1/11/2026
A website GEO and SEO analysis tells you what stops your pages being crawled, understood and cited, what they lack to rank, and where to start. The method begins with a map of the site by template, works through seven parts from crawl to AI presence, then serves to explain a traffic variation and to test a fix. To choose the software that supports this work, see our overview of SEO tools.
What a website analysis must tell you
What is blocking, what is missing, where to start
A useful analysis produces decisions: what to fix, what to consolidate, what to publish, and how to prove the impact. It looks for what genuinely costs crawl, clicks or leads by combining a crawl, Search Console and your analytics tool: a crawler on its own surfaces thousands of issues with no measurable effect.
In practical terms, the diagnosis connects visibility to value for three families of pages:
- Highly visible but rarely clicked: the problem lies in the snippet (title tag, description, a competing SERP feature).
- Clicked but rarely converting: the promise, the offer, speed or the user journey is at fault.
- Converting but under-exposed: internal linking, content or authority is lacking.
GEO targets the reuse of your content in answers from ChatGPT, Perplexity, Gemini or AI Overviews; SEO targets rankings and clicks. Both start from the same pages; GEO also requires extractable blocks (definitions, lists, tables) and trust signals (sources, dates, author).
Analysis, test or audit: three different scopes
The three words get mixed up, in tools and in quotes alike:
- Website analysis: a decision-led diagnosis, run with your Google data, that keeps visibility under control.
- Test: checking one hypothesis on a limited scope (a page, a template, a folder), with a before and an after.
- Full audit: an exhaustive review when you suspect a systemic problem (templates, CMS, migration, indexing rules).
An analysis is for deciding quickly; an audit structures a wider roadmap. If the analysis reveals a problem affecting entire sections of the site, move to a full SEO audit rather than fixing page by page.
Scope before you measure: perimeter, site map and baseline
Without framing, the figures cover averages that designate nobody, and fixes land on whichever URL happens along. Set the perimeter (domain, subdomains, folders, languages and countries, business pages) and validate the measurement chain: key events, attribution, UTMs, filters. Framing then delivers a map of the site and a measured baseline.
Mapping pages by template and by role
An actionable map groups pages by template (category, product page, landing page, article) and by role. Four families describe almost any site:
- Business pages (offers, strategic categories, contact): they offer an action.
- Hubs (pillar pages): they direct visitors and distribute authority.
- Supporting content (guides, FAQs, articles): it covers a sub-intent or answers an objection.
- Utility pages (legal notices, internal search, filters, pagination): they are kept under control to avoid noise.
Attach four values to each group: depth in clicks from the home page, internal inbound links, traffic on entry pages and visibility in search results. A map becomes actionable when it lets you decide "where" to fix, at template level, rather than "which URL" at random. A fix applied to a template propagates to all its pages; applied to an isolated URL, it is lost.
The map also makes the hub → child pages → proof (case studies, FAQs, comparisons) → conversion chain visible. Without it, you optimise an isolated page when the hub distributes nothing, or add supporting content where proof is what is missing. A broken link in the chain shows up immediately on the map, never in a page-by-page report.
Depth, internal links and orphan pages
Depth determines discovery: the further a high-stakes page sits from the home page, the more slowly it is discovered and the more expensive it becomes to crawl. Three questions read this part of the map:
- Which pages receive the most internal links, and is that consistent with their business value?
- Which strategic pages are too deep, or reached only through pagination, filters or internal search?
- Does internal linking make clear which page is the reference when several pages cover a close subject?
The same cross-check brings out orphan pages, which no internal path reaches. Compare the crawl's URLs with the sitemap, CMS exports and GA4 landing pages: those receiving traffic with no internal link go to the top of the list. Attaching them to a hub and shortening paths is a matter of internal linking.
Setting the baseline at constant scope
Without a measured starting point, you test nothing and explain nothing: you observe. A useful baseline compares at constant scope (same weekdays, same season, same campaigns) and is cut by segment:
- period A against period B, rather than one month against the previous one, which mixes seasonality with real effect;
- brand against non-brand, because brand often hides structural falls;
- mobile against desktop, because 60% of global web traffic comes from mobile (Webnyxt, 2026) and a degradation on mobile can go unnoticed in the total;
- strategic pages against long tail, so as not to drown your business pages in an average.
Capture it in Search Console (impressions, clicks, CTR, position by page and by query), GA4 (organic sessions, conversions) and a crawl or your logs (status codes, depth, crawl frequency). Document the period and segments, and keep the exports: without that, an "improvement" may be no more than a change of scope.
The seven-part method
The seven parts follow the value chain: if Google cannot crawl a page, nothing else matters. Each one is read by group of templates from the map, never as a site total.
Crawling, indexing and technical health
Start with the gap between the pages you publish and those Google actually indexes. The most expensive mistakes are rarely subtle: blocking directives, inconsistent sitemaps, contradictory canonicals, redirect chains. Search Console spots them page by page; crawling the template tells you whether the error is isolated or systematic.
For performance, look for root causes by template and by device: images, third-party scripts, JavaScript rendering, server. Only 40% of sites pass Google's Core Web Vitals assessment (SiteW, 2026), and a slow template weighs on every page that uses it. Other benchmarks are gathered in our SEO statistics.
Content, internal linking and authority
A technically clean page can earn no impressions if it answers no intent or if another page targets the same query. When two URLs alternate on a keyword, you are dealing with SEO cannibalisation: merge them, change the angle or deindex the weaker one.
On large sites, internal linking is more often the limiting factor than micro-optimisations, and the map already shows where it fails. For authority, stay pragmatic: focus the effort on a few strategic pages, not the whole site.
AI presence and conversion
At the time of writing, Search Console counts AI Overviews and AI Mode, available in France since summer 2026, in the Performance report without separating them from classic results. To measure your AI presence, ask ChatGPT, Perplexity, Gemini and Google your key questions and note whether your brand is named or your site cited.
Finally, connect each organic landing page to a conversion in GA4: a visible page that does not convert raises a question about the offer or the journey, not about SEO. The signs are read by template: weak engagement, entry pages that are heavily clicked but not engaging, journeys with no reassurance.
Analysing your site for free with Google's tools
Three free tools cover most of a first analysis, with one decisive advantage: they show Google's data, not an estimate.
Google Search Console: indexing and real demand
Our Google Search Console guide details each report; four do most of the work:
- Performance: queries, pages, clicks, impressions, CTR and average position, by country and by device.
- Page indexing: indexed pages, excluded pages and the reason for exclusion.
- URL Inspection: status of a specific page, Google-selected canonical, rendering.
- Sitemaps and Links: submitted files, most-linked pages internally and externally.
Filtered by folder or page group, these reports feed the baseline set during framing and the explanation of any variation.
Google Analytics 4 and PageSpeed Insights
GA4 tells you what visitors do once on the site. Isolate organic sessions by landing page, the associated conversions and session sources from generative engines (chatgpt.com, perplexity.ai). Pages that attract without converting, and pages that convert without attracting, are your best candidates.
For a given URL, PageSpeed Insights provides field data (real users) and a lab test, which is useful for debugging and for comparing two versions. Test one URL per template, on mobile first, and validate the effect after release.
When free is enough
Free tools work when the need is bounded: validating a landing page, checking a canonical, spotting a redirect chain. With several domains or countries, history and segmentation fall short. A one-hour loop makes the most of them:
- Diagnose (15 minutes): the page and its technical set-up, on one URL.
- Prioritise (10 minutes): blocking for indexing, improves CTR, improves content, minor.
- Fix (20 minutes): one to three actions at most.
- Measure (15 minutes): indexing and the trend in impressions and CTR in Search Console.
Repeat this cycle on five to ten priority pages before launching a redesign or large-scale production.
Explaining a variation: impressions, CTR, positions, conversions
A fall rarely has a single cause, and is almost never read at overall level. Start from the strategic pages and their queries, compare them with the baseline, rule out seasonality and on-site changes (first of all an unintentional noindex), then work back to a family of cause. The map sets the level of reading: a fall attaches to a template, not to the site total.
From observed signal to decision
Impressions, CTR, positions and conversions read with a simple grid:
When CTR falls at an unchanged position, it is the results page that has changed: a SERP analysis shows which formats and which competitors occupy the space. When the variation comes down to a single URL, take it up with an SEO page analysis, element by element.
The indicators that mislead
Some readings set off more useless projects than any technical anomaly:
- An aggregate score summarises a page's hygiene without predicting rankings: a well-scored page stays invisible if it targets no intent. Use it to spot regressions, weighted by page type.
- Aggregated average position says nothing about the pages that count: look at business pages and queries with something at stake.
- A rise in traffic is a qualified gain only once you have checked the queries carrying it.
None of these reservations forbids acting; each one forbids concluding too quickly. In B2B, the quality of entries counts more than raw volume.
Reading an SEO analyser report without getting it wrong
An SEO analyser examines URLs against a grid of technical and on-page criteria, then produces recommendations. Its findings are clues: the decision comes from cross-checking them with your Google data. Because it ignores your product and technical constraints, sort its recommendations before opening a ticket.
- Follow: error status codes, empty rendering after JavaScript, indexing directives, inconsistent canonicals or hreflang, redirect chains.
- Challenge: "lengthen the text", "add keywords", "add H2s", too generic to apply as they stand.
- Ignore: isolated alerts with no drop in impressions or clicks, or recommendations that would create duplication.
A tool that does not reproduce JavaScript rendering analyses a page Google does not see: note the analysis mode used so you can compare month to month. Then rank what remains by impact, effort and regression risk. Blockers (unintended noindex, inconsistent canonical, 5xx errors) come first; quick wins (titles of pages with high impressions and low CTR, internal links) within the month; structural projects (template duplication, indexable facets) are planned with the technical team; polish with no observable effect comes last, or never.
Testing a fix: the 10-point recipe
A test starts from the baseline set during framing. Note the biases (seasonality, releases, Google updates, SERP changes); if in doubt, narrow the scope and lengthen the observation window. The protocol applies to a page, a template or a folder:
- 1. Target: business pages, declining pages, pages in positions 11 to 20, since the click-through rate on page 2 of the SERPs is only 0.78% (Ahrefs, 2025).
- 2. Formulate a single hypothesis and the change that tests it.
- 3. Capture the baseline on the tested scope.
- 4. Check indexability: robots, noindex, canonical, redirects, duplication.
- 5. Check for errors: 4xx and 5xx status codes, broken links on internally linked URLs.
- 6. Compare the page with the SERP: expected angle, format and level of detail.
- 7. Review the markup: title, description, heading hierarchy, internal links pointing to the page.
- 8. Measure speed: LCP, INP and CLS on mobile, under constant conditions.
- 9. Test extractability: which passages could an AI reuse as they stand?
- 10. Measure the after over a comparable window, by segment, keeping the evidence (exports, SERP captures, crawl).
Allow several weeks for crawling and reindexing. Validate in three stages: technical proof (the error has disappeared in the crawl), Google proof (indexing, impressions, rankings, CTR) and business proof (conversions in GA4). If Google's signals do not move, attribute no gain to the fix.
Governing decisions and tracking the site over time
Without governance, an analysis produces contradictory optimisations (experience against SEO, content against product) and effects that are impossible to attribute. Two arrangements prevent this.
A single backlog and a written trade-off
The minimum viable set-up rests on three elements:
- a single backlog, dated and versioned, gathering the decisions that come out of the analysis;
- measurable acceptance criteria, based on the same sources as the diagnosis, with an observation period fixed in advance;
- a written list of non-priority actions, which avoids scattering and the return of the same subjects.
Every action retained gets a target, a measurable objective, an owner and a success criterion. The trade-off is pronounced in three parts: what you fix now, what you postpone, what you will not do, with a written justification. When what remains depends on the templates or on the organisation, the question is no longer how to read the site but how to change the way it is produced.
A routine with three rhythms
A one-off analysis ages quickly; a routine shortens the time between detection and action:
- Every week: alerts on indexing, errors and abnormal drops in impressions on business pages.
- Every month: decisions, tickets and content briefs.
- Every quarter: recalibrating templates, keyword groups and GEO strategy.
The dashboard has to be decidable rather than complete: visibility, CTR, indexing, engagement and conversions, read by group of templates, to say which part of the site is progressing and which one is absorbing the fall. Year-on-year comparison and rank checks at a fixed cadence are organised as part of SEO tracking.
FAQ: website GEO and SEO analysis
How do you analyse a website for free?
With Google's free tools. Search Console shows indexing, queries, clicks and average position; Google Analytics 4 connects landing pages to conversions; PageSpeed Insights measures LCP, INP and CLS. Together, they cover a first website analysis. Their limits show up across several domains or countries, for lack of long history and segmentation.
Website analysis or SEO audit: what is the difference?
A website analysis produces a decision-led diagnosis: what is blocking, what is missing, where to start. A full audit reviews the whole site exhaustively when a systemic problem is suspected (templates, CMS, migration, indexing rules). Start with the analysis; an audit is justified when it reveals problems affecting entire sections of the site.
What should you check first on a website?
Crawling and indexing. If Google cannot crawl or index your strategic pages, nothing else matters. Then check that pages match the intent of the SERP, then the mobile performance of landing pages, and finally the internal linking and authority that strengthen business pages.
How do you map a website in order to analyse it?
Group the URLs by template and by role: business pages, hubs, supporting content, utility pages. Give each group its depth, internal inbound links, traffic and visibility. The map becomes actionable when it designates the template to fix rather than a URL picked at random, and shows whether the hub, child pages, proof, conversion chain holds.
How do you explain a fall in organic traffic?
First compare with a baseline at constant scope: same weekdays, same season, brand and non-brand, mobile and desktop. Rule out an unintentional noindex and recent releases, then attach the fall to a template. Next read impressions, CTR, positions and conversions together: each combination points to a family of cause, from indexing to the snippet.
How do you include AI presence in a website analysis?
Ask ChatGPT, Perplexity, Gemini and Google your key questions, and note whether your brand is named or your site cited. Then check that your source pages contain extractable blocks: short definitions, lists, tables, dated sources. In GA4, track sessions coming from chatgpt.com and perplexity.ai.
Continue reading
Depending on what the analysis has brought out:
- Recurring technical anomalies: if crawling, rendering or performance cause problems across several templates, the fix becomes a technical project → technical SEO audit.
- Strategic pages rarely visited by Googlebot: if the map shows deep or orphan business pages, server logs tell you what the robot really crawls → log analysis.
- Reporting to a committee: if you have to account each month for the decisions taken and the AI citations before and after, the form matters as much as the substance → SEO report.
- Map and AI presence in one place: if you want the mapping of pages and keywords without assembling several exports, the Incremys 360° GEO & SEO audit module diagnoses the site and shows the brand's AI presence for each keyword: named, domain cited, domain as source or absent.
.png)
.jpeg)

.jpeg)
%2520-%2520blue.jpeg)
.avif)