How do I write alt text and SEO metadata for AI-generated product photos?

Quick Answer

Write alt text as a literal description of the product first and an attribute second: subject, key detail (color, material, angle), and use context, in 5 to 15 words and under about 125 characters. Use hyphenated, descriptive filenames like navy-leather-tote-front.jpg instead of IMG_2841.jpg, keep the on-page caption and surrounding copy consistent with the alt text, and put any "AI-generated" disclosure in the platform's disclosure field or an embedded Content Credential, not in the alt attribute.

Write the alt text for the buyer, not the bot

Alt text has two readers: a screen-reader user who cannot see the image, and a search engine deciding what the image actually shows. Both want the same thing, which is a literal description of the product.

A working pattern is [product] + [key attribute] + [context or use]. Examples:

• Navy leather tote, front view, on a marble counter
• Cream cashmere crewneck on a tall male model, three-quarter angle
• Amber glass perfume bottle, 100ml, with brass cap

Keep it between 5 and 15 words and under roughly 125 characters, which is where Google's image SEO guidance says alt text becomes useful without becoming spam, and where most screen readers stop reading aloud comfortably. Mention one relevant keyword once if it fits naturally. Do not repeat it. Do not stuff brand or category terms the image itself does not show.

If the image is purely decorative (a divider, a texture strip, a background flourish), use an empty alt="" so screen readers skip it. The W3C WAI Images Tutorial is the authoritative source on which images need text alternatives and which do not.

The full per-image metadata stack

Alt text is one field. A product image actually carries several, and they each have a job:

A few things follow from this split:

• Filenames should be lowercase, hyphenated, and descriptive before upload. Rename the file in the export, not after the platform has hashed it.
• The caption and surrounding product copy should not contradict the alt text. If the alt says "front view" and the caption says "back detail," one of them is wrong.
• Embedded provenance is a separate field. It does not replace platform disclosure rules.

The AI-specific part: accuracy and disclosure

AI-generated product imagery makes alt text more load-bearing, not less. With shot photography the file usually matches what the photographer was looking at. With AI imagery the alt text is the only ground truth a buyer or screen-reader user has for what the rendered file actually shows. Describe what is in the frame, not what the prompt asked for. If the image went through Edit Shortcuts (Try On, Recolor, Reframe) or Photoshoot variants, write fresh alt text per variant. Reusing one description across four images breaks accessibility and image search.

Resist the urge to put "AI-generated" inside the alt attribute. It does not help screen readers, it dilutes the descriptive keywords for image search, and it does not satisfy any disclosure rule. Disclosure belongs in the platform's own field (an Etsy checkbox, a Shopify product description note, a Meta ad disclosure) or in an embedded C2PA Content Credential that machine-readable systems can verify. Sibling articles on disclosure and misleading-content rules cover where each platform expects that signal.

When you are writing alt text for a Nightjar-generated Asset, open the Asset detail view in the Library to confirm the product, the ingredients used, and any model and background choices before describing the file. That detail view is the source of truth for what the rendered image actually contains.