SEO page analysis: the on-page checklist to rank and get cited

An SEO page analysis helps you make fast decisions on a specific page, without diluting the diagnosis into a site-wide average. It works as an on-page checklist to rank and get cited: it combines what Google indexes, what it understands of the page and what the SERP rewards, then asks a newer question: can a generative engine reuse your content without distorting it? This guide extends the overview of SEO tools with an on-page checklist, a way to read the data and a decision grid.

Why analyse a page rather than the whole site

One page, one intent, one SERP

Each page answers a dominant intent: learn, compare, choose, buy, solve a problem. That intent shows in the SERP, which reveals the formats Google favours for the query. Working at page level shortens the hypothesis, change, observation loop. To qualify the intent before opening a tool, rely on the method for analysing search intent.

It is also at this level that you make your answers extractable: a generative engine reuses a passage, a definition or a table, never a site-wide average.

The signals that trigger an analysis

Run a page-level analysis when a performance signal shifts (impressions, CTR, position) or when an internal change may cause a regression. Four situations justify getting started:

• CTR drops whilst impressions rise.
• The page is stuck in positions 8–20 for a revenue-driving query, so very close to where clicks concentrate.
• Two pages compete for the same intent (cannibalisation).
• A redesign, a migration or a new template affects dozens of pages.

Framing the analysis: the right page and a baseline

Selecting a high-potential page

Choose the page like an investment: value, upside, risk. The table below sums up the three criteria:

Set the baseline before any change

Without a baseline, you confuse improvement with statistical noise. Record the state of the page before touching it, over a period that accounts for seasonality:

• Impressions, clicks, CTR and average position in Google Search Console, over 28 to 90 days.
• Engagement and conversions in Google Analytics 4.
• Segmentation by device, mobile first.
• Main queries and fastest-growing queries.
• Current state of the page: rendered HTML, title, headings, main blocks.

The on-page checklist: what to check and how to fix it

The table below works as a control grid: check each row on the page, then compare it with the top 10 for the target query.

Title and meta description: the click promise

The title tag remains the number one lever on CTR. The meta description is not a direct ranking factor, but it shapes click quality: if it promises more than the page delivers, the visitor heads back to the results. Three rules cover the essentials:

• One intent, one title: no catch-all titles.
• A proof element when possible: method, checklist, comparison, update date.
• Natural wording, no keyword stuffing; test a question-style title when the SERP rewards answers.

Heading structure and answer up front

A clear heading hierarchy helps Google understand your outline and helps AI extract coherent blocks. Each H2 answers a stable question, each H3 closes a sub-question with an actionable takeaway. The most common flaws show up when you read the headings alone:

• Different H2s saying the same thing, which dilutes the page.
• Vague headings ("About", "Our solutions") that express no intent.
• No definition early on, even though the SERP favours educational content.
• Level jumps (H2 then H4) without logic.

Depth and vocabulary: cover the intent, not a ratio

A page rarely improves because it is longer; it improves because it is more useful, more precise and better structured. Keyword-density ratios can feel reassuring, but they lead to poor decisions: visible repetition, artificial headings, degraded style. Prefer intent-led coverage:

• Rule of thumb: if repetition is obvious to a human reader, you have already gone too far.
• Vocabulary: list 10–20 related terms (tools, standards, metrics, steps) and assign them by section.
• Evidence: add a reliable source whenever you state a number or a rule.
• Clarity: explain definitions and acronyms at first mention.

Speed, INP and mobile

Use two lenses for performance: lab tests to find what is blocking, field data to know which users are affected. PageSpeed Insights shows both, with the Core Web Vitals LCP, INP and CLS (at the time of writing). The causes of slowness repeat: oversized images, non-essential JavaScript, too many fonts, an overloaded top of page.

Speed things up without removing what reassures and converts: a static image instead of a hero video, a widget loaded after interaction, a static block instead of a carousel. When the problem goes beyond the page (rendering, indexing, canonicals across the site), the guide to technical SEO details the method.

Making the page extractable for AI

Generative engines (ChatGPT, Perplexity, Gemini, AI Overviews and AI Mode) compose an answer and cite sources. A brand is cited when it is named or linked in that answer. To be reused, a page must offer passages a model can lift without ambiguity, and structure helps: pages with an H1-H2-H3 hierarchy are 2.8× more likely to be cited (State of AI Search, 2025).

Lists play the same role: 80% of pages cited by AI use them (State of AI Search, 2025). More benchmarks are gathered in the GEO statistics. To make a page quotable, check these five points:

• A direct answer at the top of each section, in one or two stand-alone sentences.
• Definitions phrased as such ("X is…", "You measure Y with…").
• Steps in lists and criteria in tables.
• An FAQ where each answer makes sense out of context.
• Unambiguously named entities (products, standards, metrics) and sourced numbers.

The simplest test is to read a paragraph in isolation: if it cannot be understood out of context, it is unlikely to be reused well. Also write down the page's promise: if a model summarises it in fifteen words, what should it say without distorting the meaning? The full framework is described in the guide to Generative Engine Optimization.

Comparing the page with the SERP

A page is judged against those already holding the spot. Open the SERP for the target query and note what the top 10 has in common that your page lacks:

• The dominant format: guide, comparison, definition, transactional page, video.
• Sub-topics covered by most results and missing from your page.
• The SERP features present: AI Overview, featured snippet, People Also Ask, local pack.
• The promise of the titles and meta descriptions of the top five results: angle, proof, freshness.

Fill the gaps without copying the angle, then stand out through precision and evidence. The detailed method is in the guide to SERP analysis.

Reading the page's data

In Google Search Console, the Performance report gives queries, clicks, impressions, CTR and average position per page; URL Inspection checks its indexing. At the time of writing, these figures include AI Overviews and AI Mode without separating them. Google Analytics 4 adds engagement, conversions and session sources, including chatgpt.com and perplexity.ai. The table below links each signal to an action:

To make the most of each report, the Google Search Console guide covers them in detail.

Deciding: optimise, merge, rewrite or delete

A technically clean page can still deliver nothing if it targets no clear intent. Decide based on its role (acquisition, conversion, support, reassurance, internal linking) using this grid:

Prioritise: impact, effort, risk

A tool can surface dozens of alerts for a single page; not all deserve a ticket. Run each fix through three questions:

• Impact: can the fix move indexing, CTR, ranking or conversion?
• Effort: how much time, which dependencies, does it need a release?
• Risk: how likely is an SEO or UX regression, especially on a page already in the top 3 or heavily internally linked?

Quick wins cover the title, the meta description, the headings, an added definition or table, and internal links. Then iterate in batches over two to six weeks: snippet and structure first, then content and evidence, finally internal linking and performance.

Measuring before / after

Compare the page with itself, before and after. Google updates its algorithm continuously: dating each change protects your conclusions, which remain correlations, never causation. Observation windows vary by signal:

• CTR: 7–14 days when impressions are sufficient.
• Position: 2–6 weeks depending on crawl frequency and competition.
• Success: one binary criterion (reaching the top 10) and one business criterion (lead, conversion).

From the page to the template

An offer page should not be assessed like an article: build a grid by page type (article, landing page, category, offer page), with both GEO and SEO criteria, then apply it consistently.

If a flaw comes from the template, case-by-case fixing slows you down. Group pages by template and spot what repeats (generated titles, headings, images, scripts, structured data), then test the fix on a pilot page before rolling it out.

Depending on your situation

What comes next depends on your starting point:

• You need to assess the whole site before choosing a page: start with website analysis.
• You already track rankings and want to spot pages that are slipping: an SEO rank tracker brings them to the surface.
• The same flaws repeat across dozens of pages: a full SEO audit is needed.
• Your page is healthy but absent from AI answers: go back to the extractability grid and compare the page with the cited sources.

To run this diagnosis across the whole site, the Incremys 360 GEO & SEO audit module maps keywords and pages, and shows for each keyword whether the brand is named, the domain is cited or used as a source, or it is absent from AI answers.

FAQ on SEO page analysis

How do you analyse a page for SEO, step by step?

First choose a high-potential page, then set its intent by reviewing the SERP. Record a baseline (impressions, clicks, CTR, position, conversions), run the on-page checklist, check extractability for AI and compare the page with the top 10. Then prioritise by impact, effort and risk, fix in batches and measure before / after on the same page.

Which elements should you check first on a page?

Check first what affects the click and understanding: title, meta description, H1 and heading structure, an answer to the intent from the introduction. Next come content depth, internal links, structured data and the canonical, then mobile speed.

What keyword density should you aim for?

There is no universal density, and chasing a percentage leads to padding. Instead, cover the intent using natural variants, entities and co-occurrences, then structure your answers so they are quotable. If you need to choose, choose readability: copy that sounds optimised performs worse.

How do you diagnose a page with lots of impressions but few clicks?

First compare your title and meta description with the top five results: promise, angle, proof, freshness. Then check whether the SERP has changed, for instance with an AI Overview appearing or another dominant format. Align the introduction with the title's promise, test a rewrite, then measure CTR over a stable period.

How do you know whether a page is reusable by AI?

Read each paragraph in isolation: if it cannot be understood out of context, it will rarely be reused. Then check for single-sentence definitions, steps in lists, criteria in tables and sourced numbers. Finally, check query by query whether your brand is named or your domain cited in the answers of ChatGPT, Perplexity or AI Overviews.

How often should you analyse a site's key pages?

For business-critical pages, a monthly or quarterly cadence works, complemented by ongoing monitoring of signals: CTR, indexing, mobile performance, rankings. Add a targeted analysis after every redesign, migration or template rollout, as these are the moments most exposed to regressions.

How do you avoid cannibalisation between two similar pages?

Define a single intent per page and reflect it in the title and H1. When two pages compete for the same query, merge them and redirect one to the other. Designate the reference page through internal linking, and monitor ranking swings between the two URLs on the same queries in Search Console.