Website quality benchmark

A website benchmark that shows what changed.

Record a baseline, examine five areas and retest after implementation. See what is worth comparing and what a score alone cannot tell you.

Check a site with Insight Explore the method

Updated: 22 Aug 2026Maciej Zmitrukiewicz

The comparison rule

A score never tells the whole story.

If the pages, device or data source changed, a difference in the report may not come from your fixes. A useful benchmark also records measurement conditions and what could not be checked.

Full methodology

Comparable scope
Visible data sources
Stated limitations

What to examine

Five areas. Five different questions.

Do not compress them into one number. Open the area you want to understand and inspect its signals and limits.

PerformanceHow quickly does the page become usable? SEOCan search engines discover and read important pages? AccessibilityWhere does an automated test flag user barriers? SecurityWhich configuration signals are visible externally? CO₂eHow heavy is the page, and what does the emissions model assume?

From measurement to action

Use the same method before and after.

A result becomes useful when you can trace an issue, give it a priority and check whether the fix worked.

Explore the report structure

Define scopeDomain, site type, URL list, device, User-Agent and scan depth.
Record baselineDate, methodology version, area results, N/D, P1–P4 and sources.
Hand over prioritiesStart with evidenced issues and owners, then turn the rest into an implementation backlog.
RetestRepeat the same scope after changes and show what actually moved.

This illustrates the benchmark workflow, not a measured result for a specific site.

Keep going

Keep the method close to the result.

Read the methodology to inspect the assumptions. A sample report shows how a result becomes a set of actions.

Full methodology Report structure

Making comparisons

What is a website quality benchmark?

A repeatable measurement card: it records a site’s state for a chosen scope and date. It combines performance, SEO, accessibility, security and ecology signals so you can compare fixes against your own baseline — without relying on one magic score.

Scope: Public page, selected URL scope, device context and measurement date.
Limit: It is not a market ranking and does not predict positions or conversions.

Source: CometWeb methodology Updated: 2026-08-22

What is the difference between lab and field data?

Lab data comes from a controlled test and is useful for diagnosis. Field data describes real-user experiences and varies with devices, networks and templates. A strong benchmark shows both when available, or clearly marks that field data is unavailable.

Scope: Lighthouse or another synthetic test plus CrUX/RUM when there is enough field signal.
Limit: One lab test cannot replace real-user data.

Source: web.dev: lab and field workflows Updated: 2026-08-22

Why report the 75th percentile?

The 75th percentile shows the value below which most observations fall, reducing the influence of isolated outliers. For Core Web Vitals it is the standard way to describe a group of user experiences, not a promise that every session will behave the same way.

Scope: Field-data aggregation for a specific metric and page group.
Limit: With too few observations, a percentile can be unavailable or unstable.

Source: web.dev: measuring Web Vitals Updated: 2026-08-22

Do good Core Web Vitals guarantee higher rankings?

No. Core Web Vitals is one page-experience signal. It does not replace relevance, content quality, crawlability or other search systems. A benchmark should frame it as an improvement area, not as a promise of more visibility or traffic.

Scope: LCP, INP, CLS and page-experience context.
Limit: A CWV result alone cannot predict a ranking or traffic change.

Source: Google Search Central: Core Web Vitals Updated: 2026-08-22

How should crawl scope be documented for a comparable benchmark?

Record the domain, seed list, URL limit, discovery rules, User-Agent, date and any blocks. Compare like with like or mark scope changes. Otherwise, fewer findings may only mean a smaller crawl, not a better site.

Scope: Crawl configuration, visited URLs and the status of each request.
Limit: The result says nothing about URLs the crawler could not fetch or discover.

Source: CometWeb methodology: crawl Updated: 2026-08-22

How can an agency compare client sites without false precision?

Use a shared scope and methodology version, record the baseline, separate N/D from zero and show priorities beside scores. Show the client a trend and evidence from their site, not a table pretending different industries, templates and goals are identical.

Scope: Repeatable client card: area scores, P1–P4, sources, date and retest.
Limit: The result is not a measure of agency quality or the client’s business value.

Source: CometWeb agency materials Updated: 2026-08-22

What can a website quality benchmark not prove?

It cannot prove revenue growth, rankings, AI citations, full WCAG conformance, absence of vulnerabilities or real-world emissions by itself. It can organise observable signals, state limitations and make a repeat measurement easier after implementation.

Scope: Publicly observable diagnostic signals recorded at a specific point in time.
Limit: Legal, business, security and UX decisions need broader evidence and expert review.

Source: CometWeb methodology: limitations Updated: 2026-08-22

Search visibility

What should a benchmark check for indexability?

It should check whether a crawler can fetch the URL, whether robots.txt and meta robots block important content, whether the response status is appropriate and whether links lead to the canonical address. This is technical diagnosis, not proof that a search engine has indexed the page.

Scope: HTTP response, robots, meta robots, links and canonical signals for the chosen URL.
Limit: The search engine’s inspection tool is the authority for current indexing status.

Source: Google Search Central: crawling and indexing Updated: 2026-08-22

How should a benchmark assess canonical without confusing it with a redirect?

A canonical is a hint about which similar URL should represent a group. A redirect removes the alternate URL from the normal flow, while a canonical leaves it accessible. Check canonical consistency with content, internal links, sitemap and server responses.

Scope: Rel canonical, HTTP status, internal links and sitemap membership.
Limit: Google may select a different canonical when signals conflict.

Source: Google: consolidate duplicate URLs Updated: 2026-08-22

Does a sitemap guarantee that every URL will be indexed?

No. A sitemap helps discovery and provides extra context, but it does not replace a valid HTTP response, internal links, canonical signals or useful content. Compare the sitemap with the crawl and mark orphaned or excluded URLs in the benchmark.

Scope: XML availability, valid URLs, status codes and declared scope.
Limit: A URL in a sitemap is not proof that it is indexed.

Source: Google: sitemaps overview Updated: 2026-08-22

What does structured data add to an AEO/GEO benchmark?

Structured data can describe content types, authors, dates and relationships between entities. It is not a shortcut to AI citations or rich results. Check validity, agreement with visible content and whether the schema type actually fits the page.

Scope: JSON-LD, Organization/Person relationships, Article or FAQPage and visible-content consistency.
Limit: Schema does not guarantee a rich result, ranking or an AI mention.

Source: Google: structured data introduction Updated: 2026-08-22

Other areas

What accessibility scope can be automated honestly?

Automation catches a useful subset of repeatable problems such as contrast, missing control names and basic DOM relationships. It cannot judge all usability, content, keyboard order or component context. Report automated output as a signal for review, not as WCAG certification.

Scope: Automated checks mapped to selected WCAG 2.2 criteria.
Limit: Full conformance work requires manual testing and an appropriate evaluation scope.

Source: W3C: WCAG 2.2 Updated: 2026-08-22

Does an automated test result mean WCAG conformance?

No. An automated test can confirm one condition or flag a likely problem, while conformance concerns the whole page, process and criteria in scope. Separate pass, fail, needs review and out-of-scope items in the benchmark.

Scope: Test rules, covered elements and manual reproduction steps.
Limit: Finding no automated issue does not prove that all barriers are absent.

Source: W3C: ACT Rules Updated: 2026-08-22

How can security be benchmarked without calling it a pentest?

A public benchmark can check observable transport and header signals such as TLS, HSTS, CSP or cookie flags. This external surface review helps prioritise fixes, but it does not replace application testing, server configuration review or a penetration test.

Scope: Public HTTP/TLS and configuration signals available without authentication.
Limit: It excludes backend code, business logic, secrets and a complete threat model.

Source: MDN: Content Security Policy Updated: 2026-08-22

What does a page CO₂e estimate mean?

It is a modelled result based on transfer, page views and assumptions about energy and infrastructure emissions. It helps compare your own changes and find heavy assets. It is not a direct emissions measurement, ESG audit or environmental certificate.

Scope: Page transfer, calculation model, assumptions and measurement date.
Limit: The result depends on the model, hosting, cache, device and traffic assumption.

Source: CometWeb methodology: Ecology Updated: 2026-08-22

Start with a measurement of your own site.

Insight helps you move from an audit to priorities and a retest after changes.

Start with Insight Let’s talk