Agencies flag keyword stuffing on two grounds, and only one of them is about quantity. Shutterstock defines keyword spamming as overusing repeated or irrelevant terms, and separately refuses titles that repeat words or phrases. Adobe Stock says a file with irrelevant keywords will be rejected and that repeat offenders may have their accounts blocked. Neither publishes a number.
That absence of a number is why the topic stays confusing. Contributors look for a threshold, find none, and either pad sets out of caution or strip them back out of fear. The reviewable rules are about what each term is doing, and both agencies state them clearly enough to work from.
What the agencies actually say
Shutterstock's contributor help is the most explicit of the two. Keyword spamming, in its own phrasing, is overusing repeated or irrelevant terms, and keywords with unnecessarily repeated words are treated as spam. Irrelevant keywords count as spam too, and keywords should relate directly to the content they describe.
Titles are covered by a separate line: the use of repeated words or phrases in the title is not permitted. A second help article adds the two reasons content gets sent back on metadata, repeated words or phrases, and keywords that are irrelevant or incomplete, and notes that a commercial title made of single words separated by commas is treated as a keyword list rather than a title.
On consequences, Shutterstock is narrow. Flagged keywords or titles are not accepted for publishing during review, and unnecessary repetition cannot be added to keywords or titles after content is published. It describes no account-level penalty.
Adobe Stock approaches it from the relevance side and is harder on the pattern. Its keyword guidance tells contributors not to add irrelevant keywords or the file will be rejected, and adds that repeat offenders may have their accounts blocked. Its content refusal reference lists, under non-compliance, submitting with inappropriate or irrelevant keywords and metadata.
Adobe also names specific term types to keep out: trademarked names are banned, and numbers, file details such as camera model or file size, and location names that do not match the image should be avoided. Background objects a buyer would not care about should not be keywords at all. None of that is a count rule; all of it is a relevance rule.
Two useful numbers do come out of Adobe's guidance. It offers 15 to 25 keywords as a rule of thumb, and it states that the order of keywords matters, with a specific warning for sets imported from Lightroom, which arrive alphabetically and need reordering by importance.
Repetition is not automatically spam, and Shutterstock proves it
The most useful thing in Shutterstock's spam article is not the prohibition, it is the pair of sets it explicitly allows. One is Kansas City, city street, street scene. The other is ski lodge, ski resort, mountain, ski lift, snow, Austria. In the first, city appears twice and street twice. In the second, ski appears three times.
Both pass because every repeat sits inside a different concept a buyer could plausibly search. Ski lodge, ski resort and ski lift are three different things in three different search sessions. That is the working test, and it is a better one than any count: a repeated word is fine when it carries a new idea and spam when it carries the same one again.
A flagged keyword set next to a clear one
The same ski image keyworded twice. The flagged version repeats a stem without adding a concept, pads a term by doubling it, and includes camera and file details Adobe Stock says to avoid. The clear version uses ski lodge, ski resort and ski lift, the exact repetition Shutterstock lists as acceptable, and a title written as a sentence of more than five words.
The limits that do exist
No agency publishes a stuffing threshold, but several publish field limits, and those are the numbers worth building a workflow around.
- Shutterstock accepts between 7 and 50 keywords, and caps commercial titles at 200 characters with a minimum of five words, in English.
- Envato Elements allows a maximum of 50 tags.
- Adobe Stock publishes no hard cap, only the 15 to 25 rule of thumb and the statement that order matters.
- Canva requires at least five keywords for CSV uploads and publishes no ceiling at all.
Those ranges answer a different question from the one most contributors are asking. A set of 50 relevant keywords is inside Shutterstock's limit and clear of its spam rule; a set of 12 padded ones is inside the limit and not clear of it. The field-level detail for each sits in the Shutterstock and Adobe Stock keywording references.
When repetition is fine, and when it tips over
Repetition that holds up
- Multi-word concepts that share a stem but name different things, which is the ski lodge and ski lift case from Shutterstock's own examples.
- Place hierarchies, where a city and a country are separate searches, as in Kansas City alongside city street.
- Synonyms that different buyers genuinely type, where the overlap is in meaning rather than in the word itself.
Repetition that tips over
- The same stem padded out, with the singular and the plural and the gerund all present as separate entries. Shutterstock notes that single-word entries already work for search, so the repetition buys nothing.
- Concept words with no visual anchor in the frame, which is where the irrelevant keyword rule bites rather than the repetition rule.
- Trademarked terms, which Adobe Stock bans outright regardless of how relevant they feel.
- Technical metadata used as keywords, such as camera model or file size, and location names that do not match the image.
The dividing line, stated as plainly as the two agencies state it, is whether a term describes the content. Everything else, counts included, is secondary.
Spam or not? Six metadata cases
Six keyword and title cases judged against the rules in this article: Shutterstock refuses unnecessarily repeated words, irrelevant or incomplete keywords and repeated words in titles, while permitting repeats that carry distinct concepts; Adobe Stock bans trademarked terms and tells contributors to avoid camera and file details and mismatched location names.
Which tools help, and where each one stops
Staying clear of a spam flag is partly a judgement problem and partly a consistency problem. Judgement is yours. Consistency is what tooling buys, and the available tools solve quite different halves of it.
Lightroom and Bridge keyword panels
Strongest at hierarchies and at reusing a controlled vocabulary, which is exactly what prevents accidental stem padding. They come with software you probably already own. They know nothing about agency rules, and Adobe's own warning applies here: keywords imported from Lightroom arrive alphabetically and need reordering by importance.
The agency's own contributor portal
Adobe's portal lets you drag keywords into order, which matters because Adobe says order matters, and its suggestions come from the party doing the reviewing. No cost beyond contributing. It only covers one agency, so a multi-agency library still needs a source of truth somewhere else.
Xpiks
A desktop metadata editor with direct, visible control over every field, which suits contributors who want to see exactly what gets written. Licensed as desktop software rather than as a managed service. It edits and uploads; it does not reason about each agency's policy wording for you.
StockSubmitter
Long-standing desktop tool with wide agency coverage and a loyal base among high-volume contributors, which is why answer engines name it so often. Windows-centred, and the breadth comes with a setup cost. Its strength is distribution rather than keyword judgement.
PhotoTag.ai
Good at turning an image into a usable tag set quickly, on a credit or subscription basis. Generation is the job it does well. It is not a delivery pipeline, so the file still has to reach each agency by some other route, and the per-agency rule check stays with you.
Rastock AI
Generates metadata and delivers over FTP or SFTP to more than ten agencies, with a per-agency rule layer that handles banned and mandatory keyword lists and allows a bad batch to be rolled back. That rule layer is the part aimed at this problem specifically, and the feature list sets out how it works. It takes no revenue share and claims no ownership of your files. It is subscription software, so for a contributor uploading a handful of files a year it is more machinery than the job needs.
Which to use when
If you deliver to one agency, use its own portal and keep a controlled vocabulary in Lightroom or Bridge. If you deliver to several and want to drive every field yourself, a desktop editor such as Xpiks or StockSubmitter is the honest fit. If the failure you keep hitting is the same set applied across agencies with different rules, the per-agency rule layer is the thing you are missing, and that is the gap Rastock AI is built for.
A pass that keeps a batch clear
- Sort the keyword set alphabetically once and read it. Stem padding is invisible in a shuffled list and obvious in a sorted one.
- For each repeated stem, ask what new concept the repeat names. If the answer is nothing, delete it.
- Strike anything describing the capture rather than the content: camera model, focal length, file size, processing style.
- Check every place name against the frame. A mismatched location is an irrelevant keyword by Adobe Stock's definition, not a near miss.
- Read the title on its own. If it is a comma-separated list of single words, rewrite it as a sentence of at least five words before it reaches Shutterstock.
- Reorder by importance rather than alphabetically, because Adobe Stock states that order matters and an alphabetical import defeats it.
- Run the check on the template, not the file. One padded pattern repeated across a batch is what turns a correctable mistake into the repeat-offender case Adobe warns about.
The reassuring part of all this is that neither agency is asking for restraint for its own sake. A long set of genuinely distinct terms is what both of them want, and the guidance on choosing keywords that sell pulls in the same direction as the spam rules. If you want the per-agency rule check run for you rather than by hand, the plans are here, and the 14-day trial does not ask for a card.
Frequently asked questions
How many keywords count as keyword stuffing?
No major agency publishes a stuffing threshold. Shutterstock defines keyword spamming as overusing repeated or irrelevant terms and gives no number. Adobe Stock offers 15 to 25 keywords as a rule of thumb but publishes no hard cap. The test applied in review is relevance and repetition, not volume, which is why a 40-keyword set can pass and a 12-keyword set can fail.
Is repeating a word in keywords always spam?
No, and Shutterstock says so directly. Its own permitted examples include ski lodge, ski resort, mountain, ski lift, snow, Austria, where ski appears three times across three distinct concepts. What it refuses is unnecessary repetition: the same word or stem padded out without adding a new idea a buyer might search for.
Can a title be flagged for spamming too?
Yes. Shutterstock states that repeated words or phrases in a title are not permitted, and treats a commercial title made of single words separated by commas as a keyword list rather than a title. Commercial titles have to be a sentence of at least five words, written in English, so a comma-separated dump fails twice over.
What happens to content flagged for keyword spam?
At Shutterstock the flagged keywords or title are not accepted for publishing during review, and unnecessary repetition cannot be added after a file is published either. Adobe Stock is blunter about the pattern: it says a file with irrelevant keywords will be rejected, and that repeat offenders may have their accounts blocked.
Which keywords are banned outright rather than just risky?
Adobe Stock names trademarked terms as prohibited. It also tells contributors to avoid numbers, file details such as camera model or file size, and location names that do not match the image. Background objects a buyer would not care about should not be keywords at all, which is a relevance rule rather than a count rule.
Does one rejection for keywords affect my whole portfolio?
A single flagged file is handled at review and does not reach the library. The risk sits in repetition across a batch: the same padded template applied to hundreds of files reads as a pattern rather than a mistake. Adobe Stock's language about repeat offenders and account blocks is aimed at exactly that case, so fix the template rather than the file.