Alt text for complex images: charts, maps and diagrams in PDFs, and where AI helps

Alt text for complex images works in two parts. A short alt text names the image and its main point, and a long description carries the information the image holds: the figures in a chart, the regions of a map, the steps in a flowchart. WCAG 2.1 asks for a text alternative that serves the same purpose as the image, and for a chart in a budget book that purpose is usually the numbers. This guide shows how to write both, with full examples for charts, maps, org charts, flowcharts and infographics, where a long description can live in a PDF, and where AI drafting helps.

What WCAG 2.1 asks of images in a PDF

Success criterion 1.1.1, Non-text Content, says that non-text content needs "a text alternative that serves the equivalent purpose," with a few exceptions. The one that matters in every PDF is decoration, which must be "implemented in a way that it can be ignored by assistive technology." In a PDF, the text alternative is the image's alternate text, the Alt entry on its Figure tag (W3C technique PDF1). A decorative image is marked as an artifact, which hides it from screen readers (technique PDF4). Together, those two choices decide what screen reader users, who hear the page read aloud or feel it on a braille display, get from each image.

The W3C's Images Tutorial (opens an external site) sorts images by what they do. Six of its categories are common in public documents.

Images of text

A scanned memo pasted in as a picture, or a flyer headline saved as an image, is text that a screen reader cannot read and a magnifier user cannot reflow. The fix is real text, recovered with OCR (see our guide to scanned PDFs) or retyped in the source. If an image of text has to stay, W3C's tutorial says "the text alternative must contain the same text presented in the image." Logos are the exception.

How to write short alt text that serves its purpose

W3C's tutorial puts it plainly: "The text alternative needs to be determined by the author, depending on the usage, context, and content of an image." Write for why the image is in this document, not for how it looks.

• Lead with purpose. Write what a sighted reader gets from the image here.
• Keep it short. Microsoft's guidance says alt text "shouldn't be longer than a sentence or two."
• Skip "image of." Screen readers already announce a graphic, so WebAIM calls the phrase redundant. Naming the kind of chart, such as "Bar graph," is fine, as Section508.gov advises.
• Don't repeat the caption. If nearby text already says it, keep the alt brief, or treat the image as decorative.

The same photo can need different alt text in different documents.

Complex images need a short alt and a long description

W3C's tutorial says complex images hold "more than can be conveyed in a short phrase or sentence," so they need a "two-part text alternative." The short description names the image, its main point and where the long description is. The long description is "a textual representation of the essential information conveyed by the image": the values, relationships, trends and exceptions a sighted reader gets by looking.

Alt text examples for charts, maps and diagrams

A bar chart in a budget book

• Short alt: "Bar chart of proposed FY2027 General Fund spending by department. Public Safety is the largest at $48.2 million, about half of the $98.0 million total."
• Long description: "Proposed FY2027 General Fund spending, largest to smallest: Public Safety, $48.2 million; Public Works, $21.7 million; General Government, $12.6 million; Parks and Recreation, $9.4 million; Library, $6.1 million. Total, $98.0 million."

If the same figures already appear in a table on the page, that table is the best long description, because a screen reader user can move through it cell by cell. W3C's technique G92 offers "the data table that was used to generate the line graph" as a long description. The short alt then ends with "Figures are in Table 4, below."

A locator map and a choropleth map

A locator map in a hearing notice shows where a site is, and the notice already gives the address, so one sentence does the job.

• Locator map alt: "Locator map showing the site at the northeast corner of Main Street and 4th Avenue, two blocks north of City Hall."

A choropleth map, which shades areas by a value, needs far more. In a county health report mapping flu vaccination by township, the data is the payload: a resident wants to find their own township.

• Short alt: "Map of 2025 adult flu vaccination rates by township. Rates are highest around the county seat and lowest in the rural south. Rates for each township are listed below the map."
• Long description, as text below the map: "Townships fall into four ranges. 60% or higher: Center, home of the county seat, and Mill Creek. 50–59%: Ashford, Lake and Union. 40–49%: Birch Grove, Fairview and Oak Ridge. Below 40%: Pine Hollow and Stony Point, both on the southern county line. The countywide rate is 51%."

Don't describe the shades of blue. Describe what they mean.

An org chart and a process flowchart

• Org chart short alt: "Organization chart of the Department of Public Works: four divisions report to the director."
• Org chart long description: "The Director of Public Works reports to the City Manager. Four division managers report to the Director: Streets, Water and Sewer, Fleet Services, and Engineering. Engineering oversees Development Review and Capital Projects."

Use names and titles exactly as printed. In the source, a list of who reports to whom can replace the chart.

• Flowchart short alt: "Flowchart of the building permit review process, from application to permit issued."
• Flowchart long description: "1. The applicant submits an application and plans. 2. Staff check the application for completeness. If it is incomplete, it goes back to the applicant to resubmit. 3. A complete application goes to Zoning, Building and Fire review at the same time. 4. If any reviewer requests changes, the applicant revises the plans and they return to review. 5. When all three approve, the applicant pays the fees and the permit is issued."

Follow every branch, and say where each "no" leads.

An infographic

• Short alt: "Infographic: the library's 2025 year in numbers."
• Long description: "In 2025 the library recorded 412,000 visits, lent 1.1 million items, held 2,300 programs attended by 58,000 people, and issued 6,400 new library cards. Digital loans rose 18% over 2024."

The icons are decoration; the numbers are content. Most infographics are a list in disguise, so real text in the source is often the cleanest fix.

Where to put a long description in a PDF

PDF has no separate long-description field, so you choose among three places.

When you control the source file, default to one of the first two. The third suits a PDF you cannot re-author: use short sentences, most important fact first, since a listener can't skim.

Can AI write alt text? Where it helps and where it doesn't

AI vision models give a fast first pass: they name the kind of image and read clear labels. They also fail in predictable ways.

• They describe what they see, not why it is there. A model does not know whether your building photo is decoration on a cover or evidence in a zoning case. A person does.
• They describe appearance instead of meaning. "A bar chart with blue bars" is accurate and useless. Microsoft's own guidance says of charts: "Try to convey the insight."
• They can invent content. On a diagram with small labels, a model that isn't given the text printed in the image can misread names, merge boxes or add plausible steps that aren't there.
• Placeholders are not alt text. Microsoft 365 can generate alt text automatically, and Microsoft says to review it and "remove any comments added there such as 'Description generated with high confidence.'" Files exported from older versions of Word can carry stamps such as "Description automatically generated." Treat these as missing alt text.

When you review an AI draft, check five things.

1. Does it say why the image is in this document?
2. Are names, numbers and labels exactly as printed?
3. Does it mention anything that isn't in the image?
4. Is the image really decorative, or really text?
5. For a complex image, does a long description carry the data?

How Ockem Proof uses AI for alt text

Ockem Proof, our product, uses AI to draft and people to decide. It finds every image, including ones the document never registered in its structure. Empty "ghost" figures, tags that claim a picture where nothing is drawn, are removed without review. Word-style placeholder alt text is treated as missing.

One picture is one question. A vision model drafts a short alt and an extended description in one place, with the figure crop beside the draft. Proof flags figures that look complex, such as a drawing with dozens of text labels, and grounds the description in the words printed inside the image, so names and labels come out as printed instead of guessed. A picture that is really text is answered "It's text" and transcribed. A quality gate blocks lazy descriptions like "image" from passing.

The extended description is stored in the image's alternate text after the short alt, separated by a blank line. Acrobat and screen readers already read it there, and it survives editing in other tools, with the one-block limit described above. Proof does not add visible text to your pages, so where a caption or table would serve everyone better, make that change in the source.

Describe an image once, and wherever the same image appears in your other documents, Proof offers your approved description as a draft to review.

What a person still decides:

• What the image is for. Meaningful, decorative or text is your reviewer's call.
• Every word. A person approves or edits each draft before anything is written into the file.
• Whether the draft is enough. Some figures need a table in the source.

Drafting runs on self-hosted models on hardware Ockem operates; no customer document content goes to a third-party model. Your documents are not used for model training unless your workspace turns that on. Details are on the alt text rule page.

See it on your own files

You can try one PDF (opens an external site) of up to 10 pages free, with no account. To count your PDFs first, use the free PDF inventory tool. The alt text rule page shows how Proof handles images. For what software settles on its own, read what software fixes and what needs a person, and for chart colors, see color contrast in a PDF.

Questions

This guide explains the rules in plain language. It isn’t legal advice. Check how they apply to your organization with counsel.