How every stock agency wants your metadata (2026): CSV columns, keyword limits & embedded XMP

Every stock agency wants the same three things — a title, a description and keywords — but almost none of them want it delivered the same way. Some take a CSV with columns in a fixed order. Some read the metadata embedded inside the file you upload and ignore CSVs entirely. Get the format wrong and the import silently fails, or your file sits half-tagged. This is a contributor's reference to how the major agencies actually accept metadata in 2026 — the columns, the limits, and the two very different delivery methods.

Two ways agencies take your metadata

Before the per-agency detail, the single most useful distinction: there are two delivery methods, and knowing which an agency uses saves you a lot of wasted uploads.

• CSV upload. You upload your files, then hand the agency a spreadsheet: one row per file, columns for filename, title, description, keywords and sometimes categories. The agency matches each row to a file by its exact filename. Getty, Adobe, Shutterstock, Pond5 and Dreamstime all work this way.
• Embedded metadata (IPTC/XMP). The agency reads the title, caption and keywords written inside the file — the same IPTC/XMP fields Lightroom, Bridge and Photo Mechanic manage. There is no CSV to submit; the metadata rides along in the JPEG. Alamy, Depositphotos and 123RF work this way.

The embedded route is the one contributors miss most often, because there is no button that says "upload keywords." The keywords have to already be in the file when it arrives — which is why an XMP sidecar (a small .xmp file your catalog reads and writes into the image on export) is the bridge for those agencies.

The reference table

How each major agency accepts metadata, the delivery method, and the keyword range. Where an agency uses a CSV, the columns must match its expected layout exactly.

Limits move over time and a few agencies change columns without much notice, so when a CSV won't import, the first thing to check is that your column headers match the current template exactly — a shifted or renamed column is the usual culprit.

The CSV details that break imports

Most failed CSV imports come down to a handful of formatting issues that have nothing to do with your keywords:

• Filename mismatch. The agency pairs a row to a file by its full filename, extension included. Tag the exact file you submit — a converted clip.mp4 won't attach to the clip.mov you deliver, and a stray leading space in the name breaks the match too.
• Encoding.Dreamstime expects the Excel "CSV UTF-8" flavour (with a byte-order mark); Getty's ESP importer, by contrast, rejects a file whose header carries that mark. A tool that knows the difference matters here.
• Missing required columns. Shutterstock needs a category from its list; Pond5 needs OriginalFilename spelled exactly. An empty required field fails the whole row.
• Title length.Pond5 wants 40–80 characters; a 120-char title that's fine on Adobe can be rejected on Pond5.

Getting keywords into the embedded-metadata agencies

For Alamy, Depositphotos and 123RF there is no CSV to fix — the workflow is different. You write the metadata into the file once, and it travels with every copy you export:

• Read an XMP sidecar (or the embedded IPTC/XMP) into Lightroom Classic or Bridge — one .xmp per photo, matched by filename.
• Your catalog now holds the title, caption and keywords. Every JPEG you export carries them embedded.
• Upload those files to Alamy, Depositphotos or 123RF and the agency reads the caption and keywords automatically — nothing to type, nothing to submit.

Tagging once for all of them

The reason to understand the formats is so you don't keyword the same shoot five times. A photo or video needs one accurate set of keywords; the only thing that changes per agency is the packaging — the columns, the limits, the CSV-versus-embedded method above.

That's exactly what PixTagger automates. You drop a batch in the browser and get a buyer-focused title, a description and keywords — mapped to Getty's controlled vocabulary so they pass validation — then export the right shape for each destination: a Shutterstock CSV with categories filled in, Adobe, Pond5 and Dreamstime CSVs, or XMP sidecars for the embedded agencies. One tagging pass, every marketplace — and because photos are resized in the browser and only sidecar text is generated, your originals never leave your computer.