Alt text writing

What alt text writing is

Alt text is the written description attached to an image. A screen reader speaks it aloud, and a browser shows it when the image fails to load. Writing it properly means working out what the image is doing where it sits, and describing that. Not listing what is in the picture.

The distinction is the whole job. A photograph of two people at a counter means one thing on a careers page and another on a page about a point-of-sale system. On the careers page the description covers the atmosphere of the workplace; on the product page, the terminal in the assistant's hand. The pixels are identical. The useful description is not.

It is not the same service as the site-wide alt text fill, which is a batch job for a site with hundreds of empty image fields and is priced and run that way. This service is for the images that need judgement: charts, diagrams, product photographs, floor plans, before-and-after pairs, screenshots of a process, and any picture that carries information the surrounding text does not repeat.

When you need this

When the images are part of the argument. A report whose findings live in its charts. A guide that explains a procedure in photographs. A catalogue where the difference between two models is visible and not written down. A technical document full of diagrams. In each case a reader who cannot see the images loses the content, not the decoration.

It is also right when a document has to go to a public body, a funder or a large customer with accessibility requirements, or when someone has run an automated tool over your files and handed you a list of missing descriptions nobody in the office knows how to fill in.

You do not need this if your site has a great many images that are mostly stock photography beside text that already says everything. That is a volume problem rather than a judgement problem, and the batch service is cheaper and does the job. I will say which you need on the first call.

What I write

Every image is read with the text around it, because alt text written without the surrounding page is guesswork.

• Informative images. What the image tells the reader, in a sentence, in the voice of the page. Long enough to carry the information and no longer, because a screen reader user cannot skim a long description the way a sighted reader skims a picture.
• Charts and diagrams. Described as their finding, not their shape. “Bar chart with five bars” is useless. “Repeat orders rise in each of the five years, with the sharpest rise after the second” is the thing the chart was drawn to say. A chart carrying more than one finding gets a short alt text plus a longer description in an appendix, and I write both.
• Decorative images. Left with empty alt text, deliberately. A border, a texture, a stock handshake beside a paragraph that already says everything: describing these makes a document slower and more irritating to listen to. Knowing what to leave silent is half the skill.
• Images that contain words. Screenshots, posters, quotation cards, infographics. The words are reproduced, because otherwise they exist for some readers and not for others.
• Photographs of people. Described for what matters in context, without guessing at anything the page does not already state about the people in them.

Captions are checked at the same time. A caption and its alt text should not be identical, because a screen reader reads both and the listener hears the same sentence twice.

What you get

The description list. One spreadsheet: file name, where the image appears, how I classified it, the alt text, and a longer description where one is needed. It is ordered so somebody can work down it and paste each entry into the right field without decisions.

The marked document. Where the images sit in an editable file, I put the descriptions in and hand back the finished document rather than a list for somebody else to apply.

A short note on conventions. One page: how I handled charts, what I marked decorative and why, and the sentence patterns to reuse, so whoever adds the next image can describe it without asking anyone.

A worked example

An illustration, not a client. An Ottawa trade association publishes a 48-page annual report with 31 images: fourteen charts, nine photographs from members' sites, five diagrams of a certification process, and three decorative bands across section openers.

The automated check its designer ran reported all 31 as missing, which is where the association started. The work goes the other way. The three decorative bands get empty alt text, so a listener moves past them in silence. The fourteen charts are described by their findings, and four of them turn out to carry two findings each, so those get a paragraph in an appendix that any reader can use, sighted or not. The five process diagrams are described as numbered steps, which is how the process actually runs, and the association realises that text is also the written version of the diagram it has twice been asked for. Two member photographs carry captions identical to the drafted descriptions, and one of each pair is cut. The report goes out with 28 descriptions, 3 deliberate silences, and an appendix the membership committee decides to keep in future editions.

How it runs

1. A half-hour call. What the document is, how many images, how many are charts, and who signs off on accessibility at your end. No charge.
2. A fixed price in writing. Based on the image count and the mix, since a chart takes several times as long as a photograph.
3. You send the images in place. The document, the layout file or the page addresses, so I see each image with its surrounding text. A folder of loose images is the slow way.
4. The writing. Classify, describe, then read the document back in order to catch repetition between captions and descriptions. Two to five working days for a report of that size.
5. Delivery. The list, the marked document where applicable, and the conventions note.

What it costs

Alt text writing is quoted as a fixed price in CAD after the half-hour call. It is priced per image rather than per word, and depends on how many of your images are charts and diagrams rather than photographs. The pricing page explains the four ways I work. The number is agreed in writing before I start, and if the batch service suits you better I will say so then.

What happens next

The descriptions go in, and that part of the document is finished. If the same document has headings that are only big bold text, the copy side of document accessibility deals with that in one pass, and if it is full of links reading “read more”, rewriting the link text is a small separate job. Where a chart proved undescribable because it is genuinely unclear, redrawing it is design work I can brief and manage. This piece of work is complete on its own, and nothing further is assumed.