AIA Score Explained

What the score measures

The current calculation combines factual explicitness, a structure heuristic, and the presence of headings and meta descriptions. Human readability is reported separately and is not a fourth weighted input.

This is an AIA Matrix diagnostic metric. It does not measure whether an AI service recommends your business, and it is not a Google ranking score, a WCAG accessibility audit, or independent verification of your claims.

The current formula

The three weighted inputs are:

E = clamp((fact density - ambiguity - unanswered questions) / 2, 0, 100)
AIA Score = round(0.40 × E + 0.40 × S + 0.20 × A)

Each AI-estimated explicitness input is between 0 and 100. “Clamp” restricts the result to the stated range; “round” produces a whole-number score.

Understanding the inputs

Factual explicitness

• Fact density: the model estimates how much content contains concrete details such as names, dates, prices, locations, and quantities. Higher is better for this input.
• Ambiguity: the model estimates how much language is vague or subjective. A higher value reduces explicitness.
• Unanswered questions: the model estimates whether practical questions about the service, cost, or location remain unanswered. A higher value reduces explicitness.

A concrete claim is not necessarily true. Review extracted facts against your actual business information and supporting evidence.

Structure

The helper starts at 100 when there is exactly one H1 and 75 otherwise, with deductions for an outline starting below H1 or skipping heading levels. However, the current scan pipeline feeds it one generated H1 per scanned page, rather than each page’s original HTML heading hierarchy. A single-page scan therefore receives 100 for this input; a scan with multiple pages receives 75. This limits its usefulness as a measure of real page structure and means scan coverage can affect the score.

Accessibility

Each scanned page receives 50 points for a nonempty extracted heading and 50 for a nonempty meta description. The input is the rounded average across pages. It does not test keyboard access, screen readers, crawler permissions, or every AI system’s ability to retrieve the site.

Human readability

The report’s readability value is an AI-generated Flesch-Kincaid-like estimate. It is separate from the weighted AIA Score and should not be treated as an exact sentence-and-syllable calculation.

What a scan covers

Free scans analyze a single page. Professional scans retrieve up to 50 prioritized pages. The current AI analysis receives up to 1,200 characters of extracted text per page, with a combined input limit of 60,000 characters, alongside the selected homepage’s title and meta description.

The score therefore reflects a sample of retrieved content, not every sentence or URL on your website. Blocked pages, extraction problems, changes in selected pages, and AI estimation can affect the result.

How to use the report

1. Check coverage first. Review the scanned URLs and page count before drawing conclusions about the whole site.
2. Review factual gaps. Add accurate service descriptions, locations, hours, pricing or quote information, and answers to common customer questions where appropriate.
3. Remove ambiguity and contradictions. Replace unsupported superlatives with useful details, and reconcile inconsistent names, dates, and contact information.
4. Improve page organization. Use descriptive headings and useful meta descriptions. Good HTML structure matters even where the current scoring heuristic does not capture it fully.
5. Validate changes and rescan. Confirm the updated content is publicly accessible, then compare scans of the same URL with similar coverage.

The recommendations and gap estimates help you decide what to review; they are not a fixed checklist of guaranteed point increases. Publishing a ZIP export or an llms.txt file does not automatically add points in the current calculation.

Interpreting changes over time

Use score changes as a prompt to inspect what changed. Compare the same domain, page scope, and scan type, and review the underlying recommendations alongside the total. Single-page and multi-page results are not directly equivalent because their structure inputs differ.

AI-generated estimates can vary between runs. A small difference alone is not proof that your content improved or declined. There are no validated grade bands or guaranteed visibility outcomes defined by this implementation.