computedText(node) returns an approximation of the text which would be read aloud by a screen reader.

test('Contrived example', () => {
  document.body.innerHTML = `
    <h1>Hello World</h1>
    <div style="display: none">I am hidden</div>
    <img src="foo.jpg" alt="Bar" />
  `;
  expect(computedText(document.body)).toBe('Hello World Bar');
});

Local development:

npm install
npm run test

Use cases:

• Verify that text which should be visible/invisible to a screen reader is marked up appropriately
• Run as an approval test in CI to detect changes in the text representation of components

Heavily based on the findTextAlternatives method in GoogleChrome/accessibility-developer-tools