123RF will not review a file that has not been keyworded properly — its own contributor pages say reviewers leave such images alone rather than assess them. The platform asks for at least seven keywords in English, a factual description with no brand names, and a JPEG of no less than six megapixels. After acceptance, the text is locked.
What 123RF checks before a reviewer sees the image
Most agencies treat metadata as something that affects discoverability after approval. 123RF treats it as a precondition for approval. Its contributor FAQ is explicit: if an image is not keyworded or described adequately and accurately, reviewers will not review it. The file is not rejected on quality grounds — it simply never enters the queue in a usable state.
That changes the order the work has to happen in. Where metadata is a post-approval concern, you can upload first and tidy up later. Here the description and the keyword list are part of the submission itself, and the technical floor sits underneath both: a file under six megapixels is rejected automatically, before a human or your keywords matter at all. The general approach to writing stock metadata still applies, but the sequence here is less forgiving.
The keyword field: seven minimum, thirty to fifty ideal
The stated minimum is seven keywords per image. That number is low enough to hit without thinking and low enough to mislead, because 123RF's own keywording guide puts the useful range much higher — it describes roughly thirty to fifty relevant keywords as ideal. Submitting seven meets the rule and wastes the file.
The mechanics of the field are worth knowing precisely. Keywords are separated by commas, and everything between two commas is treated as one keyword or key phrase. A two-word entry sitting between commas is therefore a single phrase, not two separate terms. Contributors who assume the field splits on spaces end up with phrase entries where they meant individual words, and the terms they wanted never get indexed on their own.
Order carries weight here the way it does across the industry. Adobe's own metadata documentation states that the first ten positions have the greatest influence on search ranking, and the same ordering principle is the safest assumption on any agency that does not publish its weighting: put the terms a buyer would actually type at the front and leave the long tail behind them.
Check a title and keyword list
123RF asks for at least 7 keywords and describes 30 to 50 as ideal. Keywords are comma-separated, so anything between two commas counts as a single phrase. Paste a title and keyword list to see length, duplicate terms and separator problems before you upload.
Words that do not belong in a 123RF keyword list
123RF names the categories it does not want, and it names examples, which makes this easier to enforce than the vaguer policies elsewhere.
Brand and trademark terms are out. The keywording guide cites Audi, McDonald's and Teflon directly, and says copyrighted keywords of famous brands and names should not be included. This holds even when the brand is genuinely visible in the frame — the term is the problem, not the photograph.
Event dates are out: the guide asks contributors not to put dates of the event in the keyword list. Unrelated keywords are out, and 123RF calls this by its name — the FAQ uses the word spamdexing. Padding a list with high-volume terms the image does not contain is the fastest route to a file nobody reviews. And errors are out, explicitly: keywords should be free from spelling mistakes, typos and broken grammar. A misspelled keyword is worse than a missing one, because it occupies a slot and matches nothing.
The description field is factual, not atmospheric
123RF's description guidance is unusually prescriptive about tone. It asks for factual information only, concise and clear rather than evocative. Brand names and trademarked terms stay out of the description for the same reason they stay out of the keywords. If you are used to agencies where the description is a chance to sell the mood, this is a different field doing a different job.
Two specifics are easy to get backwards. Geographic locations belong in the description when they are relevant, and this is one of the few places the guide asks you to add detail rather than strip it. Model ethnicity or origin, on the other hand, is to be left out of the description, with the guide reserving that for the keyword field instead. A description reading "a Japanese woman in an office" runs against the stated rule, while the equivalent keyword is acceptable.
English is the only language the fields accept
Descriptions and keywords are to be provided in English. That is not a limit on who can buy the file: 123RF handles non-English searches by translating the query, which means the index it translates into is the English one you wrote. A keyword list in another language does not reach those buyers. It just fails to reach anyone.
Technical limits that decide the outcome before metadata matters
Metadata never gets a chance on a file that fails the technical floor. Photographs must be at least six million pixels, roughly 2000 by 3000, and the contributor FAQ states that anything smaller is rejected automatically. The upload guide adds the ceiling: the file must not exceed 30 MB. Format is JPEG only, at a minimum of 80% quality compression. Illustrations go in as EPS vectors instead, and scanned work has its own floor of at least 300 DPI for photographs, with higher requirements for negatives.
Six megapixels is a modest floor by current standards — most cameras clear it comfortably, and the files that fail are usually crops, phone exports or downsized web copies submitted by mistake. If you are working out what resolution stock agencies actually need across a mixed portfolio, 123RF sits at the permissive end and the 30 MB ceiling is more likely to bite than the megapixel minimum.
Delivery is where the two paths diverge. 123RF's own guide recommends the web form uploader only for small batches, below ten files, and FTP for anything larger. The FTP host is ftp.123rf.com, authenticated with the account username and password rather than a separate set of credentials.
AI-generated work has its own category and its own banned terms
AI-generated images are accepted, but routed. They go into the AI Generated Images category and nowhere else, and the guidance adds a keyword restriction that does not apply to photographs in the same way: AI work must not be tagged with real place names, real people's names, or brand and trademark names. The examples given are the London Eye and the Eiffel Tower for places, Abraham Lincoln and Elon Musk for people, and Nike, Disney and Marvel for brands.
The reasoning shows up in the rejection list. 123RF declines AI files for missing or extra digits, additional limbs, distorted or unidentifiable objects, wrong scale, watermarks or nonsensical text, and for near-identical variations of the same image submitted as separate files. A keyword naming a real landmark on a synthetic image makes a claim the file cannot support, which is the same logic other agencies apply in their own wording — worth comparing if you distribute AI work widely and need one metadata approach that satisfies several AI policies at once.
123RF pre-submission check
The fields 123RF states it checks before review: a minimum of 7 keywords in English, no brand names or event dates, JPEG at 6MP to 30MB and 80% quality, EPS for vectors, and AI files restricted to the AI Generated Images category.
Releases decide what your text is allowed to claim
123RF requires valid model or property releases where content depicts identifiable people, brands, places or properties. The release is not only a legal attachment — it sets the boundary on what the metadata may say. Without a model release a file cannot be described or keyworded as though it were cleared for commercial use, and a recognisable private property constrains the text the same way. Attaching releases correctly at upload time matters more here than on platforms where you can soften a description later.
The part you cannot undo
The rule that catches experienced contributors is that descriptions and keywords cannot be amended once images have been accepted. The help page puts it plainly: it is not possible for you to edit or amend after acceptance.
On most platforms a weak keyword set is a problem you can revisit, and re-optimising an older portfolio is a standard exercise. Here it is not available. Whatever went up with the file is what the file carries for as long as it stays online. That single rule is the whole argument for getting metadata right on the first pass instead of treating approval as the deadline.
An order of operations that holds up
Working backwards from the locked-text rule gives a sequence that survives contact with a real batch. Confirm the technical floor first, because nothing downstream matters on a file that is auto-rejected. Write the description second, factual and concise, with a location if one is relevant and no brand names anywhere.
Build the keyword list third, in layers: the literal contents of the frame, then the concept terms a buyer would type, then the long tail — aiming well above the seven-keyword minimum. Reorder so the ten strongest terms sit at the front. Then run a separate removal pass for brands, dates, unrelated terms and typos, because those are the four categories 123RF names explicitly and they are the easiest to miss while you are still writing. Upload last: web form under ten files, FTP above that.
If you are new to the platform rather than tuning an existing workflow, the wider 123RF contributor picture — account setup, contributor levels and how royalties are structured — sits alongside these field rules.
Where tooling helps and where it stops
The mechanical half of this automates well: counting keywords, catching duplicates, holding a title length, flagging brand terms against a list, and keeping one set of metadata consistent across agencies whose rules disagree. Rastock AI generates titles, descriptions and keywords from the image, applies per-agency banned and mandatory keyword rules, and delivers over FTP or SFTP to 10+ agencies including 123RF, with a rollback if a batch comes out wrong.
What no tool does is decide what is in your photograph, and on a platform where the text locks at acceptance the final read is yours. The arrangement that works is a tool producing a consistent first draft and a contributor checking it once, carefully, because there is no second attempt at the field. There is a 14-day free trial with no card required if you want to test that against a batch of your own before committing.
What this comes down to
123RF's requirements are not unusually demanding. Seven keywords is the lowest stated minimum among the major agencies and six megapixels is a modest floor. What is unusual is the finality: reviewers will not look at a file that arrives without adequate text, and that text cannot be changed once the file is accepted. Both rules push the same conclusion — on 123RF, metadata is not what you do after the upload. It is part of the upload.
Frequently asked questions
How many keywords does 123RF require?
The stated minimum is seven keywords per image, but 123RF's own keywording guide describes roughly thirty to fifty relevant keywords as ideal. Seven satisfies the rule and leaves the file badly indexed. Keywords are separated by commas, and everything between two commas is treated as one keyword or key phrase, so a two-word entry counts as a phrase rather than two terms.
Can I edit my 123RF keywords after the image is approved?
No. 123RF states that it is not possible to edit or amend descriptions and keywords once images have been accepted. This is the single most important difference from agencies where you can re-optimise an old portfolio later. Whatever text goes up with the file is what the file carries for as long as it stays online, which makes the pre-upload review the only review you get.
What image size and format does 123RF require?
Photographs must be at least six million pixels, roughly 2000 by 3000, and anything smaller is rejected automatically. The file must not exceed 30 MB. Format is JPEG only, at a minimum of 80% quality compression. Illustrations are submitted as EPS vectors instead, and scanned photographs need at least 300 DPI, with higher requirements for negatives.
Can I use brand names in 123RF keywords?
No. The keywording guide names Audi, McDonald's and Teflon as examples and says copyrighted keywords of famous brands and names should not be included. The rule applies even when the brand is visible in the frame, because the keyword itself is the problem. Brand and trademarked terms are also to be kept out of the description field.
Does 123RF accept AI-generated images?
Yes, but they must be submitted to the AI Generated Images category and nowhere else. AI files additionally must not be tagged with real place names such as the Eiffel Tower, real people's names, or brand and trademark names. 123RF also declines AI work for extra or missing digits, distorted objects, nonsensical text, and near-identical variations submitted as separate files.
Should I upload to 123RF by FTP or the web form?
123RF's own guidance recommends the web form uploader only for small batches, below ten files, and FTP for anything larger. The FTP host is ftp.123rf.com, and it authenticates with your 123RF account username and password rather than a separate set of credentials. Metadata still has to be complete, because reviewers will not assess an inadequately keyworded file.