Vecteezy's search engine does not treat your keywords equally. The first five carry extra weight, the system expects singular forms and no synonyms, and once a file has been submitted the keywords are locked. Getting the order right on the first pass matters more here than it does at most agencies.
That combination of weighted positions, a strict vocabulary and no second chances makes Vecteezy keywording a different job from filling a 49-slot Adobe field. The rules are published, but they are spread across Vecteezy's help centre and its contributor blog, and on one point the two sources disagree. This is what each rule says, where they conflict, and how to keyword a batch so it survives review.
The two keyword caps Vecteezy publishes, and which one to follow
Vecteezy states a keyword range in two places and the numbers are not the same. The keyword guide in the Eezy contributor help centre says the minimum is 5 and the maximum is 50, and encourages 20 to 30 keywords per resource. The contributor checklist on the Vecteezy blog says you can enter a maximum of 30 keywords and encourages at least 10.
Both pages are Vecteezy's own, so neither is a rumour. The practical resolution is to work inside the band where they agree, which is 20 to 30 keywords. That satisfies the help centre's recommendation, stays under the stricter of the two caps, and leaves you no exposure if the Content Manager form enforces 30 on the day you upload. Treat 30 as your working ceiling and 50 as a number you never need.
This also means a keyword set built for Adobe Stock, which accepts up to 49 keywords, does not transfer to Vecteezy unedited. If you deliver the same metadata everywhere, the Vecteezy copy needs trimming, and the trim should come off the bottom of a relevance-ordered list rather than at random. Mapping each agency's limits once, instead of per batch, is the only version of this that scales.
Why the first five keywords decide the file
Vecteezy's keyword guide is explicit that the most relevant keywords should be listed first, and that the first five are the most important. The contributor blog repeats it in plainer language: search gives extra weight to the first five, so the terms people are most likely to search for belong at the front.
That turns keyword entry into a ranking exercise rather than a collection exercise. Two files with identical keyword sets in different orders are not equivalent on Vecteezy. The question for position one is not whether a word is true about the image, but whether it is the word someone looking for that image would type first.
For a still life of a ceramic mug on a linen cloth, mug and ceramic earn the front. Linen, texture and minimal are real and worth including, but they describe the setting and the styling rather than the subject, so they belong further down. The usual failure is the reverse: a mood word in position one because it sounds more marketable than the object.
Order your first five for Vecteezy
Vecteezy weights the first five keywords most heavily, so the terms a buyer would actually type belong in the top slots. Subject and material first, setting and mood below.
Singular, one form, no synonyms
Three of Vecteezy's rules collapse into a single habit. Submit only one form of each keyword. Submit the singular form. Do not add synonyms or spelling variations. The guide's own example of what not to do is kitten, kitty, kitties, cat, cats, baby cat, which is six slots spent on one idea.
Vecteezy says its system already covers synonyms and variations, which is why the padding earns nothing. On an agency where your working list is around 30 slots and the first five are weighted, spending six of them on inflections is not a small inefficiency. It pushes genuinely different concepts off the file altogether.
Phrases are allowed and sometimes better, with one constraint: keep common phrases together, but do not split names or invent uncommon combinations. Rubber chicken is correct and rubber, chicken is not. Justin Bieber stays as one keyword, while Bieber, Justin is wrong. The test is whether a person would type the phrase, not whether the words happen to appear in the frame.
Keywords must be in English, and diacritics are not processed. The guide names the accented e in cafe specifically as a character the system cannot handle. If your working language is not English, spelling matters more than vocabulary here: a misspelled keyword simply does not match what buyers type, and it quietly costs you a weighted slot.
The words Vecteezy does not want in the box
Some terms are banned not because they are inaccurate but because they are useless at scale. Content-type words are the first group: the guide says not to include the content type of the files you are submitting, such as photo, vector or video. Every file on Vecteezy is one of those things, so the keyword separates nothing and wastes a slot.
The second group is AI terminology. The guide says not to include AI or any version of AI Generated in keywords. Provenance is handled as a submission property rather than a search term, so putting it in the keyword box is both unhelpful and a rule violation. If you publish generative work, the disclosure belongs in the field the agency provides for it.
The third is keyword spam, which Vecteezy defines plainly as the overuse of repeated words or phrases. The contributor blog adds the consequence: reviewers may remove irrelevant keywords, and files that keep appearing in searches without being downloaded signal weak relevance, which can push rankings down. Padding a list with popular unrelated terms is therefore worse than leaving slots empty.
A Vecteezy keyword set before and after the rules
The same mug photo keyworded twice: the first version spends slots on synonyms, content-type words, diacritics and AI terms the guide bans, while the second keeps twenty distinct concepts in relevance order.
Ethnicity, age and gender have their own rule
Vecteezy asks contributors not to combine ethnicity, age and gender into a single keyword. The guide's example is to write Caucasian ethnicity, senior man rather than Caucasian senior man. The information is the same, but split into separate terms the search index can match independently instead of one compound phrase almost nobody types.
There is a second half to it. When the subject is too far away or too covered to be identified, or the frame is cropped to a body part, the guide says to omit ethnicity, age and gender entirely rather than guess. A silhouette at distance is not a senior woman, and claiming otherwise puts a keyword on the file that the image itself does not support.
Keywords are locked once you submit
Vecteezy's contributor guidance states that keywords cannot be changed after submission. This is the rule that should change how you work, because almost every other agency habit assumes a second pass. There is no re-keywording a Vecteezy file that underperforms, and no quiet fix for a typo you notice six months later.
So the quality gate has to sit before the upload rather than after it. A keyword set should be read once against the rules, with the count inside the band, the first five ordered deliberately, singular forms only, no content-type or AI terms and nothing in the list that is not in the frame. On an agency where feedback arrives as a rejection rather than an edit request, checking metadata in advance is much cheaper than reshooting.
Titles carry keywords too, just not for internal search
Vecteezy requires a unique title on every resource, and its keywording guidance says to put core keywords in the title because titles also serve as the page title for search engines. That is a different job from the keyword box. The keywords feed Vecteezy's own search; the title does its work on Google, where a duplicate title across eleven variants of one product helps nobody.
Practically, a title should read like a sentence a person wrote, carry the one or two terms sitting in keyword positions one and two, and stop short of repeating the keyword list. The field rules themselves, including length, casing and the CSV that carries them, are a separate subject covered in the Vecteezy field reference.
A keywording order of operations that survives a real batch
The sequence that holds up across a few hundred files starts narrow and widens. Name the subject in one or two words. Add its material, type or action. Add the setting. Add mood, style and conceptual terms last. Stop between 20 and 30. Then read the first five back as if you were the buyer typing them into the search box, and reorder if they do not match.
Doing that by hand for a batch is the part that breaks. The volume problem is not generating words, it is generating words that respect one agency's vocabulary while a dozen others want something different, and keeping the ordering intact when the same file goes out to all of them. That is the work an automated metadata pass should be absorbing, and it is why a keyword set built for one agency rarely survives the trip to another unedited.
Rastock AI writes titles, descriptions and keywords from the image itself, applies each destination's rules including banned and mandatory terms, and delivers over FTP or SFTP to 10+ agencies including Vecteezy, with per-destination status and rollback if a batch goes wrong. There is no revenue share, no ownership claim on your files, and IPTC and CSV export stay open. The free trial runs 14 days and does not ask for a card.
Running Vecteezy alongside the rest of your agencies
Vecteezy's rules are not unusual in kind, since most agencies weight early keywords and dislike padding. The specific combination is what makes it restrictive: a tight working band, singular forms only, no synonyms, and no edits after submission. A portfolio going to several agencies needs the Vecteezy copy treated as its own version rather than as a truncated Adobe set.
If your existing Vecteezy files were keyworded before you read the guide, the honest answer is that you cannot repair them. What you can do is stop the pattern on everything new, and clean up the keyword sets you still control elsewhere, so the same habits do not follow your work from one agency to the next.
Frequently asked questions
How many keywords does Vecteezy allow?
Vecteezy publishes two different figures. The keyword guide in the Eezy contributor help centre states a minimum of 5 and a maximum of 50, with 20 to 30 recommended per resource. The contributor checklist on Vecteezy's blog states a maximum of 30 and encourages at least 10. Working in the 20 to 30 band satisfies both documents, so that is the safe target for any batch.
Do the first five keywords really matter on Vecteezy?
Yes. Vecteezy's keyword guide says the most relevant keywords should be listed first and that the first five are the most important, and its contributor blog confirms that search gives those positions extra weight. Two files carrying the same keywords in a different order are not treated equally, so position one should hold the word a buyer would type first.
Can I edit Vecteezy keywords after a file is approved?
No. Vecteezy's contributor guidance states that keywords cannot be changed after submission. There is no re-keywording pass for a file that underperforms and no correcting a typo six months later, which is why the metadata check has to happen before the upload rather than after approval.
Should I add plurals and synonyms to cover more searches?
No. Vecteezy asks for one form of each keyword, in the singular, and says its system already covers synonyms and variations. The guide's own example of what to avoid is kitten, kitty, kitties, cat, cats, baby cat, which spends six slots on one concept. On a list capped at 30, that padding pushes out genuinely different terms.
Is it against the rules to put AI Generated in Vecteezy keywords?
Yes. The keyword guide says not to include AI or any version of AI Generated in keywords. The same applies to content-type words such as photo, vector and video, since every file on the site is one of those. AI provenance is declared where the agency asks for it during submission, not in the search index.
How should ethnicity, age and gender be keyworded?
Separately rather than combined. Vecteezy's example is Caucasian ethnicity, senior man instead of Caucasian senior man, which gives the index three terms it can use independently. When the subject is too distant, covered or cropped for those attributes to be visible or relevant, the guide says to leave them out altogether rather than estimate them.