
- Opening answer — the core question answered within the first 60 words, in a self-contained sentence.
- URL slug — short, descriptive, lowercase, hyphenated, no dates or parameters.
- Internal links in — at least three relevant pages linking to this one with descriptive anchor text.
- Internal links out — links to related pages that genuinely help the reader continue.
- Image alt text — descriptive, not keyword-stuffed, present on every meaningful image.
- Canonical tag — self-referencing unless you intend otherwise, and pointing at an indexable URL.
- Indexability — no accidental noindex, not blocked in robots.txt, present in the sitemap.
- Structured data — Article, Product, FAQ, or Breadcrumb as appropriate, validating without errors.
A page passing all twelve with strong content and still stuck below position 20 usually has an off-page problem. That is a different investigation.
Cannibalization: the check most audits miss
Two of your own pages competing for the same query is one of the most common causes of a stuck ranking, and no single-page checklist catches it. Search your site for the target keyword and see how many of your URLs Google considers relevant. If two rank alternately for the same term, they are splitting signals.
The fix is usually consolidation: merge the weaker page into the stronger, redirect it, and update internal links to point at the survivor. Occasionally the right fix is differentiation — narrowing each page to a distinct intent. What almost never works is leaving both and hoping.
Page experience: where it matters and where it does not
Core Web Vitals are a real ranking input and a small one. A page that loads in 1.2 seconds does not outrank a better page that loads in 2.4 seconds. But a page with a Largest Contentful Paint above four seconds on mobile, or visible layout shift on load, is losing users before it loses rankings.
Check field data rather than lab scores where you can — real-user measurements from actual visitors reflect what people experience, while lab tests reflect a simulated device on a simulated connection. Fix the outliers, ignore the difference between a score of 92 and 98, and spend the time saved on content instead.
Doing this at scale without losing a week
Twenty minutes per page is fine for ten pages and impossible for four hundred. At scale you need a crawl that flags the mechanical failures site-wide, so human attention goes only to the pages that need judgment.
A useful crawl surfaces the actual evidence, not just counts: the specific duplicate title text, the URL missing an H1, the exact meta description that runs to 210 characters. Counts tell you a problem exists; evidence tells you what to type.
SEO Rocket runs an instant quick scan of roughly 25 pages for a fast read, and deep full-site crawls verified past 900 pages, with concrete per-issue evidence — real titles, URLs, and H1 text — plus PageSpeed and Core Web Vitals including real-user field data. Whatever tool you use, insist on evidence-level output; anything less means you re-do the diagnostic work by hand.
Prioritising the fix list
A site-wide audit typically returns hundreds of items. Sort them by traffic at stake rather than severity label. A missing meta description on a page with 8,000 monthly impressions is worth more than a broken canonical on a page nobody visits.
A practical order of work: pages ranking 5 to 15 with mechanical failures first, because they are closest to a real gain; then pages with high impressions and a click-through rate under 1.5 percent, which are usually title-tag problems; then indexability errors anywhere, because those are pure loss; then everything else in traffic order.
Measuring whether the analysis worked
Record positions before you change anything, then check at 14, 30, and 60 days. Title-tag changes often show within two weeks. Content expansion typically takes four to eight. Internal link changes need a recrawl of the linking pages first, so allow a month.
Read trends, not single readings — a swing of two or three positions between checks is ordinary noise, and treating it as a result leads to reverting changes that were working. Use Search Console as ground truth for your own pages, since third-party position and volume figures are modeled estimates built from periodic crawls. Change one variable per page where you can, and you will learn what actually works on your site rather than what works in general.
