Adobe Stock gives you four metadata fields that decide whether a file gets found: title, keywords, category, and the language your account writes in. Adobe asks for titles under roughly 70 characters, allows up to 49 keywords per submission, and treats the first ten keywords as the ones with the greatest influence on search ranking. Everything else is technical compliance.
Most contributors learn these fields by trial and rejection. That is an expensive way to learn, because a file with weak metadata is not refused — it is accepted and then quietly never surfaces. This reference collects what Adobe's own contributor documentation specifies for each field, plus the technical limits that sit behind them.
The title field: describe what is visible, in under 70 characters
Adobe's guidance on titles and keywords asks for a brief, clear title that accurately describes the content, ideally kept under 70 characters. The title should focus on what is most visually important in the frame. Overly technical or gear-heavy language is discouraged, and anything involving intellectual property, artist names or identifiable real people is off limits.
The examples Adobe publishes are plain descriptive phrases rather than keyword strings. A photograph of two people embracing outdoors is titled as a couple hugging in the park. An illustration of a researcher at a microscope is titled as a woman in a laboratory with a face mask and gloves. Both read like something a person would actually say, which is the point: the title is the first thing a buyer reads, and it also feeds the search index.
Two practical rules follow. If a word is important enough to appear in the title, it belongs in the keyword field as well, since the two fields reinforce each other rather than dividing the work. And resist padding: a title stuffed to 70 characters with synonyms reads worse to a buyer and adds nothing the keyword field cannot carry.
The keyword field: 49 slots, and the first ten decide ranking
Adobe allows up to 49 keywords per content submission. The important detail is not the ceiling but the ordering rule: Adobe states that the most important and relevant keywords should sit in the first ten positions, because those carry the greatest influence on search ranking. Keywords 11 through 49 still help a file surface in long-tail searches, but they do not compete for the same weight.
Adobe groups useful keyword types into a handful of buckets, and running through them is a fast way to fill 30 accurate slots without inventing anything:
- Descriptive elements — white, fluffy, young animal, pup
- General and specific pairs — animal, mammal, carnivore
- Location, disambiguated — London, England or London, Ontario
- Concepts — solitude, childhood, milestone, cold
- Number of people — one person, three people, alone
- Setting and conditions — indoors, outdoors, day, night, sunny, cloudy
- Viewpoint — high-angle view, aerial view, portrait
- Model information — demographics and gender terms buyers actually filter on
Adobe also asks for local spelling where it applies, so color and colour are a choice rather than a pair. Each keyword should be used once; repeating a term in singular and plural form wastes a slot without adding reach.
The ordering rule is worth an example. For a photograph of an older woman flexing her arms on a beach, Adobe's own sample opening ten runs: woman, back, muscular, flexing, muscles, beach, Caucasian, senior adult, adult, one person. Notice what is missing — no mood words, no viewpoint, no weather. Those belong further down. The first ten answer the literal question of what a buyer typed.
If you want the broader logic behind ordering and the difference between literal and conceptual terms, that is covered separately in our guide to structuring tagged keywords. The Adobe-specific rule is simply that position matters and the top of the list is scarce real estate.
The category field: one choice, pre-suggested by Adobe Sensei
Every submission gets one category. When you upload to the Contributor Portal, Adobe Sensei suggests a category automatically based on the content. Adobe describes this as a good starting point and explicitly recommends reviewing it rather than accepting it blindly.
The list Adobe publishes runs to twenty-one options: Animals; Buildings and architecture; Business; Drinks; The environment; States of mind; Food; Graphic resources; Hobbies and leisure; Industry; Landscape; Lifestyle; People; Plants and flowers; Culture and religion; Science; Social issues; Sports; Technology; Transport; and Travel.
Several of these overlap in practice, which is where the automatic suggestion tends to drift. A photograph of a family cooking dinner could be Food, Lifestyle or People. Adobe's advice is to weigh not only the primary subject but the context, mood and intent — and if nothing fits perfectly, pick the closest match to the core message rather than the most populated category.
Language: the setting most contributors set once and never check
Titles and keywords must match the language selected in your Contributor account, chosen in the dropdown labelled “I'm writing title and keywords in”. This is a single-language field, not a multilingual one. Mixing English keywords into a submission set to another language, or the reverse, undercuts search matching for the whole file.
It matters more than it sounds. Contributors who write in a second language often draft keywords in their own language, translate a few and leave the rest, which produces a mixed set that matches neither audience cleanly. Pick one language, set it in the portal, and keep every field in it.
Releases and the generative AI checkbox
Two attachments sit alongside the text fields and are treated as part of the submission rather than as extras. Recognisable people require an individual model release. Identifiable locations, artwork or objects require a property release. Missing either one is a refusal reason on its own, no matter how good the keywords are.
The generative AI checkbox is a separate declaration. Adobe asks contributors to tick “Created using generative AI tools” when generative tools changed, augmented or added a primary subject in the file. Adobe also splits the file type: photo-realistic AI images that could plausibly have been captured with a camera are submitted as Photos, while everything else generated goes in as Illustrations.
Agencies differ sharply on how AI-generated work must be declared, and getting it wrong on one platform can affect an account across a whole batch. We cover the cross-agency picture in metadata for generative AI images.
The technical limits behind the metadata
Metadata does not rescue a file that fails the technical requirements. For photos, Adobe specifies a resolution range of 4MP to 100MP, a maximum file size of 45MB, JPEG format with an sRGB colour profile, and content free of watermarks, timestamps or branding. Sharp focus and clean lighting are stated requirements, not preferences.
Larger files inside that ceiling give buyers more flexibility, so resizing down to the minimum is rarely the right call. Resize to preserve detail rather than to shrink the upload.
One thing worth knowing if you upload by FTP rather than through the browser: the title and keyword values are read from the file's embedded IPTC and XMP fields, so what you write in your editor is what Adobe receives. Our primer on IPTC, XMP and embedded keywords explains which field maps to which.
Metadata mistakes that get content refused
Adobe names a short list of habits that lead to refusal, and they are all metadata-side rather than image-side:
- Irrelevant keywords. Adding terms because they are popular rather than because they describe the file is grounds for refusal, not just poor ranking.
- Trademarks and brand names. Keep them out of the title and the keyword field, even when the object is visible in frame.
- Personal information. Names of the people in the photograph do not belong in metadata.
- Duplicate terms. Each keyword should appear once; repetition burns a slot and signals padding.
The image-side reasons — focus, noise, over-editing, similarity — are a separate list, and we walk through them in Adobe Stock rejection reasons.
Holding these fields steady across a whole batch
Reading the rules is easy. Applying them identically to four hundred files from one shoot is where the work actually is, and it is where most portfolios drift: the first fifty files get careful ordering and the last three hundred get a copied keyword block.
A batch stays consistent when three things are true. The title stays under Adobe's length guidance without being padded. The first ten keywords are re-ordered per file rather than inherited from a template, because a template cannot know which subject dominates a given frame. And the category is checked against the frame instead of accepted from the automatic suggestion.
This is the part Rastock AI was built for. It reads each file, writes a title and an ordered keyword set inside the agency's own limits, and delivers to Adobe Stock and 10+ other agencies over FTP or SFTP, with per-file status you can check afterwards. Banned and mandatory keyword rules are enforced before delivery rather than discovered at moderation, and a batch can be rolled back if you do not like the result. You can see how the metadata and delivery pipeline fits together, or check the plans and free trial — 14 days, no card required. There is no revenue share and no claim on your files; IPTC and CSV export stay open.
The field reference in one pass
Title: brief, clear, ideally under 70 characters, describing what is visually dominant, with no IP, artist names or real people. Keywords: up to 49, each used once, most important ten first, local spelling, no trademarks or personal data. Category: one of twenty-one, auto-suggested and worth reviewing. Language: single, matching the account setting. Releases: model for recognisable people, property for identifiable places and objects. Generative AI: declared with the checkbox, and filed as Photos or Illustrations depending on realism.
Adobe updates this documentation regularly — the titles and keywords page was last revised in August 2026 — so it is worth re-reading the source pages once a quarter rather than trusting a copy of the rules from a year ago.
Frequently asked questions
How many keywords can you add to an Adobe Stock submission?
Adobe allows up to 49 keywords per content submission. That is a ceiling rather than a target. Adobe asks for relevant nouns, adjectives and concepts specific to the content, and warns that irrelevant keywords can cause a file to be refused. Most well-described photographs are covered by roughly 25 to 40 accurate terms, with the strongest ones placed in the first ten positions.
How long should an Adobe Stock title be?
Adobe asks contributors to keep titles brief and clear, ideally under 70 characters. The title should describe what is visually most important in the frame, in natural language, without camera or gear jargon. Adobe's published examples are short descriptive phrases rather than comma-separated keyword strings, and they exclude trademarks, artist names and identifiable real people.
Does keyword order matter on Adobe Stock?
Yes. Adobe states that the first ten keywords have the greatest influence on search ranking, so the most important and relevant terms belong at the top of the list. Terms in positions 11 to 49 still help a file appear in narrower searches, but they do not carry the same weight. Reordering per file, rather than reusing one template, is what makes this rule pay off.
Do I have to accept the category Adobe suggests?
No. Adobe Sensei suggests a category automatically when you upload, and Adobe describes that suggestion as a starting point to review rather than a final answer. If another of the twenty-one categories describes the content better, select it from the dropdown. Adobe advises weighing context, mood and intent alongside the primary subject, and choosing the closest fit when nothing is perfect.
When do I need to tick the generative AI checkbox?
Adobe asks contributors to select the "Created using generative AI tools" checkbox when generative tools changed, augmented or added a primary subject in the file. Adobe also separates the file type: photo-realistic generative images that could have been captured with a camera are submitted as Photos, while other generative images are submitted as Illustrations.
What are the technical limits for photos on Adobe Stock?
Adobe specifies a resolution range of 4MP to 100MP, a maximum file size of 45MB, JPEG format with an sRGB colour profile, and files free of watermarks, timestamps or branding. Sharp focus and good lighting are listed as requirements. Larger files within that ceiling give buyers more flexibility, so resizing down to the minimum is usually a mistake.