How Varazm measures AEO
Every number we report is either modelled from your page or measured from real engine answers. This page states which is which — and what each one does not prove.
AEO Readiness
Computed from the page itself: 25 weighted factors. Deterministic and repeatable. It tells you whether the page is quotable — not that it has been quoted.
Live AI Visibility
Extracted from real stored answers: brand mentions, citations, share of voice and sentiment. It depends on the prompts you track and varies between runs.
1. What we measure
The analyzer scores 25 factors in five groups. Each factor carries its own score, evidence and fix guidance.
Crawlability & AI crawler access
robots.txt rules per bot, sitemap presence, HTTP status, redirects and whether retrieval bots such as OAI-SearchBot or PerplexityBot are permitted.
Structure & semantic HTML
Single H1, correct heading order, landmarks, list and table semantics, and whether the main answer is present in server-rendered HTML.
Metadata & structured data
Title and description quality, canonical correctness, Open Graph completeness, and valid, non-contradictory JSON-LD with consistent entity identifiers.
Answerability & attribution
Whether questions are answered directly and early, factual density, dates, named sources and outbound references that let an engine verify a claim.
Delivery & accessibility
Rendering without client-side gating, viewport and mobile usability, image alternatives, internal linking and page weight.
2. Engine modelling
Each engine profile weights the factors differently — an engine that fetches pages at answer time is more sensitive to crawler access and server-rendered HTML. This weighting is an informed estimate based on published engine behaviour, not access to any internal algorithm.
That is why the interface always shows the number of engine readiness models and the number of live monitored answer engines as two separate figures.
3. Live monitoring
Every tracked prompt is sent to the connected engines and the full answer is stored with its engine, language, locale and timestamp. Brand mentions (Unicode-aware, word-boundary matching across all registered aliases), citations (normalised to domain and path), competitors and sentiment are then extracted from that stored text.
4. Repeated sampling & variability
One run is not a result. Runs are stored individually so multiple executions of the same prompt can be compared and the real spread becomes visible — your brand may appear in one run and vanish in the next without anything changing on your site.
5. Confidence
Confidence is computed only from real data: the number of stored samples and the observed variance between them. When there are too few samples the figure is labelled low confidence — no synthetic confidence interval is ever generated.
AI Visibility 61 ± 7 Confidence: High Based on 10 repeated samples
6. Definitions
- Mention — the brand name appears in the answer text, with no source link.
- Citation — the answer links to a specific URL as a source.
- Sentiment — the tone of the sentence containing the mention; always stored with an evidence snippet.
- Share of Voice — your share of the answers that mention any registered brand, yours or a competitor's.
7. Sentiment & claim accuracy
Sentiment is about tone, not accuracy. Claims an engine makes about your brand are kept separately with their evidence so they can be marked confirmed, needs review or incorrect. No claim is ever stored as fact without evidence.
8. Limitations
- A readiness score is not a rank and does not guarantee a citation.
- Visibility is only valid for the prompts you track; a different prompt set gives a different result.
- Engines change without notice, and answers can be personalised or region-specific.
- Some sites block crawlers; the audit then uses fallback retrieval and the report says so explicitly.
- No tool, including this one, has access to the internal algorithms of answer engines.
Frequently asked questions
- What is Answer Engine Optimization?
- Answer Engine Optimization (AEO) is the practice of preparing a website so AI answer engines can read, understand, quote and cite it. It combines technical access, semantic clarity and directly answerable content.
- How is AEO different from SEO?
- SEO optimises for a position in a list of results; AEO optimises for inclusion inside a generated answer. Most technical SEO work still helps, but AEO adds entity clarity, factual attribution and AI crawler permissions.
- What is AI Visibility?
- AI Visibility is how often your brand actually appears in answers produced by AI engines for a tracked set of prompts. It is measured from stored responses, not estimated from your page.
- How does Varazm calculate readiness?
- Readiness is computed from the page itself: 25 weighted factors grouped into technical, structural, semantic and content categories. The same page always produces the same score until the page changes.
- Why do AI responses change between runs?
- Generative models sample from a probability distribution, so the same prompt can produce different brands and sources each time. This is why a single run is never treated as a result — repeated runs are.
- Does Varazm use estimated or live data?
- Both, clearly separated. Readiness scores are estimated from your pages; visibility, citations, share of voice and sentiment come from live answers stored at run time. The interface never presents one as the other.