How to Write Stock Photo Titles and Descriptions

The title is the one piece of metadata every stock agency reads, and no two agencies agree on what it should look like. Shutterstock rejects a title that is not a sentence of at least five words. Adobe Stock wants a short phrase, ideally under 70 characters. Pond5 caps it at 80. Getty keeps the title and the description as two separate fields, and Shutterstock folds them into one. Put an event photo into the editorial lane and each agency asks for the date in a different order. This guide puts every agency's published rules side by side, shows the one line that passes all of them, and adds something the help pages cannot tell you: how often a real batch ships the same title twice. We measured it, and before we added a check it was one file in four.

Title, description, caption: which field is which

Before the rules, the vocabulary, because the agencies use the same words for different fields:

• Adobe Stock has a Titleand no separate description. Its sample upload CSV has exactly five columns — Filename, Title, Keywords, Category, Releases — and marks the title "up to 200 characters".
• Shutterstockhas one field it calls "description/title". It is what buyers see as the title, it can run to 2,048 characters, and it is where an editorial caption goes.
• Getty Images and iStock keep a title and a description as two columns in the ESP upload CSV. Buyers see both, and they are not interchangeable: the title is the headline, the description is the caption.
• Pond5has a required Title (40–80 characters) and an optional Description meant to "expand on what is already stated in the title".
• Alamy calls it a caption. A caption plus at least five tags is the minimum for an image to go on sale, and Alamy reads it straight from the IPTC Description field of an uploaded file.
• 123RF asks for a description, which works as the title.

So when an agency says "title", check whether it means a headline or the only descriptive text the file will ever carry. That decides how much you can say in it.

Every agency's title rules on one page

Dreamstime also requires a title, a description, keywords and categories on every file before it is properly findable; its help pages are the place to check its current limits. Depositphotos reads the title and description embedded in your file when you upload it.

The one line that passes everywhere

Lay the rules on top of each other and the space they share is narrower than it looks, but it exists:

1. 40 to 70 characters.Long enough for Pond5's floor, short enough for Adobe's ideal and well inside Pond5's cap.
2. At least five words, reading as a clause. Shutterstock needs a sentence; Adobe wants a phrase, not a paragraph. A plain subject-verb clause with no full stop satisfies both.
3. No commas.Pond5's own examples use them, but Adobe's CSV guidance does not want them, and on Shutterstock commas between single words are what make a title look like a keyword list.
4. Subject, then action, then setting. Pond5 recommends this order outright. It also puts first the words Adobe tells you to repeat in your first ten keywords.
5. Only what the picture shows or you know for certain. No brand, no person's name, no camera model, no adjective a reviewer cannot verify.

Where the agencies genuinely disagree

Mood, or only facts?

Shutterstock tells you to "capture the mood and emotion behind your content". Adobe asks for descriptive, not subjective, language, and Pond5 tells you to avoid filler such as "stunning" and "beautiful". You can satisfy both by writing the mood as something visible. "Laughing" and "hugging" are things a reviewer can see in the frame. "Joyful", "enjoying" and "beautiful" are your opinion of it.

Ethnicity in the description?

123RF says to leave a model's ethnicity out of the description and put it in the keywords instead, its example being "Man drinking coffee" rather than naming his background. Shutterstock does not forbid it, but requires identity terms to be accurate and respectful. Either way, the only reliable source for how a person identifies is the model release or the person, not your reading of the photograph.

How much description?

Pond5 wants "a few interesting and entertaining sentences or a brief paragraph": weather, time of day, season, number of people, and even Latin species names. 123RF prefers the short version, "Three business people having a discussion", over a longer one that lists outfits and expressions, since those details belong in the keywords. Write the long version once and cut it for the agencies that want it short.

One file in four shared its title with another file

Pond5 says each item should have a unique title. Vecteezy requires one. Shutterstock asks for a description that "uniquely identifies your content". None of them says how often contributors fail at this, and we can.

Until August 2026 PixTagger warned about repeated titles in an ordinary batch but did not stop them. That gave us a clean record of what ships when nothing enforces it. We took every batch of five or more distinct files tagged before the check existed: 164 batches, 7,485 files, from 12 accounts.

• 26.1% of files carried exactly the same title as at least one other file in the same batch.
• 73.8% of batches contained at least one repeated title.
• In the worst batch, 30 files shared a single title.
• It was not one account's habit: the largest account repeated on 27.1% of its files, and all the other accounts together on 24.2%.

These titles were written by an AI model, not by hand, but the mechanism is the same one that catches people. A shoot produces runs of similar frames, and whoever describes frame 14 sees the same woman, the same bicycle and the same garage as in frame 13. The honest description really is the same.

Then we forced every title in an export to be unique, and a second number showed that exact uniqueness is a low bar. Across 4,161 files in batches of five or more tagged since, there are no exact repeats, but 14.4% of files still open with the same four words as another file in their batch. A title that differs only in its last word is unique to a database and identical to a buyer scanning a results page. The shortcut of adding a number to make it unique is the one Pond5 rules out by name.

Watch the verbs that read minds

The most common subjective word in a stock title is not an adjective. It is a verb. "Enjoying", "admiring", "loving" and "dreaming of" all claim to know what is going on inside someone's head. A reviewer cannot confirm it, and a buyer searching for a woman drinkingcoffee is not typing "enjoying".

We learned this from our own pipeline. Our instructions to the model forbid subjective and mind-reading words in titles, and the model still slipped often enough that we added a final pass that rewrites them regardless: "enjoying" becomes "having", "admiring" becomes "looking at", and adjectives like "excited" and "beautiful" are removed. If a model that has been told explicitly still writes them, a tired contributor at file 200 will too. Search your finished CSV for "-ing" verbs of feeling before you upload.

What only you know goes in the description

Everything so far can be read off the pixels. The facts that make a title worth more usually cannot:

• Place. A cobbled street looks like fifty cities. 123RF says to give the exact location name when it matters to the image, and only when you are confident about it.
• Roles and relationships. Two people at a workbench are just two people. Only you know that one owns the shop and the other started this week.
• The event. Shutterstock asks you to name the event in the title of any event coverage, especially where credentials were required.
• The date. Mandatory for documentary editorial, as the next section shows.

We hit the roles problem while auditing a real shoot tagged in our own tool. A frame of a manager giving a maintenance worker instructions came out described as "two colleagues reviewing documents in an office corridor". The shoot note named both roles, but not which frame showed which pair, and no camera can tell a manager from a colleague. We did not write a rule to guess. That detail now comes only from a note attached to that specific file, because a confident wrong relationship in a title is worse than none at all.

Another audit taught us the opposite risk. A fact that is true of one frame will drift into its neighbours if you write a batch together. We caught a town name from one file turning up in the titles of other files from a different town, because the titles already written were being passed along as context. If you batch-paste a description, reread the place names on every file.

So write the long, specific version in the description, where Pond5 and Getty expect it, and keep the title to what is true of that one frame: "Shop owner teaching a new mechanic to true a bicycle wheel" is 58 characters, and every word of it came from knowing the people.

Editorial captions: same facts, different order

Documentary editorial is where titles stop being a style choice. Each agency publishes a format, and they do not match:

One invented event, written for each:

• Shutterstock: Novi Sad, Serbia – March 14, 2026: Cyclists ride through the city centre during a car-free day.
• Pond5: Novi Sad, Serbia – 2026, March 14: Cyclists ride through the city centre during a car-free day.
• 123RF: Novi Sad, Serbia, March 14, 2026 – Cyclists ride through the city centre during a car-free day.

One practical catch: that caption is 94 characters, which is over Pond5's 80-character title limit. Pond5 lets the caption live in the description instead, so give Pond5 a normal title and put the dateline there.

Video titles name the shot

For footage, the type of shot belongs in the title. Pond5's recommended order ends with it, and its model example is "African Oystercatcher Bird Feeding in Estuary, Eats Worm, Wide Angle". Adobe's metadata guidance for video likewise asks for the camera technique in the title as well as the keywords. So a clip becomes "Slow push-in on a woman truing a bicycle wheel in a repair shop". Leave out what describes the file rather than the content: Vecteezy asks for no "4K" or "footage" in the title. Resolution, codec and frame rate have their own fields, and the full video spec comparison is in stock video metadata requirements for every agency.

A pre-upload title checklist

1. Is it 40–70 characters and at least five words?
2. Does it read as a clause, not a comma-separated list?
3. Is it free of brands, people's names, camera models, links and emojis?
4. Does it contain only facts: no "beautiful", no "enjoying"?
5. Does it differ from its neighbours in the first few words, not just the last one, and with no number codes?
6. Are the title's key words also in the first ten keywords?
7. Is it in English, apart from place names and Latin species names?
8. For editorial: is the dateline in thisagency's order?
9. Is it spelled right? 123RF does not let you edit a description after the file is accepted.

Sources & further reading

• Adobe Stock — Titles and keywords
• Adobe Stock — sample upload CSV (column layout, 200-character title)
• Shutterstock — Content Publishing Standards: Contextual Metadata
• Shutterstock — Why was my content rejected for title or keywords?
• Shutterstock — Description and keyword best practices
• Shutterstock — How do I submit editorial content?
• Pond5 — Tagging & pricing your files (titles, descriptions, editorial captions)
• Pond5 — Master your metadata
• 123RF — Complete guide to describing and keywording photos and vectors
• 123RF — Complete guide to describing and keywording editorial images
• Alamy — Captioning and keywording FAQ
• Vecteezy — Contributor checklist
• How every stock agency wants your metadata
• Why stock photos and video get rejected