What file metadata is
Every stock platform expects three descriptive fields: a title, a description and keywords — plus housekeeping like the category, an editorial flag, and model or property releases.
Inside the file, metadata sits in two layers. EXIF is technical camera data: exposure, ISO, capture date, GPS. IPTC/XMP holds the descriptive fields that search actually uses. When people say "metadata" about stock, they mean the IPTC/XMP layer.
Stock search is text search
The engine compares the buyer's typed query to your title and keywords. Systems like Adobe Sensei do analyse the image, but mostly to check that your keywords match what's visible in the frame — not to rank by pixels.
The practical consequence: precise, honest text is the whole game. A strong image with vague keywords stays invisible.
Title and description
A title is a natural phrase about what's in the frame, not a pile of words. Shutterstock rejects titles that are just keyword lists. Write it in your account language.
For commercial files, keep dates, place names and people's names out of the title — that's editorial style, and platforms reject it in commercial work. Write "protesters raise their hands during a march", not "May 30, 2020, protesters…". The description carries context the title has no room for.
Keywords: two layers
Literal — what is physically in the frame: object, place, colour, action. Conceptual — what the image means and what a buyer would use it for: "success", "work-life balance", "loneliness".
Most commercially valuable licences come from conceptual searches: a buyer looks for something that fits a brief, not "a photo of a person". A conceptual keyword still has to be defensible by the content — don't tag "happiness" on a neutral face.
Order matters
The first keywords carry the most weight. Adobe Stock states the first 10 positions have the most influence on ranking. Vecteezy puts extra emphasis on the first 5.
Other platforms don't publish a number, but putting your most precise, most commercial terms first never hurts and usually helps.
Fewer, but sharper
Filling every keyword slot is not the goal. Irrelevant keywords dilute your relevance signal, and stacked near-synonyms count as spam on some platforms.
Shutterstock's own example of bad keywording is a set like "autumn, fall, foliage, October, gold, yellow, orange, amber, fall colors" — two dozen words for just two ideas, the season and the colour, multiplied out with synonyms.
Broad terms ("nature", "woman", "business") put you up against millions of files. A long specific phrase ("female entrepreneur working from a mountain cabin") competes with almost nobody and lands exactly on a buyer's intent.
Singular vs plural — there is no single rule
Adobe Stock and Vecteezy officially say: enter one form of a word and the engine covers the rest — duplicating "dog / dogs" just wastes a slot.
Other platforms make no such promise, so don't count on the engine supplying the other form. Where both forms genuinely matter, add them yourself; otherwise default to the singular.
Getty and iStock use a controlled vocabulary
Getty and iStock don't take free tags — each word is matched to a fixed approved list. A matched word inherits broader terms, synonyms and translations: "Labrador Retriever" connects to "dog", "pet", "animal".
A word that isn't in the list still works as a literal search term but connects to nothing. This is the most common reason a file passes review and still gets no views.
Metadata in the file, or in a CSV
Adobe Stock and Getty read the IPTC/XMP fields embedded in the file on upload — fill them in once and the portal picks them up. Shutterstock is built around a separate CSV.
Most contributors do both: embed the metadata into the file once, then export the same data as a CSV in each platform's format. Note that embedded metadata is widely read for images, but not for video.
Video and audio
Video has dimensions a still image doesn't: camera movement (pan, tilt, zoom, drone), motion within the frame, and a scene that changes over the clip. Describe those alongside the subject and the concept.
Audio adds tempo (BPM), key, mood and genre for music; a podcast episode needs a summary and the platform's category (Apple Podcasts has its own fixed list). Either way, the metadata has to describe the whole file, start to finish.
When metadata gets a file rejected
Almost every reason comes down to one thing: the metadata doesn't match what's actually in the file.
- Keyword spamming — repeating a word's root or stacking unrelated terms.
- Keywords that don't match the frame.
- An editorial-style title (specific date, place, names) in a commercial file.
- Brand names, trademarks or the names of known people used as keywords without the rights.
How Meta VisBreeze handles it
Meta VisBreeze takes this work off your hands. Each generator already knows the rules of the platform you pick — its limits, keyword format and categories — and writes the title, description and keywords to fit them, keyword order included where the platform weights it. The literal-and-conceptual balance is kept, and trademarked names can be stripped automatically. You choose the output language.
Then you review and edit — across the whole batch or per file — and export: a CSV in the platform's format, or the metadata written straight into the file (EXIF/XMP for images, ID3/XMP for audio, XMP for video). Files are processed and deleted right after — nothing is stored.
FAQ
Do buyers see my keywords?
No. Keywords are only used by the search engine to match a query. The buyer sees your title and the image.
Should I fill every keyword slot?
No. Stop when you run out of terms that are genuinely accurate. Irrelevant fillers can hurt ranking and, on some platforms, count as spam.
Can I use the same keyword list on every platform?
The core list travels well; the rules around it don't. Keyword-order weight, singular vs plural, controlled vocabulary and category systems all differ. A preset per platform is what handles that.
Do platforms read the metadata inside my file?
Adobe Stock and Getty read embedded IPTC/XMP for images on upload; Shutterstock uses a CSV. Support for embedded video metadata is patchy — a CSV is the safe route there.
What about AI-generated work?
Metadata works the same way. But some platforms don't accept AI content at all, and some require you to disclose it — that's a platform rule, not a metadata one.