How to Generate Metadata for Adobe Stock (2026 Guide)

To generate metadata for Adobe Stock, give every file its own short title (under 70 characters) that names the main subject and what makes this frame different. Then add up to 49 keywords in order of importance, because Adobe says the first 10 positions have the greatest influence on search. Keyword the shoot as a series, not file by file: the facts every frame shares (place, people, season) stay the same, while the title and the first 10 keywords describe what only this frame shows. If similar files carry identical titles and keywords, Adobe treats that as a sign of similar content, the files compete for the same searches, and they risk refusal.

One thing up front: good metadata makes good content findable. It does not make weak content sell, and we do not promise more sales.

What Adobe Stock asks for

These are Adobe's own rules, checked on 6 November 2026, next to how we apply them.

Field	Adobe's rule	How we apply it
Title	Brief and clear, ideally under 70 characters. Name the most important visual element. No gear-heavy terms, no IP, artist names or real people. The CSV template caps titles at 70 characters, plain text, no commas.	Main subject plus what is unique in this frame, in plain words a buyer would type.
Number of keywords	Up to 49 per file (Adobe's keyword guide; the CSV spec lists 50). Use each keyword once.	49 is a ceiling, not a target. We stop when the relevant words run out.
Keyword order	Most important first. The first 10 positions have the greatest influence on search ranking.	First 10 = what this frame shows that its neighbours don't, plus the core subject.
Relevance	Only relevant keywords, or the content can be refused.	Every keyword must be visible or clearly implied in the frame.
Brands and personal info	Keep trademarks, brand names and personal information out.	We also strip camera names, file-name text, client and model names.
Language	One language, matching your contributor account. Local spelling where it applies.	English.
Similar content	One of the top refusal reasons. Different files with identical titles and keywords are named as a sign of similar content.	A unique title per file and a distinct top 10 across the shoot.
Category	Optional CSV column, entered as a category number.	One category per file.
Releases	Optional CSV column with the exact release file name.	Filled from your release files.
Generative AI	Tick "Created using generative AI tools" and, where it applies, "People and Property are fictional". No names of artists, real people or fictional characters.	Never written into titles or keywords.
CSV file	UTF-8 .csv with the exact template headers, up to 5,000 rows and 1 MB, file names up to 30 characters.	Our export follows the template.

Step by step: Adobe Stock metadata for a whole shoot

Write the shoot brief first. Who is in it (how many people, age groups, roles), what happens, where and when. These facts are the same for every frame, so you decide them once and never contradict them later.
Group the files by scene and cast. Frames from the same scene and the same people belong together. In Stock ImageIQ, Sort for you proposes these folders; nothing moves until you confirm.
Find the angle of each frame. Ask what changes from the frame next to it: the action, the number of people, the framing (wide, medium, close-up), the focus object, the emotion, the camera angle. For video, also the camera motion.
Write the title from that angle. Subject plus what is unique, under 70 characters, no commas if you upload by CSV.
Order the keywords. The first 10 carry what is specific to this frame plus the core subject. After that come the shared shoot facts (place, season, people), then the concepts and moods a buyer would search for.
Check before export. No brands, camera names, file names, client or model names. No keyword used twice. No two files in the shoot with the same title. Ages, number of people and location consistent across the series.
Add category, releases and the AI label where they apply, then export the Adobe Stock CSV.
Cull before you upload. If two frames would need the same title and the same keywords, they are probably the same picture for a buyer. Upload the stronger one.

Similar images compete with each other

Imagine 80 frames from one morning on a balcony, all uploaded with the same title and the same 49 keywords. Every one of them answers exactly the same search. A buyer who types "woman relaxing on balcony" gets a row of near-identical results, and a buyer who types "woman drinking coffee with a city view" never finds the frame at the railing that would have fit. The files cannibalize each other: they fight for one search instead of covering many.

Three frames from the same balcony shoot. One title on all three answers one search; a title per frame answers three. Searches are examples. Photos: azmanl

Adobe sees the same problem from its side. Its distinct content guidelines name identical titles and keywords on different files as a sign that the files serve the same customer need, and say that too many similar files can reduce how visible your work is. So identical metadata does not only waste the series. It can get the series refused.

The fix is to keep what the frames share and change what they don't. The same three frames with their titles and first keywords (simplified):

Frame	Title	First keywords
With coffee	Relaxed young woman holding a coffee cup in a balcony chair	coffee cup, relaxing, wicker chair, morning coffee, young woman
Reading	Young woman reading a book in a chair on a sunny balcony	reading, book, concentration, leisure, young woman
At the railing	Woman drinking coffee at the balcony railing with a city view	drinking coffee, railing, city view, standing, apartment balcony

All three still carry the shared facts further down the list: balcony, apartment, city, summer, morning, one person, lifestyle. The top of the list is where they differ.

What our data shows. We measured 19 real shoots (1,577 files) processed with Stock ImageIQ. In 77% of the files, the set of the first 10 keywords (order ignored) appeared on no other file from the same shoot. Of all the keywords two files from the same shoot use between them, about two fifths appear on both: the shared facts stay, the rest changes. This counts keywords, not buyer searches or sales. More on this, with a full series: Same Keywords on Every Photo? Your Series Is Competing With Itself.

What we learned working with top contributors

The rules behind Stock ImageIQ were not written from agency help pages alone. They were built with top contributors, from their confirmations and suggestions, and checked against measured results and sales. These are the ones that matter most for Adobe Stock:

Facts before style. A wrong age group, species or city does more damage than a missing adjective. Buyers filter on facts, and a wrong fact puts the file in front of the wrong buyer.
Keep the series consistent. If the brief says two people in their thirties, every frame says the same. Contradictions inside one shoot look careless to a reviewer and confuse search.
Describe situations, not object lists. Buyers search for what is happening: "father teaching daughter to ride a bike", not "bicycle, girl, man". The title should read like that situation.
Concepts only when you can see them. "Teamwork" belongs on a frame where people visibly work together, not on every office photo. Concepts you can't see are irrelevant keywords, and Adobe refuses for those.
Locations from specific to broad. City, region, country, never two places that contradict each other.
Video has its own vocabulary. Camera motion (aerial, tracking shot, slow motion), what happens in the clip, and when the action changes, the action that lasts longest. More in Does Your Keywording Tool Watch the Whole Clip?
Nothing from outside the frame. Camera brands from file names, the client's name, the agency's name: none of it describes the picture, and some of it breaks the trademark rule.
Don't chase 49. Padding the list with weak words does not add searches. It adds refusal risk and dilutes the words that matter.

What metadata cannot do

We don't promise more sales, and no metadata tool honestly can. Sales start with content buyers need: a clear subject, a commercial use, technical quality, valid releases and a topic that isn't already covered by thousands of similar files. That is the precondition. Metadata comes after it.

What metadata does decide is whether a good file is found for the searches it deserves, and whether a series is accepted instead of refused as similar content. If an image does not sell with accurate, distinct metadata, the answer is usually in the image or the topic, not in the keywords.

What we can measure, we publish: whether each file is described correctly and differently from its neighbours. Our next measurement counts every manual correction an experienced contributor makes across 2,309 files from 38 shoots.

FAQ

How many keywords can I add on Adobe Stock? Up to 49 per file, according to Adobe's keyword guide. Use only as many as are truly relevant. The first 10 matter most.

Does keyword order matter on Adobe Stock? Yes. Adobe says the first 10 positions have the greatest influence on search ranking, so put the most specific and important keywords first.

How long should an Adobe Stock title be? Ideally under 70 characters. If you upload by CSV, the template caps titles at 70 characters and does not allow commas.

Can I use the same keywords for similar photos? Not if you want them all to work. Adobe names identical titles and keywords on different files as a sign of similar content, which is one of the top refusal reasons. Keep the shared facts and change the title and the first 10 keywords.

Can I use brand names or camera names in keywords? No. Adobe asks you to keep trademarks, brand names and personal information out of titles and keywords.

Do I have to label AI-generated content? Yes. Tick "Created using generative AI tools" and, when it applies, "People and Property are fictional". Never use names of artists, real people or fictional characters in the metadata.

Will better metadata increase my sales? We don't promise that. Metadata helps good content get found and accepted. Whether it sells depends first on the content itself.

Can Stock ImageIQ generate Adobe Stock metadata for me? Yes. It reads the whole shoot, writes a title and ordered keywords for each file, and exports a ready Adobe Stock CSV. With the Stock ImageIQ desktop app for macOS, it also writes the metadata straight into your original photos, so it travels with every file:

Title → XMP dc:title and photoshop:Headline, plus IPTC-IIM Object Name and Headline
Description → XMP dc:description, IPTC-IIM Caption-Abstract and EXIF ImageDescription
Keywords, in their order → XMP dc:subject and IPTC-IIM Keywords

JPEG and TIFF files get both XMP and IPTC-IIM. PNG and WebP get XMP only, because those formats have no IPTC-IIM block. XMP always carries the full text; the IPTC-IIM copies respect that standard's byte limits. Every other field in the file stays untouched. The Mac app uploads a copy of each photo at most 1280 px on its long side for analysis (small photos and GIF or BMP files are uploaded as they are), then writes the title, description and keywords back into the originals on your disk. For videos, the app saves a metadata CSV next to the clips.

A photo uses 1 credit and a video uses 3 credits. On the largest pack a credit costs under €0.10 (excl. VAT). See pricing or download the desktop app.

Sources: Tips for effective titles and keywords · Distinct content submission guidelines · CSV requirements for Adobe Stock · Generative AI content guidelines