IPTC, XMP and EXIF Metadata Explained: Which Fields Stock Marketplaces Actually Read
A technical breakdown of legacy IPTC-IIM datasets, Adobe XMP Dublin Core packets, container behavior across JPEG, EPS, and MP4, and why embedded titles sometimes disappear on upload.
Technical Guides · · 9 min read · by Salman Ahmad
Understand the difference between EXIF camera data, IPTC-IIM binary records, and XML-based XMP Dublin Core packets—plus which exact tags Adobe Stock, Shutterstock, and Freepik read across JPEG, TIFF, EPS, AI, and MP4 files.
Every contributor has experienced the same puzzling moment: you type a careful title and 45 keywords into an image editor, upload the file to two different stock agencies, and discover that Agency A populated everything cleanly while Agency B left the title blank or truncated accented characters into garbled symbols.
The culprit is almost never random server failure. It happens because "embedded metadata" is not a single unified block inside a media file—it is three overlapping historical standards (EXIF, IPTC-IIM, and XMP) stored in separate binary segments that often disagree with one another when only one standard is updated.
Once you understand which block each file format carries and which exact field names stock ingestion servers prioritize, missing titles and encoding bugs become straightforward to diagnose and prevent.
The Three Metadata Layers Inside a Stock File
Camera telemetry, legacy press records, and modern XML packets live side by side.
When you inspect a photograph or illustration in ExifTool or Adobe Bridge, three distinct standards can exist inside the same file:
• EXIF (Exchangeable Image File Format): Created for digital cameras. Stores shutter speed, aperture, ISO, lens focal length, color space, orientation flags, and optional GPS coordinates. Stock portals rarely use EXIF for buyer search—though some display camera specs or strip GPS data for privacy.
• IPTC-IIM (Information Interchange Model): A compact binary standard developed in the 1990s for news agencies. Inside JPEG (`APP13` Photoshop IRB segment) and TIFF files, it stores fixed record numbers such as `2:05` (ObjectName / Title), `2:120` (Caption-Abstract / Description), and repeatable `2:25` (Keywords).
• XMP (Extensible Metadata Platform): Introduced by Adobe as UTF-8 encoded XML using RDF and Dublin Core namespaces (`dc:title`, `dc:description`, `dc:subject`). Unlike legacy IPTC-IIM, XMP natively handles Unicode characters, works inside vector PostScript/PDF streams (`.ai`, `.eps`, `.svg`), and embeds cleanly into QuickTime/ISO-BMFF video containers (`.mp4`, `.mov`).
| Concept | IPTC-IIM Record | XMP Dublin Core Tag | Primary Agency Readers |
|---|---|---|---|
| Title / Headline | 2:05 ObjectName (64 B legacy) | dc:title (rdf:Alt / x-default) | Adobe Stock, Freepik, Dreamstime, Vecteezy |
| Caption / Description | 2:120 Caption-Abstract (2000 B) | dc:description (rdf:Alt / x-default) | Shutterstock, 123RF, Dreamstime, Vecteezy |
| Keywords / Tags | 2:25 Keywords (repeatable) | dc:subject (rdf:Bag -> rdf:li) | All 6 agencies on JPEG; XMP on EPS/SVG |
| UTF-8 Marker | 1:90 CodedCharacterSet (ESC % G) | Native XML encoding='UTF-8' | Prevents mojibake on accented characters |
Minimal Standards-Compliant XMP Dublin Core Packet Inside APP1 / PostScript:
<?xpacket begin="\uFEFF" id="W5M0MpCehiHzreSzNTczkc9d"?>
<x:xmpmeta xmlns:x="adobe:ns:meta/">
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
<rdf:Description xmlns:dc="http://purl.org/dc/elements/1.1/">
<dc:title>
<rdf:Alt><rdf:li xml:lang="x-default">Barista steaming milk at espresso machine in specialty cafe</rdf:li></rdf:Alt>
</dc:title>
<dc:description>
<rdf:Alt><rdf:li xml:lang="x-default">Close-up of barista steaming milk in stainless pitcher behind cafe counter</rdf:li></rdf:Alt>
</dc:description>
<dc:subject>
<rdf:Bag>
<rdf:li>barista</rdf:li>
<rdf:li>espresso machine</rdf:li>
<rdf:li>steaming milk</rdf:li>
<rdf:li>specialty coffee</rdf:li>
</rdf:Bag>
</dc:subject>
</rdf:Description>
</rdf:RDF>
</x:xmpmeta>
<?xpacket end="w"?>Key takeaways
- EXIF describes how the camera captured the frame; IPTC-IIM and XMP describe what the frame means to a buyer.
- XMP (`dc:title`, `dc:description`, `dc:subject`) is always UTF-8 XML and works across raster, vector, and video containers.
- Writing both synchronized IPTC-IIM (UTF-8 coded character set `1:90`) and XMP Dublin Core maximizes compatibility on JPEG uploads.
Title vs Description Collisions: Why Agencies Read Different Fields
Adobe Stock has one Title field, while Shutterstock uses Description as its primary caption.
The single most common source of confusion across stock agencies is how each ingestion pipeline maps `dc:title` (ObjectName) versus `dc:description` (Caption-Abstract):
• Adobe Stock: Displays a single Title field (up to 200 characters). On upload, Adobe Stock checks XMP `dc:title` / IPTC `ObjectName` first, and falls back to `dc:description` / `Caption-Abstract` if `dc:title` is empty. Caution: if Windows Explorer or Lightroom wrote the camera filename (`DSC_0492`) into `ObjectName`, Adobe Stock will display `DSC_0492` instead of your rich caption!
• Shutterstock: Does not expose a separate short title to contributors; its 200-character Description field reads IPTC `Caption-Abstract` / XMP `dc:description` first.
• Dreamstime & Vecteezy: Read both fields separately (`ObjectName` → Title, `Caption-Abstract` → Description).
Because of these differing rules, leaving an old camera code in `dc:title` while only editing `dc:description` causes inconsistent imports across portals.
Worked example
Weak or risky title: dc:title = 'IMG_8821' | dc:description = ' Barista steaming oat milk in specialty coffee shop'
Clear descriptive title: dc:title = 'Barista steaming milk at espresso machine in specialty cafe' | dc:description = 'Close-up of specialty coffee barista steaming milk in stainless pitcher behind counter'
Populating both XMP/IPTC Title and Description intentionally ensures single-field portals (Adobe Stock) and caption-first portals (Shutterstock) both receive clean human prose.
Why Vector (.EPS / .AI) and Video (.MP4 / .MOV) Behave Differently From JPEG
Browsers can rewrite JPEG APP1 segments in memory, but PostScript and video atoms require container-aware writers.
In a JPEG file, XMP sits inside a well-defined `APP1` marker (`http://ns.adobe.com/xap/1.0/\0`) near the start of the byte stream. Even lightweight browser code can inject or replace that packet cleanly.
Vector and video formats are structurally very different:
• EPS (Encapsulated PostScript): Stores XMP inside PostScript DSC comment blocks (`%begin_xml_code` ... `%end_xml_code`) or embedded XMP packet wrappers with whitespace padding so byte offsets remain valid without breaking Illustrator's PostScript parser.
• AI (Adobe Illustrator): Modern `.ai` files are dual-stream PDF containers; modifying bytes naively corrupts the PDF cross-reference (`xref`) table unless updated through a PDF/XMP packet space or companion workflow.
• MP4 / MOV: Store metadata inside `moov/udta/meta` or top-level `uuid` XMP boxes. Moreover, not every stock portal reads embedded video atoms on upload—several agencies still require a companion CSV (see our Marketplace Metadata Limits reference) even when video XMP is valid.
Compliance Note: Never use a generic text editor to paste XML into an .ai or .mp4 binary—shifting byte offsets without updating container length headers or xref tables will corrupt the file and trigger an instant upload failure.
Fixing Garbled Characters and Ordered Keyword Arrays
UTF-8 coded character markers and rdf:Bag vs rdf:Seq behavior.
Have you ever seen smart quotes or hyphens turn into `—` after uploading a JPEG? That happens when software writes UTF-8 text into legacy IPTC-IIM without writing the Envelope Record `1:90` (`CodedCharacterSet = ESC % G`), causing older ingestion parsers to assume Latin-1 / MacRoman encoding.
Keyword ordering has a subtle technical detail too: in XMP Dublin Core, `dc:subject` is defined as an unordered `rdf:Bag`, yet in practice most XML parsers iterate elements in document order (`<rdf:li>` index 1 through N). Keeping your `<rdf:li>` items and IPTC `2:25` datasets in identical priority order preserves your top-10 keyword ranking on portals that respect tag order.
Practitioner Tip: Stick to clean ASCII punctuation (straight apostrophes and standard hyphens) in stock titles so legacy IPTC parsers never misread smart typography.
Summary
Embedded metadata works reliably when you stop treating it as a black box: populate both Title and Description deliberately, never leave camera filenames in ObjectName, keep XMP and IPTC synchronized in UTF-8, and keep a formatted CSV ready for portals whose video or vector pipelines ignore embedded packets.
Pre-Submission Review Checklist
- Clear camera codes ('IMG_001', 'Untitled-1') out of ObjectName / dc:title before exporting
- Write synchronized UTF-8 XMP (`dc:title`, `dc:description`, `dc:subject`) and IPTC-IIM records on JPEGs
- Preserve top-to-bottom keyword priority order across `<rdf:li>` tags
- Replace smart quotes and em-dashes with plain ASCII equivalents for legacy IPTC safety
- Spot-check one uploaded JPEG, EPS, and MP4 on each target portal before uploading a 200-file batch
Official Documentation & Standards Referenced
- IPTC Photo Metadata Standard (2024.1 Specification)
- Adobe XMP Developer Center — Part 1: Data Model, Serialization & Core Properties
- Dublin Core Metadata Initiative (DCMI) — Element Set 1.1
Related contributor guides
- How to Write More Distinct AI Stock Prompts: A Six-Part Framework
- Why Strong-Looking Stock Photos Still Get Rejected: Technical, IP and Moderation Diagnostics
- Seven Compliance Mistakes That Can Put a Stock Account at Risk
Tags: IPTC, XMP, EXIF, Dublin Core, Metadata Embedding. Written and reviewed by Salman Ahmad — see our editorial standards.
More resources: metadata limits by marketplace, free metadata checker, free browser tool, all guides.