Getty Images reads a small, strict set of metadata on every file you send: a descriptive title, a description that answers who, what, where and when, an accurate keyword set, creator and copyright details, and location and date. Files that arrive with those fields thin, vague or wrong get held or rejected long before anyone judges the picture itself.
What Getty counts as metadata
Two things travel with your file, and contributors often confuse them. The first is embedded metadata: the IPTC and XMP fields written inside the JPEG or video file itself, which you set in Lightroom, Bridge, Photo Mechanic or a dedicated metadata tool. The second is the record Getty builds in its contributor portal, where files are grouped into batches and reviewed.
Both matter. Embedded fields are what a reviewer sees first and what survives if the file moves between systems. The portal record is what ends up attached to the asset customers search. Getting them to agree is most of the work, and it is the single biggest reason organised contributors clear review faster than disorganised ones.
The fields, one at a time
Title
The title is a plain description of what is in the frame, not a caption with personality and not a sentence stuffed with search terms. Think of it as the line a picture editor would write in a hurry: subject, action, setting. A title like Young woman checking phone at a bus stop in the rain does more work than either Urban solitude or Woman phone rain city street lifestyle commuter.
Titles also feed Getty's own keyword suggestions, so a vague title costs you twice: the asset reads badly to a buyer, and the tooling around it has less to work with.
Description
The description carries the context the title cannot: who is in the shot, what they are doing, where it was taken, when, and where it matters, why. For creative content this can be written naturally. For editorial content it has to be factually correct, because the caption is part of the licence, not decoration.
A useful discipline is to write the description as if the reader cannot see the image at all. If your sentence still fully explains the scene, it will hold up under review and it will read well on the asset page.
Keywords
Keywords are where most Getty rejections start, and the failure is almost always accuracy rather than volume. Every term has to be defensible against the frame in front of you. If the picture shows two people at a table with laptops, meeting is fair, negotiation is a guess, and startup funding round is a story you invented. The keywording rules Getty enforces reward restraint far more than they reward reach.
A workable structure is to build outward in layers: literal subject first, then action, then setting, then the conceptual terms a buyer might search. Stop at the point where you would have to argue for a word. That is the line reviewers are checking, and it is also the line beyond which extra terms dilute the asset rather than surfacing it.
Creator, copyright and credit
These are the fields contributors set once and then forget to check. The creator name and the copyright line have to match the name on your contributor agreement, not a studio nickname or an old business name. Mismatches here are slow to resolve because they are an account question rather than an image question, and they can hold an otherwise clean batch.
Location and date
City, state or region, and country, plus the date the content was created. For creative work this supports search and rights checks. For editorial work it is load-bearing: a wrong date or a wrong city can make a caption factually false, which is a different class of problem from a weak keyword.
Embedded metadata versus what you type in the portal
If you write your metadata into the file before upload, most of the portal work disappears. If you do not, you are typing the same information twice and inviting the two copies to drift apart. Understanding how IPTC, XMP and embedded keywords actually work is the difference between a workflow that scales past a few hundred files and one that does not.
Two habits prevent most of the pain. Write metadata at the point where you cull, not at the point where you upload, so the information is attached while the shoot is still fresh in your head. And export in a way that preserves IPTC rather than stripping it, which some resizing and conversion scripts do silently.
Write your metadata in English. Getty's contributor documentation, review process and search all operate in English, and a keyword set in another language will not do the job even if the image is flawless.
The technical requirements that sit next to your metadata
Perfect metadata will not save a file that fails on specification. Getty's own guidance on preparing content sets out the constraints you need to clear before metadata matters at all:
- Photos must be at least 3 megapixels and no larger than 256 megapixels. A 2000 by 1333 pixel file falls under the minimum.
- Use RGB colour space, save as JPEG at the highest setting, and do not upsize to reach the minimum.
- Vector EPS files must be CS6 compatible and must be paired with a high quality raster version sharing the same filename.
- Raster illustrations have a minimum of 1900 by 2800 pixels, with 4075 by 4075 recommended.
- Video clips should run at least 5 seconds, with 30 seconds as the suggested ceiling unless the content justifies more, and files cap at 6GB.
Getty also asks contributors to submit their strongest frames rather than a full series. Near identical files from the same shoot can be flagged as duplicate content and rejected, which means the culling decision is a submission decision, not just a taste one.
Creative and editorial need different metadata
Creative content is licensed for commercial use, so the metadata is about findability and the legal state of the frame: releases attached where people or private property appear, no visible trademarks or logos, keywords that reflect both the literal scene and the concepts a designer would search.
Editorial content is licensed on the basis that the caption is true. Event names, dates, places and the identity of the people in frame have to be accurate and specific. Treating an editorial caption as a creative description is one of the more expensive mistakes a contributor can make, because the correction is rarely just a metadata edit.
Getty does not accept generative AI submissions
This is the point where Getty diverges sharply from most of the microstock market. Getty's contributor guidance states plainly that it does not accept content created with generative AI models such as Stable Diffusion, DALL-E or Midjourney. It draws a line around that: 3D renders are still acceptable, and ordinary digital editing tools like Photoshop and Illustrator are unaffected.
If you work across several agencies, this is a routing decision rather than a metadata one. Content that carries generative AI disclosure for Adobe Stock or Shutterstock should simply not be in the Getty batch. The wider picture of how agencies treat metadata for generative AI images varies enough that a single universal batch is no longer realistic.
Where metadata quietly gets files rejected
- Keywords that describe an interpretation rather than the frame. Reviewers check against the image, not against your intent.
- A description that repeats the title instead of adding the who, what, where and when.
- Creator or copyright fields that do not match the contributor account.
- Metadata stripped during export, so the file arrives technically fine and informationally empty.
- An editorial caption written in creative language, or a creative file captioned as if it were news.
- Releases missing on frames where a recognisable person or private property appears.
A Getty metadata workflow you can repeat
- Cull first and hard. Pick the strongest frames and drop the near duplicates before you write a single keyword.
- Set creator, copyright and location once as a template, and apply it to the whole shoot.
- Write the title and description per file, while you still remember the shoot.
- Build keywords in layers: literal subject, action, setting, then concept. Stop where you would have to argue.
- Embed everything into the file, then export without stripping IPTC.
- Attach releases, split editorial from creative, and only then build the batch.
The order matters more than the tooling. Everything that can be set once at shoot level should be set once, and the per file work should be limited to the things that genuinely differ. If you want the submission side rather than the metadata side, the Getty upload workflow covers batches, FTP and limits.
Doing this across more than one agency
Getty's field set is not Adobe's, and neither is Shutterstock's. Contributors who deliver to several agencies end up maintaining one clean master record per file and mapping it outward, because rewriting metadata per destination does not survive contact with volume. That mapping problem is what Rastock AI is built around: generate titles, descriptions and keywords once, hold them to each agency's rules including banned and mandatory terms, then deliver over FTP to 10+ agencies with per agency status you can actually see.
There is no revenue share and no ownership claim on your files, and IPTC and CSV export stay open so you can leave with everything. The 14 day free trial does not ask for a card, which is enough time to run a real Getty batch through it and compare the result against your current process.
Frequently asked questions
What metadata does Getty Images actually require?
Every submission needs a descriptive title, a description covering who, what, where and when, an accurate keyword set, creator and copyright details that match your contributor agreement, and location and date information. Editorial submissions additionally need factually correct event names, dates and places, because the caption forms part of what is being licensed.
Should I embed metadata in the file or type it into the portal?
Embed it in the file. Writing IPTC and XMP fields before upload means the information travels with the asset, survives moves between systems, and removes most of the manual work in the contributor portal. Typing the same content twice invites the embedded copy and the portal record to drift apart, which is a common source of avoidable corrections.
Does Getty Images accept AI-generated images?
No. Getty's contributor guidance states that it does not accept content created with generative AI models such as Stable Diffusion, DALL-E or Midjourney. The restriction is specific: 3D renders remain acceptable, and ordinary digital editing in tools like Photoshop or Illustrator is unaffected. If you produce generative work, it needs to go to agencies that accept it.
What are Getty's minimum file requirements?
Photos must be at least 3 megapixels and no more than 256 megapixels, in RGB, saved as JPEG at the highest quality setting with no upsizing. Raster illustrations start at 1900 by 2800 pixels with 4075 by 4075 recommended, and vector EPS files must be CS6 compatible and paired with a raster version. Video runs from 5 seconds, capped at 6GB.
How many keywords should I add to a Getty submission?
Enough to describe the frame accurately and no more. Getty reviews keywords for accuracy rather than rewarding volume, so terms you could not defend against the image work against you. Build outward in layers from literal subject to action, setting and concept, and stop at the point where a reviewer could reasonably disagree with a word.
Why do clean images still get rejected at Getty?
Usually for reasons that sit beside the picture rather than in it. Near-duplicate frames from the same series get flagged as duplicate content, metadata gets stripped during export, creator and copyright fields do not match the account, releases are missing where a recognisable person appears, or an editorial caption is written in creative language instead of factual terms.