On page SEO analysis: how to score one page against the query it is meant to win

A seven check sequence for a single URL, the free tools that answer each check, and a scorecard that ends in a verdict rather than a number out of 100.

Published by AI Knows Us · Updated 12 October 2026

An on page SEO analysis scores one page against the one search it is meant to win. Run the checks in this order: the intent match against the results page, the title and the H1, the coverage of the sub questions a searcher expects, the indexing basics for that exact URL, the structured data, the internal links pointing at it, and page experience. The order matters because a page that answers the wrong question gains nothing from perfect tags, and an unindexed page gains nothing from anything. Three free tools cover every check: Google Search Console, the Rich Results Test and PageSpeed Insights, plus your browser's view source.

This page covers the analysis of a single page. For site wide problems such as redirect chains and orphan pages, see what a free SEO checker tells you, and for the fixes themselves see the eleven on page techniques in order.

Step zero: name the one query

Write down the single search this page exists to win, and the audience behind it. A page analysed against three queries produces three contradictory verdicts, which is why so many audits end in a list nobody acts on. If the page genuinely serves two different intents, the finding is that it should be two pages, and that is often the largest gain available.

The seven checks, in order

1. Intent match

Open an incognito window, run the query, and read the top five results as a set. Note the format they share, the depth, and the sub questions they all answer. That is the specification Google is currently rewarding. Score your page against it honestly: same format, same depth, same questions covered. A mismatch here outweighs every other finding on the list.

2. The title and the H1

Read the title tag in view source, not the one your CMS preview shows. The query, or the natural language form of it, belongs near the front. The H1 should say the same thing in human words. Google documents how it chooses the link text it displays in its guidance on title links, and the practical reading is that a clear, specific, unique title is what survives rewriting.

<title>On page SEO analysis: a seven check sequence</title>
<meta name="description" content="One page, one query, seven checks and a verdict.">
<link rel="canonical" href="https://example.com/on-page-analysis">
<meta name="robots" content="index,follow">

Those four lines answer four questions at once: what the page is called, what the snippet says, which URL is the real one, and whether the page is allowed to appear at all.

3. Coverage

List the sub questions a person with this query also has. Then mark each one as answered, mentioned or missing on your page. Missing entries are the content work, and they are usually worth more than any tag edit. A page that answers the main question and nine follow ups is the page an engine can quote from.

4. Indexing basics for this URL

Use the URL Inspection tool in Google Search Console on the exact address. You are checking four things: that Google has it indexed, that the canonical Google chose matches the one you declared, that the fetched HTML contains your main content, and that no robots directive is blocking it. Google's robots meta tag specification is the reference for what each directive does.

5. Structured data

Run the page through the Rich Results Test. The useful outcome is not a badge, it is confirmation that the markup you intended is present, valid and consistent with the visible page. Mark up what the page actually is: an article, a product, a FAQ, a local business. Markup that describes something the reader cannot see is the one kind that causes trouble.

6. Internal links into the page

Count the internal links pointing at this URL and read their anchor text. A page with one link from a sitemap and none from a relevant article is being told, by your own site, that it is unimportant. Three to five links from related pages, with anchor text that names the topic, is a cheap and reliable improvement. Our piece on internal linking structure covers how to find the gaps at scale.

7. Page experience

Run PageSpeed Insights on the URL and read the field data section first, because that is what real visitors experienced. Fix the largest contentful paint and layout shift issues on the template rather than the page, since one template fix lifts every page built on it.

The scorecard

CheckPass testAct when
Intent matchSame format and depth as the top fiveAny mismatch. Rewrite or split before anything else
Title and H1Query near the front, unique across the site, under roughly 60 characters of visible textGeneric, duplicated or missing the topic entirely
CoverageMain question plus the follow ups answered on the pageTwo or more follow ups missing
IndexingIndexed, canonical agrees, content present in the fetched HTMLAny disagreement. This blocks everything
Structured dataValid, and it describes the visible pageErrors, or markup for content that is not there
Internal linksThree or more relevant internal links with descriptive anchorsFewer than two
Page experienceField data in the good band for the templatePoor field data on a template used by many pages

Finish with one line per check and a single verdict: edit, rewrite, split or leave alone. A verdict is what gets acted on. A score out of 100 is a weighting a vendor chose, so read the findings underneath it instead.

What to leave out of the analysis

  • Keyword density targets. Write the way a person asking the question would speak, use the words that naturally follow, and move on.
  • Meta keywords. The tag is inert in Google. Delete it and spend the minute on the description.
  • Exact match headings for every variation. One clear heading per real sub question beats eight headings stuffed with phrasings.
  • Word count as a goal. Match the depth the top results actually show, which is sometimes shorter than yours.

What people ask about on page SEO analysis

How long does an on page analysis take?

Twenty five minutes for one page once you have the sequence, and most of that is the intent and coverage work. The tool checks take five minutes between them.

What is the difference between an on page analysis and an SEO audit?

Scope. An on page analysis judges one page against one query. An audit judges a whole site against crawlability, indexing, structure and speed. Both are worth doing, and the audit comes first when a site has never had one.

Do I need a paid tool for this?

No. Search Console, PageSpeed Insights, the Rich Results Test and view source cover all seven checks on a single page. Paid tools earn their place when you need the same analysis across hundreds of URLs at once.

Which check moves positions fastest?

Intent and coverage, by a wide margin, followed by internal links. Tag work matters and it is rarely the thing holding a page back.

How often should I re analyse a page?

Re run the analysis on your top twenty pages once a quarter, and immediately after any change to the template or the results page for your query. Working from current data is the whole point.

What do I do when the page passes everything and still does not rank?

Compare authority and freshness against the incumbents: how many sites link to their page, how recently it was updated, and how completely it answers the follow ups. That comparison tells you whether the gap is content depth or off page evidence, and sends you to the right fix.

Scoring the page against the answer, not only the results

Your analysis scores a page against the query it should win. Ask that query in an assistant and a second score appears, because the reply either uses your page, uses a rival's, or ignores both and writes something general.

That second result tells you which passages are being treated as the answer. Run it for the query you just analysed in ChatGPT, Gemini and Perplexity, then open the pages each one cites. AI Knows Us reports the cited pages for the questions you give it.

See what AI says about you.

The first scan is free, it needs no account, and it asks the questions your buyers ask.

Free. No card. We ask 5 real buyer questions on 2 AI apps.

Already sure you want this? Join the waitlist. We open on 31 October 2026. It costs an email address. No card, and nothing is charged. Or talk to a person if you have a question first. Or read how we fix what it finds.