Clear expectations
What the score really means.
A useful assessment starts with knowing what was measured and what was not.
A baseline from your public website
CSFunnel fetches the submitted page and selected supporting resources, such as robots and sitemap information, to evaluate observable website signals. It is not an exhaustive crawl of every page or a complete browser performance and accessibility audit. Findings depend on the content available to the analyzer.
Six areas of readiness
- AI readiness: business context, clear explanations, and content signals relevant to discovery.
- Structured data and AEO: machine-readable context and answer-oriented structure.
- On-page SEO: titles, descriptions, headings, links, and related signals.
- Technical and crawlability: access, crawler directives, and supporting technical information.
- Content and trust: authorship, evidence, and verifiable business information.
- Performance and accessibility: detectable foundations, not a substitute for a full accessibility review or real-user performance measurements.
From checks to scores
Checks receive a status and contribute to category scores using the analyzer’s weighting rules. Your report shows the findings and priorities behind the result. A score summarizes this assessment; it is not an estimate of a platform’s private ranking system. Blocked resources and unavailable checks limit the evidence.
Evidence scoring v2
New audits use evidence scoring v2. A pass earns full credit, a warning earns 25%, and a failure earns none. High-impact checks have twice their base weight; medium-impact checks have 1.5 times their base weight. Category scores are the weighted earned points divided by possible points. The overall score remains a weighted average of the six categories. A category with no scored evidence receives zero, not a perfect score.
Individual crawler directives are listed for diagnosis and counted once through the combined search-access check. Optional files, copyright notices, marketing keywords, and unverified certification claims do not independently earn points. Missing optional information does not create a penalty. Quantitative claims need accompanying source links, quotations need attribution, and content dates must be specific and valid. This scan detects evidence; it does not independently prove every claim.
Form-label and image-dimension checks require complete coverage for full credit. Structured business data needs a meaningful name and website URL. The AI readiness estimate combines the AI readiness category (50%), structured data (30%), and content/trust (20%). These weights are CSFunnel’s assessment rubric, not a platform’s ranking formula.
There is no arbitrary score ceiling: a page that passes all applicable checks can earn 100. Even 100 describes the checked website evidence, not guaranteed recommendations. Earlier saved reports keep their original scores and are labeled with their earlier model. Run a new audit to use the updated rubric, and compare reports using the same model.
Reference guidance: Google’s AI search guidance, content dates, and structured data.
Unified report assessment v3
Saved report pages and PDFs now present one overall assessment across eight areas: AI search and GEO (20%), structured data (15%), content and trust (15%), search and SEO (15%), crawlability (10%), measured performance (15%), measured accessibility (5%), and browser best practices (5%). These weights are our assessment rubric, not a search platform formula.
The first five areas use the saved website evidence. Search and SEO combines the website score (70%) with browser SEO (30%) when both are available. The three measured areas use the selected device’s browser scores directly. Mobile is the default; selecting desktop also changes the PDF snapshot. Missing measured categories are excluded and available weights are normalized. Coverage and provisional scores are clearly labeled. Real visitor data and live AI answers provide context and do not change this score.
Historical raw audits are preserved. The unified assessment can differ from the original website-only score because it includes measured browser evidence. The report history retains the original website-check score. Compare the same device and assessment version. Pending measurements appear immediately with loading indicators; completion notifications require this report tab to remain open.
Measured website experience
New audits request mobile and desktop browser tests in the background through Google PageSpeed Insights and Lighthouse. Performance, accessibility, browser best practices, and search foundations contribute to the unified report as described above. We show measured load timings, grouped improvement actions, estimated opportunities, and checks that still need human review. A single run can vary; compare the same URL, device, and Lighthouse version.
Experimental AI usability checks describe whether agents can operate a page. They do not measure business recommendations, AI rankings, or guaranteed inclusion. Unsupported categories are shown as unavailable.
Real visitor experience
Where enabled and available, the separate Chrome UX Report API supplies historical Chrome visitor measurements at the 75th percentile. We show the collection dates and identify website-wide data when page-level data is missing. Core Web Vitals require LCP, INP, and CLS; lab blocking time is not visitor INP. Missing data or a provider failure never becomes a zero score.
Visitor data comes from the Chrome UX Report, licensed under CC BY 4.0, and is summarized by CSFunnel. Browser measurements come from Google PageSpeed Insights / Lighthouse. These sources do not endorse CSFunnel.
Availability and measurement limits
We reuse matching measurements for up to 24 hours to limit repeated requests. Failed tests display a reason and may be retried after a cooldown; daily and queue limits also apply. Your baseline audit remains available during provider outages. Measurements are versioned attachments: a fresh audit does not overwrite the previous report. URLs with query strings are not submitted to external measurement services because they can contain private tokens.
Evidence corrections v2.1
New readiness audits account for SVG artwork, modern-browser script applicability, and accessible names on image controls. Unmeasured compression and full document fetch duration are informational, not substitutes for measured compression or time to first byte. Saved scores retain their original model.
Competitor context
An optional comparison runs the same baseline assessment for another public URL. Compare pages with similar purposes and read the findings alongside the scores. A competitor’s score does not reveal their traffic, AI mentions, or commercial success.
Live Gemini evidence
If available, a separate action on your report can request a Gemini answer and record its sources and business mentions. Cached results may be reused and fresh requests have daily limits. The report identifies the available evidence; a website score does not increase just because one answer mentions the business.
What to do next
Review the highest-priority findings in the context of your business. Fix access barriers and unclear information first, then improve useful answers and supporting evidence. Recheck after meaningful changes and assess business outcomes separately.