Keywords rarely vanish because a platform rejected them. They vanish because a cap was hit, a comma split a value, or a field got overwritten somewhere between your editor and the agency. Adobe Stock allows up to 49 keywords per submission, Shutterstock states a range of 7 to 50, Canva publishes a minimum of five and no maximum at all, and none of them send you a list of what was dropped.
The limits, and where they are actually published
Three things are worth noticing before the table. The caps differ by one on the two largest platforms, which is enough to shorten a list you thought was portable. Only some platforms publish a minimum. And several publish nothing about what happens when you exceed the ceiling, which means the honest answer for those is that the behaviour is undocumented rather than safe.
Published keyword and title limits by platform
Adobe Stock allows up to 49 keywords and advises titles under 70 characters, Shutterstock states 7 to 50 keywords with a 200-character title, Envato caps tags at 50, and Canva requires at least five keywords with no published maximum.
The Adobe row carries the detail that matters most for truncation. Adobe says keyword order is essential and that the most important terms belong in the first 10 positions because those have the greatest influence on search ranking. It also says to use each keyword once and to submit content in one language only, with titles and keywords matching the language set on your contributor account. The ranking consequences of that ordering are unpacked in Adobe Stock keywording rules.
Five ways a keyword list gets shortened
1. You went over the cap
The simplest cause, and the one people assume least often, because a 55-keyword set looks generous rather than broken. If you keyword against the most permissive destination and deliver the same set everywhere, the stricter platform silently takes what fits. Since Adobe and Shutterstock differ by a single keyword, a set built to 50 fits one and not the other.
2. A comma inside a keyword value
This is the one that produces the strangest symptoms. In a CSV, the keyword column is itself usually comma-delimited, so a single keyword containing a comma either splits into two values or, if quoting is broken, shifts every column after it by one. The result is a row where the keywords look short and the category looks wrong, which sends people hunting for a category bug that does not exist. Canva makes the convention explicit by stating that keywords are separated by a comma and a space; the corollary is that a comma inside a term is not a term.
3. The field you wrote is not the field they read
Embedded keywords live in more than one place. A writer that fills the legacy IPTC keyword field but not the XMP equivalent, or the reverse, produces a file that looks fully tagged in one application and half tagged in another. If the agency reads the field your tool left empty, the list does not arrive short, it arrives missing. Which field holds what is covered in photo tag basics for IPTC, XMP and embedded keywords.
4. Something downstream rewrote the file
An export preset that strips metadata, a resize step that writes a fresh file rather than copying the original's fields, a watermarking pass, a cloud sync that re-encodes: any of these can replace a complete keyword set with a partial one after you have already checked it. The giveaway is that the file on your disk is correct and the file you delivered is not. The full list of culprits is in metadata disappears after export.
5. Order decided what survived
When a platform takes the first N keywords and discards the rest, the order you shipped in becomes the selection criterion. This is where alphabetical keyword lists do real damage: an alphabetised set that gets cut at the cap keeps everything beginning with A to M and throws away the specific terms that happened to start with S. Adobe's own guidance already asks for importance order, so an alphabetised export is working against the platform twice.
What a truncation-prone set looks like
Most sets that arrive short were fragile before they left. The comparison below takes one straight export and shows which terms were going to cause trouble, which were spending a slot for nothing, and what the same file looks like once both problems are removed.
A fragile keyword set and a portable one
One exported keyword set carrying a comma inside a value, a duplicate, a second-language term and two near-synonyms, rebuilt so it survives a CSV round trip and fits Adobe's 49-keyword cap and under-70-character title guidance.
The rebuilt title is 60 characters, inside Adobe's suggestion of under 70 and well inside Shutterstock's 200-character ceiling, so the same string travels to both without editing. The rebuilt keywords also follow the coverage pattern Adobe asks for: descriptive elements, a general and specific pair, location, number of people, indoor or outdoor, and time of day.
When to use which safeguard
- Keywording for one platform: build to that platform's cap and ordering rule and stop. Nothing else is worth the effort.
- Keywording for several platforms by hand: build one set to the strictest published cap among your destinations, in importance order. A set that fits 49 fits every platform in the table above, and importance order means a cut on an unknown-limit platform removes your weakest terms rather than a random slice.
- Delivering by CSV: test one row end to end before the batch. Write a deliberate multi-word keyword, send it, read it back from the platform, and confirm it arrived as one value. This single test catches the comma problem permanently.
- Delivering embedded metadata: verify the keyword field on the exported file rather than the working file, and verify it with a different tool than the one that wrote it. A writer and a reader that share assumptions will agree with each other and both be wrong.
Diagnosing a batch you have already sent
Work backwards along the chain, because each step narrows the problem to one of the five causes above. Read the keyword field out of a delivered file and count it: if the count matches what you intended, the loss happened at the platform and you are looking at a cap or an ordering cut. If the count is already short on your own disk, the loss happened before delivery and you are looking at an export step or a field mismatch. If the count is right but the values look wrong, with one keyword appearing where a category should be, you have the comma problem.
One thing worth doing once rather than per batch: write down the cap and the title limit for every destination you actually use, including the ones that publish nothing, and keep the list next to your keywording step rather than in a browser tab. The same reasoning applied to titles and descriptions is set out in standardising titles and descriptions across agencies, and the two lists are usually maintained together.
For the two platforms most people ship to first, the field-level references are worth reading in full rather than from a summary table: the Shutterstock metadata requirements breakdown covers the description and keyword fields together, which is where the 7-keyword floor tends to surprise people migrating from a platform that has none.
Making the limit the tool's problem
Every cause above is a bookkeeping failure rather than a judgement failure, which is exactly the kind of thing worth handing to software. Rastock.ai holds each platform's field and keyword rules, so one keyword set is trimmed and ordered per destination instead of being sent identically everywhere and quietly cut, and it delivers over FTP and SFTP to more than ten agencies with no revenue share and no claim on your files. The feature overview covers the per-platform checks, including banned and mandatory keyword handling, and pricing includes a 14-day free trial with no card required. IPTC and CSV export stay unrestricted either way, so a test batch tells you what you need to know without committing anything.
Frequently asked questions
How many keywords does Adobe Stock allow?
Up to 49 per content submission, according to Adobe's own contributor help on titles and keywords. The same page says keyword order is essential and that the first 10 positions have the greatest influence on search ranking. It does not state what happens to keywords beyond 49, so treat the cap as a limit to respect rather than a threshold to test.
What is Shutterstock's keyword limit?
Shutterstock's stated range is 7 to 50 keywords, with a descriptive title capped at 200 characters. The floor matters as much as the ceiling: a set of five keywords that satisfies Canva will not clear Shutterstock's minimum, which is a common surprise for contributors adding Shutterstock to an existing pipeline.
Why did only some of my keywords upload?
There are five usual causes: the set exceeded the platform's cap, a keyword contained a comma and split in the CSV, the keywords were written to one embedded metadata field and read from another, an export or resize step rewrote the file after you checked it, or the platform took the first N keywords and your ordering decided which survived.
Can a keyword contain a comma?
Not safely. The comma is the delimiter in both CSV columns and keyword lists, so a term containing one either splits into two values or breaks the row's quoting and shifts every following column. Canva states outright that keywords are separated by a comma and a space. Use spaces or hyphens inside multi-word terms and keep commas for separation only.
Should I keyword to the highest limit or the lowest?
Build one set to the strictest published cap among the platforms you actually deliver to, in importance order. A set that fits 49 fits every major platform, and importance order means that if an undocumented platform trims the list, it removes your weakest terms rather than an arbitrary slice. Keywording to the most generous limit guarantees silent loss elsewhere.
How do I check what actually arrived?
Read the keyword field back out of a delivered file and count it, then compare against what you intended. A matching count means the loss happened at the platform. A short count on your own disk means an export step or a field mismatch. A correct count with values in the wrong columns means a comma inside a keyword. Use a different tool to read than the one that wrote.