Keywords Getty Refuses — Even From Its Own Vocabulary

Getty and iStock keyword against a controlled vocabulary: a fixed list of terms, and anything outside it is not a keyword as far as the system is concerned. So the obvious rule is "use terms from the vocabulary". The obvious rule is not enough. Contributors keep sending us keywords that sit squarely inside the vocabulary and were still refused on submission, and after comparing the vocabulary we ship against 33,027 published iStock images, we can finally say why — and why you cannot tell which terms those are by reading them.

Two questions that look like one question

"Is this a valid keyword?" is really two questions, and they have different answers:

1. Is the term in the controlled vocabulary? This is a lookup. The answer is yes or no and it never changes between images.
2. Will the term be accepted on this image, by this agency, today? This depends on what the term actually denotes, on whether the picture supports it, and on decisions the agency does not publish.

Almost every keywording tool answers the first question and presents the answer as though it settled the second. It does not. The gap between the two is where submissions get held up, and it is wider than most contributors expect.

What contributors actually get refused

Acceptance data is easy to come by — every published image is proof that its keywords were accepted. Rejection data is rare, because a refusal happens inside a contributor's own submission screen and nobody publishes it. When contributors send us theirs, we keep it.

One contributor submitted a batch of video clips and sent us the exact terms the system would not recognise on submission:

• inside · health · hair type · emotion · individual · looking to the camera · editorial · active · black

Read that list again. There is nothing exotic in it. These are ordinary English words, the kind any honest description of a photograph would use, and several of them are in the vocabulary. They were still refused.

Since then we have had further reports — a single word one time, a two-word phrase another. The single word had 395 uses across the published images we sampled, and other contributors are still shipping it. So "refused" is not even stable across accounts and dates. That is worth saying plainly rather than pretending there is a clean rule underneath.

23% of the vocabulary never appears on a real image

To get past anecdote we compared the controlled vocabulary as we ship it — 24,632 terms — against the keyword lists of 33,027 published iStock images.

Nearly a quarter of the vocabulary does not appear on a single one of those images. That sounds alarming until you look at what is in it: Aardvark, 5th Century BC, Acadia National Park, 10 Seconds or Greater. The vocabulary covers the whole world of imagery. A sample of 33,027 images cannot possibly touch all of it.

The commonest cause: the term is a name, not a description

Here is the one we can explain end to end, because it happened on a real batch last week.

A Christmas-market shoot came back with City Lights on most of its frames. Reasonable-looking keyword for a city at night; it is in the vocabulary; the submission system flagged it.

The reason is visible the moment you look at the neighbouring entry in the vocabulary:

The vocabulary entry is a proper noun. It is not the general idea of a lit city. And the give-away in the data matched: that term appears on none of the 33,027 published images, while it was going out on 4.6% of ours. Published contributors describe that scene with City Life, Cityscape, City Street, Illuminated and Night — all common, all safe.

It is a whole family, not one term

Once you know to look, the pattern is everywhere in the vocabulary. These are all real entries:

This is why the vocabulary uses those Name - Qualifier forms. A term written Belgrade - Serbia or Split - Croatia is unambiguous; the bare word is a coin toss the system has to resolve without you. Where a qualified form exists, use it — it costs you the same slot and removes the guess.

The second cause: the term describes nothing

Go back to that refused list. inside, health, emotion, active, individual. Every one of them is a category, not a description of a picture. Nobody searching a stock library types "emotion"; they type what the emotion is.

Two of them also have exact, accepted counterparts, which is the clearest sign that the vocabulary has an opinion about wording:

These are not synonyms the system is being fussy about. They are the difference between a term that exists and a string that does not. A controlled vocabulary matches the exactstring — a lowercase "video" does not register against Video, and "looking to the camera" does not register against Looking At Camera at all.

The third cause: the word means something else here

black was on that refused list, and it is the most instructive one. In a stock vocabulary that word sits in the middle of a people-description system, so a keyword you intended as the colour of the coat lands in the same space as terms about ethnicity. The vocabulary has separate, unambiguous entries for both jobs — a colour term for the coat, a people term for a person — and the bare word is neither.

The general rule: if a word does two jobs in English, the vocabulary almost always has two entries, and the bare word is the one that fails. Reach for the specific entry.

Terms the agency adds for you

There is a happier reason a keyword can be missing from a published list: the agency applies it at ingest and does not need it from you. Getty stamps facets like the medium, whether the image is colour, and the orientation onto assets as they come in.

So spending keyword slots on those is not an error that gets you refused — it is just waste. On a 50-slot limit, three or four slots spent restating what the system already knows are three or four concepts a buyer might have searched for that you did not include. We cover the ingest-side detail in the ESP CSV guide.

What to do about it

1. Prefer the qualified form. If the vocabulary offers Name - Qualifier, use it. It is the same slot and it removes the ambiguity that gets terms held up.
2. Be suspicious of two-word phrases that read like a title. If the phrase could be the name of a film, a shop, a band or a landmark, check what the vocabulary means by it before you ship it.
3. Drop the category words. Terms like emotion, health, activity or individual describe a shelf, not a photograph. Replace each with the specific thing it was standing in for.
4. Match the exact string. Case and wording both count. Bind your list to the vocabulary before export rather than after a refusal.
5. Keep your own rejection list. When a term is refused, write it down. It is the only feedback the system gives you, it is not published anywhere, and — as our own reports show — it does not always agree with itself between accounts.

Sources & further reading

• Getty Images — Preparing your content (contributor login required)
• The Getty controlled vocabulary, explained
• The Getty ESP CSV, field by field
• Why stock photos get rejected, by reason
• How many keywords each platform wants
• The complete Getty & iStock contributor guide