How to keyword for Getty Images: controlled vocabulary

If you contribute to Getty Images or iStock, the single biggest factor in whether your photos get found is not how many keywords you add — it is whether those keywords exist in Getty's controlled vocabulary. Here is what that vocabulary actually is, why it quietly decides your sales, the traps that catch even careful contributors, and a practical keywording workflow that survives Getty and iStock review.

What "controlled vocabulary" means

A controlled vocabulary is a fixed, curated list of approved terms. Instead of letting every contributor invent their own words, Getty maintains one master list — closer to a structured dictionary than a free-text box — and every keyword on your image is matched against it.

Think of it like a library. You can describe a book however you like in conversation, but the library catalog only files it under official subject headings. If you invent your own heading, nobody browsing the catalog will ever land on your book. Getty's search works the same way: buyers find images through the controlled terms, so a keyword that is not in the vocabulary does little or nothing for discoverability.

Why it decides whether you get found

There are three practical consequences for a contributor:

• Unmatched keywords are wasted.Type "cosy autumn vibes" and there is no such entry. The vocabulary has Cozy and Autumnas two separate terms, and nothing for "vibes". The slot is spent, the discoverability is not gained.
• Irrelevant keywords can hurt you.Stuffing popular but inaccurate terms ("keyword spamming") degrades Getty's search quality, and reviewers reject or down-rank files that do it. More keywords is not better — accurate keywords is better.
• The right term unlocks many searches at once. Because the vocabulary is structured, one precise term can quietly attach your image to a whole family of related searches.

The hidden superpower: hierarchy, synonyms, and translations

Getty's controlled vocabulary takes a single valid keyword and expands it in three directions automatically:

• Hierarchy. A specific term inherits its broader parents. Tag Labrador Retriever and the vocabulary already knows that a Labrador is a Dog, and that a dog is an Animal — you do not have to spend slots saying so.
• Synonyms. A term is mapped to its synonyms, so a buyer typing a near-equivalent word can still reach a file tagged with the approved one.
• Translations. The same term is searchable in other languages — a buyer in Madrid searching perro or in Berlin searching Hund can still land on your Dog.

So one precise keyword quietly becomes a whole web of searchable access points, across many languages, without you typing any of them. That is exactly why specificity beats volume: the broad terms come free, the specific ones do not.

The forms you cannot guess

This is the trap that catches experienced contributors, and it caught us: when we checked every example on the first version of this page against the vocabulary, ten of them were not in it. Each was a perfectly sensible English word. Each was the wrong form.

And there is no rule you can learn instead of checking. People by gender are plural (Women, Boys), but family roles are singular (Mother, Son), and so are Child and Teenager. Some concepts are shortened (Beginnings), some are lengthened (Mental Burnout), and some carry a qualifier after a dash because the bare word means more than one thing (Partnership - Teamwork). That dash form is the same mechanism behind a whole class of rejections — see why Getty rejects keywords that look valid.

The practical rule: never type a Getty keyword from memory. Pick it from the vocabulary (the ESP autocomplete, or a tool that matches against the list) so you get the exact form the search index uses.

Women and Men are not a head count

Because the vocabulary puts gender in the plural, it is easy to read Womenas "several women". Getty does not use it that way. We measured it on 33,028 published iStock images:

Half of all images tagged Women show one person. Women and Men say who is in the frame, not how many. The count lives in its own terms: One Person, One Woman Only, Two People, Three People, Small Group Of People, Large Group Of People.

This matters beyond tidiness, because other terms depend on the count. Diversity or Multiracial Group on a single-person portrait is simply wrong, and so is anything that implies a second person. We learned it the hard way: our own pipeline once read Women as evidence of a group and let a collective term through on frames showing one person. So give the count its own keyword, and never let a gender term stand in for it.

Some terms ask a question you have to answer truthfully

A few vocabulary terms are not about what is in the frame but about how the person relates to the camera — and they are easy to get backwards.

• Looking At Camera means your camera. A photo of someone taking a selfie shows a person looking straight into a lens — their phone's, not yours. From where the buyer stands, they are looking away. That frame is Selfie or Photographing, not Looking At Camera. The selfie itself, taken by the phone, is the one where the tag is true.
• Candid and Looking At Camera never go together. Candid means the subject is not engaging with the photographer. Across our 33,028 published images, the two appear together on none of the 1,076 tagged Candid. If both are on your list, one of them is wrong.

What keyword order does, and does not, do

You will read everywhere that "the first ten keywords carry the most weight". That rule is real, but it is Adobe Stock's, and Adobe states it in its own guidance. We have not found Getty publishing an equivalent rule for its search, so we do not repeat it here as fact.

Leading with your strongest terms still costs nothing and helps in two practical ways: it keeps the list honest while you trim it, and the same list is ready to reuse if you also submit elsewhere. A sensible order:

1. How many people — One Person, Two People, Small Group Of People.
2. Gender and age range — Women, Mid Adult Men. Ethnicity only as the model release states (more on this below).
3. What they are doing — the main action.
4. Time and place — Day, Indoors, location, country.
5. Key objects in the frame.
6. Then broader and conceptual terms.

If you submit the same image to Adobe as well, order stops being optional — and the list itself has to change, because the two agencies' vocabularies overlap far less than you would expect. We measured it in one photo, two keyword lists.

A repeatable method: Who / What / Where & When / Why / How

The fastest way to produce a complete, balanced keyword set is to interrogate your own image with five questions:

• Who? people count, gender, age, ethnicity (from the release), relationships.
• What? the action and the objects.
• Where & When? location type, country, season, time of day.
• Why? the concept or emotion a buyer is searching — Togetherness, Working At Home, Beginnings.
• How? the technique — Selective Focus, Aerial View, Slow Motion.

Worked example — a photo of a father and son at a dining table with a laptop, every term checked against the vocabulary: Two People, Mid Adult Men, Boys, Father, Son, Using Laptop, Working At Home, Sitting, Dining Room, Day, Indoors, then concepts like Connection, Real Life, and Concentration. The five questions force you to cover both what is literally in the frame and what the image is about.

Literal vs. conceptual keywords

Getty wants both kinds of terms, and strong files balance them:

• Literal keywords describe what is visibly in the frame: Women, Laptop, Kitchen, Window, Morning.
• Conceptual keywords describe the idea, mood, or use: Working At Home, Efficiency, Solitude, Beginnings.

Buyers searching for a feeling (Freedom, Mental Burnout, Togetherness) buy on the conceptual terms — and those are exactly the terms inexperienced contributors leave out. But every conceptual term still has to exist in the vocabulary and has to be honestly defensible from the image. Do not tag Happiness on a neutral face just because it sells.

Do not spend slots on what Getty adds for you

Some terms appear on almost every published Getty and iStock image, and it is tempting to conclude you must type them. You do not — Getty applies them itself when the file is ingested. In our sample of 33,028 published iStock images:

• Photography is on 96.2%, Horizontal on 90.7%, Color Image on 82.2%. These are technical facets of the file, stamped at ingest.
• Ethnicity terms come from the model release attached to the submission — which is the authoritative source anyway.

Every slot you spend on a facet Getty would have added is a slot not spent on something only you can say about the picture. The Getty ESP CSV guide covers the rest of what the upload file does and does not need.

The mistakes that sink most submissions

1. Free-text terms. Slang, brand names, hashtag-style phrases, and made-up compounds that are not in the vocabulary.
2. Keyword spamming.Adding 50 loosely related terms to "cover all bases." It lowers relevance and invites rejection.
3. Guessing the form. Woman for Women, Burnout for Mental Burnout — see the table above. A near-miss matches nothing.
4. Over-broad only. Tagging just Animal, Nature, Outdoors on a photo that clearly shows a fox in snow. You are competing with millions of files on the broad terms and invisible on the specific ones: Fox, Snow, Winter.
5. Letting a gender term count people. Women on a solo portrait is fine; Diversity on the same portrait is not.
6. Inviting the wrong associations. Because Getty auto-expands every term, a sloppy or wrong specific keyword drags in a whole branch of irrelevant searches. Pick the term that is exactly right, not merely close.

Titles do double duty

A title is not decoration. A concise, accurate title describing the main subject helps buyers scan results, supports search ranking, and — inside Getty's submission tool — actually fuels keyword suggestions. Write the title as a plain description of what the image is, not a poetic caption. We cover the shape that works across agencies in how to write stock photo titles.

Keywording people: be factual and respectful

Keywords describing a person's age, gender, or ethnicity should reflect how that person represents themselves, and the accurate source for that is the model release — not your assumption from the photo. This is both an accuracy issue and a respect issue: guessing demographics from appearance is unreliable and can misrepresent a real person. When you do not have that information, keep demographic terms general or leave them out.

Tips from working contributors

• Reverse-engineer the best-sellers. Search your exact concept on the agency, sort by best-selling, and study the metadata of the top files. The vocabulary that already sells is the vocabulary to learn from.
• Think in neighbors — then look them up. For Business: Corporate Business, Professional Occupation. For Teamwork: Cooperation, Partnership - Teamwork. Notice that two of those four are not the word you would have typed.
• Build presets. Keep reusable keyword sets for recurring subjects (portraits, product, landscapes) and apply them per batch — then trim to each file.

A practical Getty keywording workflow

Here is the routine that consistently passes review and ranks — whether you do it by hand or with a tool:

1. Describe the literal scene first. Subject, setting, action, count, time of day, season, dominant colors.
2. Add the most specific true terms. Exact breed, location type, object names. Specificity is where the hierarchy pays off.
3. Layer in honest conceptual terms. The idea or emotion a buyer would search to license this exact image.
4. Check the count and the gaze. A count term of its own; Looking At Camera only if they look into your lens; never alongside Candid.
5. Match everything to the controlled vocabulary. Drop any term that is not an approved Getty term, skip the facets Getty adds itself, then cut to the strongest you can defend to a reviewer. Getty allows up to 50, but quality beats filling the cap.

Where PixTagger fits in

The slowest, most error-prone step above is matching every term against Getty's controlled vocabulary by hand. That is exactly the step PixTagger automates.

PixTagger runs vision AI over your photo or video, generates literal and conceptual keywords, and then maps every term to its exact form in Getty's vocabulary before it ever reaches your export — a near-miss is mapped to the entry it means, and a term with no genuine match is dropped instead of wasting a slot. The checks in this guide are built in: the people count comes from the picture, not from a gender term, and Looking At Camera is held to your lens. The Getty CSV it produces is import-ready for the Getty ESP. For video, it samples three frames across the clip (at 10%, 50% and 90%, skipping the head and tail padding), or five on request for clips where the action changes.

Don't want to match every term against the list by hand? PixTagger generates valid Getty keywordsfor you — each one mapped to the controlled vocabulary and checked against what's actually in the image — then exports the ESP-ready CSV.

Sources & further reading

• Getty Images Contributor Community — keywording help
• Keywording for stock photography — Shutterstock blog
• Our own measurements: every term on this page checked against Getty's controlled vocabulary (24,631 entries); co-occurrence figures from a sample of 33,028 published iStock images collected in 2026. The sample was gathered by keyword search, not at random, so treat the percentages as strong evidence of how Getty uses a term, not as exact library-wide rates.