The complete 2026 guide to EXIF, IPTC and XMP for image SEO

Every image carries three parallel metadata systems, and they were designed by different committees for different reasons. EXIF describes the capture (camera, exposure, GPS). IPTC describes the content (title, caption, keywords, creator, copyright). XMP is Adobe’s XML wrapper that can hold both and is the format most modern tools actually read. Getting image SEO right is mostly about knowing which block a given consumer reads.

The fields that actually matter

After embedding metadata across thousands of production images, a short list does the heavy lifting:

• IPTC Object Name / Title — a concise, specific headline. This is what many CMSs and stock platforms surface first.
• IPTC Caption/Description — one or two full sentences. Search engines increasingly parse this for context.
• IPTC Keywords — a flat list of 8–12 focused terms. More is not better; relevance is.
• IPTC/XMP Creator & Copyright — required by every serious stock validator and the basis of image licensing signals.
• XMP Alt Text (Iptc4xmpCore:AltTextAccessibility) — the accessibility layer crawlers and screen readers both use.

Where files go wrong

The most common failure isn’t missing metadata — it’s metadata that gets silently stripped. Resizing pipelines, social platforms and even some CMS upload handlers discard everything but the pixels. If you write a beautiful IPTC block and then run the file through an unaware optimizer, you ship a blank image.

The fix is to make metadata the last step of processing, and to verify the output rather than trust it. Read the embedded fields back out of the final file — not the source — before you publish.

Writing at scale

Hand-editing metadata doesn’t survive a catalog of any size. Standardise a template per image type (product, editorial, stock), generate the caption/keywords from what’s actually visible, and embed in one pass. That is exactly the workflow Image SEO King automates: decode, write EXIF/IPTC/XMP, convert, and hand back a file you can verify.

Takeaway: pick the eight fields above, write them honestly from the image content, and confirm they survived the pipeline. That single discipline outperforms any amount of keyword stuffing.