HTML Inspector - SeleniumBase for Rust

The HTML Inspector scans a page for common accessibility and markup issues. It is a fast static check that can catch missing attributes, invalid references, and structural problems before they reach production.

Info

What you will learn

How to run an inspection.
Which checks are performed.
How to use the inspector in CI.
The scope and limitations of static analysis.

Run an inspection#

BaseCase::inspect_html takes the current page source and inspects it:

# use seleniumbase_rs::BaseCase;# async fn demo(sb: &BaseCase) -> Result<(), seleniumbase_rs::SeleniumBaseError> {let inspection = sb.inspect_html().await?;assert!(inspection.is_clean(), "{inspection:?}");# Ok(())# }

BaseCase::assert_no_html_issues does the same and fails with an AssertionFailed error that lists every [rule] message.

The inspector needs no browser. HtmlInspector::inspect takes any HTML string:

use seleniumbase_rs::api::html_inspector::HtmlInspector;# fn main() -> Result<(), seleniumbase_rs::SeleniumBaseError> {let html = r#"<html lang="en"><head><title>T</title></head>              <body><img src="logo.png"></body></html>"#;let inspection = HtmlInspector::inspect(html)?;assert!(inspection.issues.iter().any(|issue| issue.rule == "image-alt"));# Ok(())# }

Checks performed#

Every finding has a stable rule ID and a severity. is_clean() is true only when there are no findings at all, warnings included; has_errors() is true when at least one finding has InspectionSeverity::Error.

Rule ID	Severity	What it flags
`document-lang`	error	The `html` element has no non-empty `lang`.
`document-title`	error	The document has no non-empty `title`.
`meta-viewport`	warning	No viewport meta tag whose content contains `width=device-width`.
`skip-link`	warning	No link to `#maincontent` or `#content`.
`image-alt`	error	An `img` without an `alt` attribute (use `alt=""` for decoration).
`interactive-name`	error	A link (`a[href]`) or button without a static accessible name (text, `aria-label`, `title`, `alt`, or a valid `aria-labelledby`).
`link-text`	warning	A link with no visible text, or with generic text: “click here”, “here”, “read more”, “more”, “link” or “details”.
`nested-interactive`	error	A button inside a link.
`form-control-label`	error	A form control without a `label`, wrapping `label`, `aria-label`, or valid `aria-labelledby` reference.
`radio-group-legend`	warning	A named radio group of two or more not inside a `fieldset` with a `legend`.
`duplicate-id`	error	An ID used more than once.
`invalid-aria-reference`	error	`aria-labelledby` or `aria-describedby` naming an ID that does not exist.
`empty-heading`	warning	A heading with no text.
`heading-order`	warning	A heading level that jumps by more than one.
`page-h1`	warning	No `h1`, or more than one.
`table-headers`	warning	A table with no `th` cell.
`main-landmark`	warning	No `main` element or `role="main"`.

Access results#

# use seleniumbase_rs::api::html_inspector::InspectionResult;# fn print_issues(inspection: InspectionResult) {let issues = inspection.issues;for issue in issues {    println!(        "{:?} {} at {:?}: {}",        issue.severity, issue.rule, issue.selector, issue.message    );}# }

selector is an Option<String>: a best-effort selector for the element, or None for document-level findings.

Use in CI#

Fail a build when new markup issues appear:

# use seleniumbase_rs::api::html_inspector::InspectionResult;# fn gate(inspection: InspectionResult) {if !inspection.is_clean() {    std::process::exit(1);}# }

Tip

To fail only on errors and let warnings through, test inspection.has_errors() instead.

Why it matters#

Clean HTML improves accessibility, SEO, and test reliability.

Scope and limitations#

The inspector uses an HTML5 parser and deterministic static rules. Stable rule IDs and best-effort element selectors make findings suitable for CI allowlists. It does not execute JavaScript, compute styles or contrast, test keyboard behavior, or inspect the browser accessibility tree.

Warning

Use it as an early check alongside browser-based accessibility testing such as axe-core and manual keyboard and assistive-technology review. A clean result is not a claim of WCAG conformance.