Methodology
How AI engines decide what to cite
Answer engines do not rank pages the way search engines do. They fetch, parse, and decide whether a page is safe to quote. Sona built a 209-point rubric of 35 checks across six weighted categories to measure exactly that. This page documents the whole thing, openly, so you can audit your own site against it whether or not you use Sona.
3 in 4
websites are partially or fully invisible to AI engines
60%
of Google searches now end without a click
800M+
people use ChatGPT every week and ask it for recommendations
Source: Sona AI Visibility Checker, ai-visibility.sona.com
The scoring model
Every check is pass, partial, or fail, and each carries a fixed point value. Weights are not evenly distributed. They reflect how decisive each category is in practice. Crawlability and content structure together account for 133 of 209 points, because a page that cannot be fetched or parsed is never a candidate for citation regardless of how good the content is.
| Category | Points | Share of total |
|---|---|---|
| Crawlability | 66 | 32% |
| Content Structure | 67 | 32% |
| Content Quality | 40 | 19% |
| Performance | 14 | 7% |
| Accessibility | 13 | 6% |
| Security | 9 | 4% |
| Total | 209 | 100% |
What each category measures
Crawlability
66 ptsIf an AI crawler cannot fetch and parse your page, nothing else matters. This is the largest single failure category and the most fixable.
- Content visible without JavaScript execution
- AI bot permissions in robots.txt
- Live GPTBot fetch probe
- Canonical URL correctness
- llms.txt presence and validity
- Load speed for crawlers
- Internal link integrity
Content Structure
67 ptsStructure is how a model locates the answer inside your page. Unstructured prose forces the engine to guess what your page is about.
- Semantic HTML
- Heading hierarchy (single H1, ordered H2/H3)
- Structured data: FAQ, Article, Organization, Breadcrumb, Contact, HowTo
- JSON-LD validity
- Title and meta description quality
- Open Graph tags
- dateModified present
Content Quality
40 ptsEngines prefer sources that define entities explicitly, make specific claims, and show who wrote them and when.
- Entity definition (is it clear what you are?)
- Citable, specific claims
- Content depth
- Topical authority
- Named authors
- Visible timestamps
- Freshness within 90 days
Performance
14 ptsCrawlers operate under time budgets. Slow pages get partially fetched or skipped entirely.
- Core Web Vitals (LCP, CLS)
- Page load speed
- Resource optimization
Accessibility
13 ptsAccessibility markup doubles as machine-readability: alt text and contrast compliance describe content a parser cannot see.
- Image alt text
- Automated checks (Lighthouse / axe)
- WCAG AA contrast compliance
Security
9 ptsTransport and header hygiene are trust signals, and mixed content can break a crawl outright.
- HTTPS
- Mixed content
- Security headers (HSTS, CSP, X-Content-Type-Options)
Free tool · No account
Is your website invisible to AI?
Sona's free AI Visibility Checker runs 35 technical checks across 6 categories and scores your site out of 209 points across crawlability, schema markup, content structure, performance, security, and accessibility. Results in under 30 seconds, no signup.
66
Crawlability
67
Content Structure
40
Content Quality
14
Performance
13
Accessibility
9
Security
209 points total across 35 checks.
Frequently asked questions
How do AI engines decide which sources to cite?
AI answer engines evaluate a different signal set than classic search rankings. Before a page can be cited it has to be fetchable without JavaScript, parseable as structured data, clearly scoped to a defined entity, and recent enough to be treated as current. Sona groups these into six categories (crawlability, content structure, content quality, performance, accessibility, and security), weighted across 35 checks for a 209-point total.
What is AI visibility?
AI visibility is how often and how accurately your brand appears in answers generated by AI engines such as ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews. It has two halves: whether an engine can technically read and cite your site, and whether your brand actually surfaces in the answers your buyers see. The first is an audit problem; the second is a monitoring problem.
Why does crawlability carry the most weight?
Crawlability is a gate rather than a ranking factor. If GPTBot cannot fetch the page, or the content only renders after JavaScript executes, no amount of good writing or schema markup can compensate, because the engine never sees it. That is why crawlability and content structure together account for 133 of the 209 available points.
Does an llms.txt file actually matter?
It is a small but real signal, and it is checked directly. An llms.txt file gives answer engines a plain-text map of what your site is and which pages matter most, in a format designed to be read as guidance rather than marketing copy. It does not override crawl permissions, and it will not rescue a site that fails the crawlability checks.
How fresh does content need to be?
The freshness check looks for content updated within the last 90 days, alongside a visible timestamp and a machine-readable dateModified field. Engines discount sources they cannot date, and they discount them further when the visible date and the structured-data date disagree.
Is the AI Visibility Checker free?
Yes. The checker at ai-visibility.sona.com runs all 35 checks with no account and no credit card, and returns a scored report in under 30 seconds. Ongoing monitoring of how often your brand appears in AI answers, across engines, prompts, regions, and competitors, is the paid Sona platform.
Auditing your site is step one. Knowing what AI says about you is step two.
Sona tracks how often your brand appears in answers across ChatGPT, Perplexity, Claude, Gemini and 10+ engines, then ties that visibility to pipeline and revenue.
