Alt Text and Accessibility Labels

The Alt Text and Accessibility Labels parameter measures how well a page labels its images and interactive controls so AI crawlers and assistive technology can understand non-text content. friendly4AI scores it from the combined coverage of image alt text and interactive-control labels: a page passes (100) at 90% or higher, scores partial (50) between 60% and 90%, and fails (0) below 60%.

What does this parameter measure?

The score rests on two signals: do your images carry alt text, and do your interactive controls carry accessible labels? friendly4AI scans the page HTML (DOM) and counts every <img>, tracking how many have an alt attribute. It then counts interactive controls — <input>, <button>, <select>, <textarea> — and how many are labelled through aria-label, aria-labelledby, a title attribute, or, for non-input controls, visible text content. That gives two coverage ratios, which it averages into a single accessibility-coverage figure.

Why does it matter for AI-readiness?

Alt text and ARIA labels are how AI systems read non-text content. An image with no alt is opaque to a crawler. A button labelled by nothing but an icon means nothing to a machine. Label these well and answer engines — ChatGPT, Perplexity, Gemini, Google AI Overviews — can tell what your visuals show and what your controls do, so they describe and cite your page more completely. The coverage figure works as a quality signal too: a well-labelled page reads as well-built to assistive technology and AI alike. It pairs with related structure signals such as semantic HTML, heading hierarchy, and page metadata.

How is it scored?

Under the v4.5 methodology, this Content Structure parameter is a gradient based on combined coverage. The processor works out image alt coverage (images with alt / total images) and control label coverage (labelled controls / total controls), then takes the mean of the two. If a page has no images, or no controls, that dimension counts as full coverage (1.0) — you are not penalised for elements you simply do not have.

• Pass (100) — combined coverage is at least 90%.
• Partial (50) — combined coverage is between 60% and 90%.
• Fail (0) — combined coverage is below 60%.

These bands match the published rubric directly.

How do I fix common issues?

• Add a descriptive alt attribute to every meaningful image, for example <img alt="Product dashboard screenshot">.
• Mark purely decorative images with alt="" (empty alt) so they are skipped rather than counted as unlabelled.
• Give every <button> and icon control an aria-label, or visible text that names its action.
• Label form inputs with aria-label, aria-labelledby, or an associated <label>. Visible text alone does not count for inputs.
• Fix your weakest dimension first. The two ratios are averaged, so a high image score cannot rescue a page full of unlabelled controls.
• Re-scan to confirm combined coverage reaches 90% or higher.