Describing People in Alt Text Without Guessing Identity, Age or Mood

Separate observation from interpretation

Describing a person begins with a distinction that is easy to overlook: what is visible is not necessarily what is true about the person. A photograph can show someone wearing a uniform, standing beside a medical instrument or holding a book. It cannot automatically establish that the person is a doctor, a patient or a teacher. Those interpretations may be correct, but the image alone does not provide enough evidence.

A useful description reports observable details and uses surrounding editorial context for established facts. For example, "A person wearing a white coat stands beside a laboratory bench" describes the scene. If a reliable article identifies the person as a laboratory scientist, that information can be included when it matters to the image's purpose. The distinction prevents descriptions from turning plausible guesses into factual statements.

• Observation: what can be seen directly.
• Context: information established by reliable accompanying material.
• Inference: a conclusion that the image alone does not establish.

Identity needs reliable context

Names can be important in photographs of public figures, interview subjects, event participants or people whose identity is central to a story. Include a name when the publication has reliable information establishing who is pictured. Do not identify someone simply because an automated tool suggests a familiar face or because the person's appearance resembles a known individual.

The same principle applies to gender, ethnicity, nationality and relationships. Appearance does not establish how someone identifies, where they come from or how they are related to another person. If the information is relevant, use a reliable source or the person's own stated information. If it is unavailable, describe the visible activity or context instead.

This approach is especially important when writing alt text for people in photos used in education, journalism and social media. A mistaken identity can change the meaning of an image and may attach a person to an event, statement or organization without justification.

Age and mood are difficult to establish visually

Exact age is rarely necessary in alternative text, and visual estimates can be unreliable. A photograph may support broad language such as "a child," "an older adult" or "a young-looking person" when the distinction is genuinely relevant, but even these terms should be chosen carefully. Avoid invented ages such as "a 42-year-old man" when no reliable information supports the number.

Mood presents a similar problem. A smile is visible; happiness is an interpretation. A person looking down may be reading, thinking, resting or simply looking at something outside the frame. For example, "A woman smiles while speaking at a microphone" reports observable behavior. "A woman happily celebrates her success" adds an emotional explanation that the photograph may not establish.

When the surrounding story establishes an emotional context, that information can be included with appropriate attribution or wording. An article may explain that a person is celebrating an award. The description can reflect that context without claiming that facial expression alone proves an internal feeling.

Choose details that help readers understand the scene

A practical description usually needs only a few details: who or what is central, what the person is doing and which surroundings matter. Consider a photograph of two people at a table. "Two people sit across from each other at a small cafe table, with notebooks open between them" gives useful spatial and activity information. It does not need speculative labels such as colleagues, friends or business partners.

Clothing can be relevant when it distinguishes people or contributes to the story. A description of a sports event may mention team uniforms. A workplace photograph may mention protective equipment. For an unrelated portrait, a complete clothing inventory may add little. The editor should select details based on the image's role, not on a fixed requirement to mention every visible feature.

Position and action can also help identify the central subject in group photographs. Phrases such as "the person at the left" or "the speaker in the center" are useful when there are several people and the accompanying content provides no names. Avoid long descriptions that force readers to remember multiple visual identifiers without a clear reason.

AI-generated alt text can sound certain while being wrong

Machine-generated alt text often produces fluent descriptions of people, which can make uncertain details appear authoritative. A tool might infer a person's profession from clothing, describe a facial expression as an emotion or assign a relationship to people standing together. These statements can read naturally even when the image provides insufficient evidence. Fluency is a writing quality, not proof of accuracy.

AI-generated alt text accuracy depends on the image, the system, the prompt, available context and the way a description is used. An automated description may correctly identify a person holding an umbrella while missing that the umbrella is the subject of a product comparison. It may also overlook a relevant sign or introduce a confident but incorrect identification. Different tools can produce different descriptions of the same image.

Descriptions produced by image description tools should therefore be treated as drafts. An editor should compare each factual claim with visible evidence and reliable context. Remove unsupported names, exact ages, occupations, relationships and emotional conclusions. Keep details that contribute to understanding and rewrite vague language into a clear account of the scene. This is a more useful standard than asking whether an automated description sounds convincing.

• Check names against reliable publication information.
• Replace guessed occupations with visible clothing or activities.
• Remove exact ages that lack supporting information.
• Distinguish facial expressions from claimed emotions.
• Compare the description with the image's actual editorial purpose.

A practical editing method for descriptions of people

Begin by identifying the reason the image appears. A portrait introducing an interview may need the subject's established name. A photograph illustrating a public activity may need only the action and setting. A group photograph may require a brief explanation of who is central, especially when the image accompanies a discussion involving several participants.

Next, underline every claim in the proposed description that goes beyond visible details. Ask whether the claim comes from reliable context, whether it matters and whether the wording makes its source clear. If the evidence is weak, remove the claim or replace it with an observable detail. For example, change "A tired employee looks frustrated after a long shift" to "A person sits at a desk with their head resting on one hand."

Finally, read the revised description without looking at the image. It should communicate the important information without relying on assumptions or unexplained references. Check that it does not repeat a nearby caption and that its language respects the people depicted. Apply these three editing steps to the next portrait or group photograph in the publishing queue.

Questions readers ask