Which Alt Text Is Most Appropriate? Real Examples and AI Mistakes

Which alt text is most appropriate for an image?

The honest answer is that it depends on why the image is there, what information it provides and what the surrounding content already explains.

A detailed description is not automatically better than a short one. An empty alt attribute is not automatically a mistake. A technically accurate description can still be completely unhelpful if it focuses on irrelevant details.

Alternative text is sometimes called an alt tag, and our main guide explains what it is and how it should be used. This article looks at real examples and explores how to choose the most useful wording for a particular image.

The examples began with 3 prominent stock images that I found on the home page of a website. The alternative text attached to them could hardly have been more different.

Work, lifestyle and business concept. Good-looking focused European female in trendy glasses sitting in cafe near laptop, working on notebook, surrounded with books, making notes for lesson with kids
About_hero
An empty alt attribute

One description was extremely long and subjective. One was simply a filename. One said nothing at all.

Which was the most appropriate?

The answer cannot be determined by counting the words. We need to understand the purpose of each image first.

Why Context Determines The Most Appropriate Alt Text

Alternative text is not a literal inventory of everything visible inside an image.

Its purpose is to communicate the information or function that somebody would otherwise miss because they cannot see the image.

The same photograph may therefore need different wording on different pages.

Imagine a photograph of a woman sitting at a cafe table with a laptop and notebook.

On a page about flexible working, the important detail may be that she is working remotely from a cafe.

On a page promoting a particular laptop, the important detail may be the product she is using.

On a page about accessible workspaces, the important information may be the layout, lighting or equipment visible in the scene.

On a page where the photograph is merely decorative, the image may need an empty alt attribute.

The pixels have not changed. The purpose has.

Ask What The Image Adds To The Page

A useful starting question is:

What would somebody miss if this image disappeared?

If the answer is important information, the description should communicate that information.

If the answer is only visual decoration, an empty alt attribute may be correct.

If the image functions as a button or link, the wording should describe the action or destination rather than the appearance of the picture.

The W3C alternative text decision tree provides a useful process for deciding whether an image is informative, functional, decorative or complex.

Do Not Repeat Everything Already Written Nearby

Alternative text does not need to repeat an identical caption, heading or paragraph.

Suppose a staff photograph appears beside the visible text:

Jane Williams
Head of Customer Services

The description does not need to say Jane Williams, Head of Customer Services unless the image itself is being used as a link or the visible wording is not reliably associated with it.

The image could use a short visual description if appearance is part of its purpose. It could also use an empty attribute if the photograph is merely decorative.

Repetition makes pages longer and more tiring without necessarily making them more accessible.

Comparing Three Very Different Alt Text Examples

The original website used 3 stock images grouped prominently on its home page.

The descriptions demonstrate 3 common approaches: an excessive stock-library description, a meaningless filename and an empty attribute.

The Overlong Stock Photograph Description

The first example read:

Work, lifestyle and business concept. Good-looking focused European female in trendy glasses sitting in cafe near laptop, working on notebook, surrounded with books, making notes for lesson with kids

This resembles text supplied by a stock photography library. It appears designed to help people find the image rather than to help somebody understand its purpose on the website.

Several parts are problematic.

Good-looking is subjective and unnecessary.
Trendy is another opinion rather than useful information.
European may be an assumption based on appearance.
Work, lifestyle and business concept sounds like stock-library categorisation.
The description lists objects without establishing which details matter.
The sentence is much longer than most simple photographs require.

The description may contain accurate visual details, but accuracy alone does not make it appropriate.

A more useful version for a page about remote working might be:

Woman taking notes beside her laptop while working in a cafe

The revised wording communicates the activity and setting without the stock-library clutter.

The Meaningless Filename

The second example was:

About_hero

This probably began as an internal filename or label used by the designer.

It tells the visitor almost nothing about the image.

The word About may reveal where the image appears. The word hero may indicate that it is a large banner. Neither communicates the content, purpose or function of the picture.

File naming conventions belong behind the scenes. They should not become the experience announced to a screen reader user.

If the image provides useful information, it needs a meaningful description. If it is decorative, the correct solution is an intentionally empty attribute rather than a filename.

The Empty Alt Attribute

The third image had an empty alt attribute:

alt=""

This might have been completely correct.

An empty attribute tells assistive technology that the image should be skipped. It is the appropriate choice when an image is decorative, repeats information already available or adds no useful meaning.

An empty attribute is different from a missing attribute.

The empty attribute is a deliberate instruction. A missing attribute may cause a screen reader to announce the filename or other unhelpful information.

We cannot decide whether the empty description was correct without seeing the image in context.

This is exactly why the longest description is not automatically the winner.

When Empty Alt Text Is Better Than A Description

There is a persistent belief that every image must contain a written description.

That belief can create pages where screen reader users must listen to descriptions of decorative shapes, repeated logos, background textures and generic stock photographs that add no meaningful information.

Accessibility is not measured by how many words a page announces.

The objective is to communicate useful content without unnecessary noise.

Decorative Images

A decorative image may make a page more visually attractive without contributing information.

Examples may include:

Background flourishes
Decorative dividing graphics
Generic stock photographs used only for atmosphere
Icons repeated beside visible text
Images that duplicate an adjacent caption completely

These images should normally use:

alt=""

This allows the screen reader to move past them without interruption.

Icons Beside Visible Text

A telephone icon displayed beside a visible phone number does not usually need to be announced.

The number itself provides the useful information.

Announcing Phone icon before every phone number would add repetition without improving understanding.

The same principle applies to envelope icons beside email addresses and location pins beside written addresses.

Images Described Fully In Nearby Text

A chart may be followed immediately by a complete explanation of its findings. A photograph may have a visible caption that communicates everything important. An illustration may repeat the same point made in the paragraph beside it.

Repetition may not help.

The decision depends on whether the image itself contributes anything beyond the nearby text.

The W3C guidance on decorative images explains why these images should normally be hidden from assistive technology.

Can AI Generate Appropriate Alt Text

I placed the original images into PowerPoint and used its automatic image-description feature.

The generated descriptions were:

A person sitting at a table using a computer
A group of people sitting at a table
A group of young girls running in a hallway

These descriptions were clearer than About_hero and less cluttered than the stock-library paragraph.

They were still only guesses based on the pixels.

The software could identify people, tables, computers and movement. It could not understand why the images appeared on the page or which details mattered to the organisation.

AI Can Provide A Useful First Draft

Automatic descriptions can help when a website contains a large image library or when an editor feels uncertain about where to begin.

They may correctly identify:

People and objects
Basic actions
Locations and settings
Visible text
Colours and simple relationships

A generated description may therefore be a reasonable first draft.

It should not be published without review.

AI Does Not Understand Page Purpose

A photograph of somebody using a computer may appear on a page about remote working, cybersecurity, accessible software, staff training or office furniture.

The visible content is the same, but the relevant meaning changes.

AI can describe the scene. It may not understand why that scene matters in this particular context.

A generic description such as person using computer may therefore be technically correct while failing to communicate the point of the image.

AI Can Invent Or Misinterpret Details

Automatic systems can make confident mistakes.

They may misidentify objects, guess ages, infer gender, misunderstand an activity or claim to see details that are not present.

Descriptions involving race, ethnicity, disability, emotion or relationships need particular care.

A person standing beside a wheelchair is not necessarily its user. Two people smiling at each other are not necessarily a couple. Somebody looking down may be concentrating rather than sad.

Human review is needed before assumptions become published facts.

AI Can Produce Too Much Detail

Some generated descriptions attempt to mention every object, colour, expression and item of clothing.

The result can be grammatically impressive and practically exhausting.

A blind visitor does not necessarily need a miniature novel every time somebody uploads a photograph.

Useful descriptions prioritise meaning.

How To Improve AI Generated Alt Text

Automatic descriptions should be treated as drafts rather than finished content.

The following process can turn a generic description into something more useful.

Check Every Claimed Fact

Confirm that the people, objects, actions and setting have been identified correctly.

Remove assumptions that cannot be verified.

If the software describes somebody as a young woman, ask whether age and gender are known or merely inferred from appearance.

Remove Irrelevant Detail

Delete objects and visual details that do not help explain the image.

A computer, notebook, coffee cup and pile of books may all be visible. The reader may only need to know that somebody is studying at a cafe table.

Add The Missing Purpose

AI may identify the scene without understanding why it appears on the page.

Add the detail that connects the image to the content.

Generic AI draft:

Person using a laptop at a desk

Improved description on an accessibility-audit page:

Blind tester reviewing a booking form with a screen reader

The second description explains why the image is relevant.

Decide Whether The Image Needs A Description At All

AI systems tend to assume that the objective is always to generate text.

The correct answer may be an empty attribute.

Ask whether the image adds information or merely supports the visual design.

Creating a description for every decorative stock photograph can make a page less pleasant to navigate.

A Simple Test For Choosing Appropriate Alt Text

Use the following questions whenever you are unsure what to write.

Why is the image on this page?
What information would somebody miss if they could not see it?
Is that information already available in nearby text?
Is the image informative, functional, decorative or complex?
Does the description focus on the image purpose rather than every visible detail?
Have any subjective judgements or assumptions been included?
Would a shorter description communicate the meaning more clearly?
Could an empty alt attribute be the correct choice?
Has any AI-generated wording been checked by a person?
Does the description make sense when read as part of the complete page?

Reading the page aloud can expose repetition and unnecessary detail.

Testing it with a screen reader gives a much clearer understanding of how the description fits into the wider experience.

An automated checker may confirm that an alt attribute exists. It cannot reliably decide whether the wording is appropriate.

That requires context, judgement and an understanding of the people using the page.

Our website accessibility audits combine technical review with testing by disabled people who use assistive technology in everyday life.

This helps organisations understand whether their descriptions provide meaningful access rather than merely filling an HTML field.

The most appropriate alt text is not necessarily the longest, the most detailed or the version generated by the cleverest tool.

It is the wording that communicates the relevant meaning of the image within the context of the page.

Sometimes that will be a clear sentence. Sometimes it will be a few words. Sometimes the correct answer will be silence.

Good accessibility is not about making every image speak. It is about making sure every image communicates appropriately.

Clive Loseby

Access by Design