What your assessment measures
Method version: nansen-visibility-20261013-v2. This is a checklist of observable page signals. It is not a Google ranking model or a test of live AI recommendations.
260 checks, with evidence
The starting page receives 142 SEO, 52 AEO and 66 GEO checks. An optional focus phrase adds five checks. The checklist covers metadata, HTTP response, headings, content structure, links, images, structured data, crawler permissions, accessibility signals, business context, attribution, freshness and answer formatting. Every finding includes its weight, evidence and suggested next step. Context-dependent signals can be informational rather than failures.
Scoring
Each applicable check carries a weight of 1, 2 or 3. A pass earns its weight; an issue earns zero. Informational or inapplicable checks have zero weight and are excluded. Category score = 100 × earned weight ÷ applicable weight, rounded. Overall score is the rounded average of SEO, AEO and GEO. Exports contain individual weights. Platform readiness cards use relevant subsets of these same checks; they are heuristics, not measured citation probabilities.
The component layout and check coverage were informed by the public SEOScore tool. Nansen implements its own checks and weights. SEOScore’s private scoring algorithm is not available, so scores need not match theirs.
Public page crawl
Crawl mode visits up to 20 same-origin public HTML pages, following eligible links found on each page. It respects applicable robots.txt rules, skips private portal/account/payment paths, and stops within a 35-second crawl budget. Size, network and access limits may yield fewer pages. The report lists checked and skipped pages, repeated issues and duplicate metadata. The displayed overall score is the starting-page score. Up to five starting-page internal links are sampled for HTTP status. This is not an exhaustive crawl.
Public web search and AI assessment
Parallel anonymous MCP search is used first. A confirmed free-limit response can switch to Tavily’s keyless free search. Results are cached for 24 hours, and provider cooldowns are respected. No paid search credentials or automatic upgrade are used. Results include source links and provider attribution. No returned sources does not establish absence from Google. Search order is not a Google position.
DeepSeek receives measured public website facts and available public search excerpts to explain the findings and prioritize fixes. Submitted lead names, emails, private workspace data and provider credentials are excluded. Its write-up does not change deterministic scores. AI interpretation may contain errors; review its recommendations against the evidence.
Performance data
When available, a separate Google PageSpeed Insights request returns actual mobile Lighthouse lab metrics and any available field summary. Lab results depend on run conditions. Total Blocking Time is not Interaction to Next Paint. Failed or quota-limited requests appear as unavailable; no metric is invented.
Limits
The scanner does not execute page JavaScript, load every asset, verify all schema properties, assess source credibility, run a full accessibility audit, verify index inclusion, measure Google positions, or query live ChatGPT or Claude recommendations. Language and readability checks use English-oriented heuristics. Length thresholds are editorial checks, not Google rules. Image sizes supplied through CSS and dynamic content may be missed.
Only public HTTP/HTTPS URLs are accepted. Private/reserved networks, credentials and unusual ports are rejected. Queries and fragments are removed. Better readiness does not guarantee rankings, traffic, sales or AI mentions.