Alt text examples for the images people actually publish

W3C/WAI separates images by purpose: informative images add content, functional images start actions, and decorative images add no information. Complex images, such as detailed charts, may need a short alt text plus a longer description in the page.

Page context can change the answer. A flower photo beside an article about the flower is informative. The same photo used as a background behind a heading may be decorative.

A product name is useful when it identifies the item, but it often misses the visible reason for showing another photo. I add the color, material, angle, or variant detail that distinguishes the image. I leave out claims the photo cannot prove.

On a product page, nearby text may already name the item. The alt text can then concentrate on the view or detail. In a product grid, the product name may be needed because there is less surrounding context.

For a screenshot, I name the application or page and the state the reader needs to notice. I transcribe an error message when the wording matters. I do not list every button in the interface when the article only points to one field.

The surrounding instructions still do the teaching. Alt text makes the relevant screen state available when the screenshot is missing.

A simple chart can fit in one sentence when the main point is the information the page needs. A detailed chart cannot. W3C/WAI recommends a short description that identifies a complex image and a longer text alternative that carries its essential values, relationships, and trends.

I put the detailed explanation in normal page content when possible. Readers can scan the same values, and the alt attribute does not turn into a dense paragraph.

For an editorial photo, I describe the people, objects, setting, and action that matter to the story. I do not guess a person's identity, mood, background, or other details that the image does not establish.

A staff portrait may only need the person's name when the photo identifies them. If the nearby heading already gives the name and the portrait adds no other information, the author can decide that the image is decorative.

When an image is the only content inside a link or button, I name what happens. W3C/WAI uses this rule for functional images: describe the action or destination instead of the pixels.

If visible link text already names the same destination, an icon or logo inside that link can use empty alt text to avoid repeating it.

I use a generator to get the visible facts onto the page, then edit the draft for its role. The model can see the image. It cannot reliably know which nearby sentence already explains it, where a link goes, or which chart values the reader needs.

For one image, I upload it, generate a draft, and compare the draft with the page. For a catalog, I generate rows in bulk and review variant details before import. I run the final text through the validator when I want a quick check for empty, generic, long, or keyword-stuffed wording.