When a Phrase Works Better Than a Single Word

On microstock, a compound keyword phrase is not decoration and not a trick. It is simply a more precise way of telling the algorithm and the buyer what is actually in the file. Sometimes one word is too broad. Doctor leaves far too many possibilities open, while doctor consultation already defines a specific scenario. The same goes for ice cream, remote work or seamless pattern: these are not random couplings but established search constructions that describe content more accurately.

The largest platforms allow this in plain language. Shutterstock words its keyword requirement like this: "use specific and unique words (or compound words)" — compound words are not merely tolerated, they are recommended. Adobe, for its part, stresses the essentials: keywords must be relevant, not excessive, and ordered by importance.

This is exactly why the "single words versus compound phrases" debate is a false one. Good attribution needs both. Single words provide reach, phrases add context. If a photo shows a doctor and a patient in an examination room, tags like medical, doctor, patient are often not enough on their own. A more precise pairing such as medical appointment or doctor consultation conveys the meaning of the frame without drifting into abstraction. But this only works when the phrase is genuinely natural language and genuinely matches the image.

Where the Benefit Begins — and Where the Harm Starts

Compound keywords start paying off where they answer a real buyer query. Adobe Stock advises contributors to add words describing the subject, the action, the setting and any unique details, then to arrange them by importance: the first 10 keywords carry the most weight in search placement. Shutterstock likewise bets on precision, requiring you to stay within a range of 7 to 50 keywords and not to repeat the same words or word stems unnecessarily.

The harm begins where a phrase stops being a description and becomes an attempt to game the system. Shutterstock supplies its own examples of what counts as spam: "car ride, car road, car drive, car move, car adventure, car trip, car journey, car vector" or "seamless pattern argyle, seamless pattern green, seamless pattern blue". Here a single stem is dragged across the entire list — that is no longer metadata, it is a simulation of relevance. Adobe puts the consequences even more bluntly: irrelevant keywords hurt your search placement and reduce your chance of sales, they can get a file rejected, and repeat offenders risk having their account blocked.

And here is the detail that is most often misread. Not every repeated stem is forbidden. In that same help article, Shutterstock explicitly lists "Kansas City" and "ski lodge, ski resort" as acceptable. The difference is not the repetition itself but whether the second occurrence adds new meaning. Two tags sharing a stem while describing different things are fine. Eight tags repeating a stem for the sake of volume are spam. A simple practical marker: if a stem appears in your list more than twice, stop and reread it.

The same principle governs titles. Shutterstock separately describes title spam and gives examples: "run, run, run" or "Summer Season. Summer Wallpaper. Summer Time. Happy Summer. Summer Day". Yet the title "Runner running on city street" is considered perfectly fine — because there the repetition is a natural part of the phrase rather than padding.

A Simple Rule: The Phrase Must Sound Human

Here is a good test: say the keyword phrase out loud. If it sounds like a normal query, it is probably appropriate. If it sounds like words mechanically glued together, leave it out. Remote work, renewable energy, slow motion, financial planning all sound natural. Phrases assembled to fill a quota rather than to convey meaning give themselves away as spam very quickly.

This is not just folk wisdom. Adobe recommends a short title — ideally under 70 characters — and explicitly advises checking whether it sounds natural when spoken aloud. The platform itself is proposing precisely this test.

Titles also carry formal requirements that are forgotten more often than any other. Shutterstock requires the title to be written in English, to be at least five words long, and to be a complete descriptive sentence rather than a list of keys: a description should "read like a sentence or phrase and not like a list of words". The English-language requirement extends to keywords as well, with stated exceptions for scientific Latin names, place names, and foreign expressions that have entered common English usage.

So trying to cram your entire body of work into the title through compound phrases is a bad idea. Title and keywords must operate as one system, not as two separate sets of random words.

Other Platforms: Same Principle, Different Rules

Adobe and Shutterstock are not the whole of microstock. On other platforms compound phrases live by their own rules, and the differences are not cosmetic.

Getty Images and iStock work on a fundamentally different model. You cannot simply write any phrase there: keywords are drawn from a controlled vocabulary — a pre-approved taxonomy. Every keyword carries invisible synonyms that the platform itself feeds into search, so a contributor does not need to list variants of the same concept. The practical consequence: a compound phrase you invent may simply not exist in their vocabulary. What wins here is not inventive phrasing but an exact match to an existing list of terms.

Alamy sits closer to the familiar model: up to 50 tags, while a caption plus five tags is enough to get a file live. But it has its own priority mechanism — up to 10 "supertags", which the search engine weights more heavily. In essence this is the same principle as Adobe's top ten: weight depends not only on which words you chose, but on which ones you put first.

123RF requires a minimum of 7 keywords, English only, and warns explicitly against spamdexing — adding unrelated keywords to surface in extra searches. Different wording, same substance as Shutterstock: "be precise and descriptive."

Pond5 advises contributors in its help pages to keep 40–50 keywords and separately cautions against keyword stuffing — "adding irrelevant keywords or an excess of keywords." Its title requirement is stated very briefly: "nothing fancy" — it should convey exactly what is depicted.

One practical conclusion follows from this spread: there is no single number. The minimum on 123RF is 7 keywords, Alamy needs 5 to put a file on sale, Pond5 asks for 40–50, Adobe allows a maximum of 49, Shutterstock up to 50. What they all share is something else: the demand for relevance and the penalty for padding. So build the set from the meaning of the frame first, then trim it to the specific platform's limit — not the other way round. And before your first upload to a new platform, it is worth opening its own help pages: the numbers change, and nobody is going to check them for you.

How I Would Build Attribution in Practice

The best approach is to think not in terms of "how many phrases" but in terms of "which layers of meaning are covered." A single frame usually needs four levels: who is in it, what is happening, where it happens, and which commercial query it might serve.

For a photograph of a doctor and a patient in the same room, for instance, you can assemble several distinct meanings rather than one overloaded block: doctor consultation, medical appointment, patient care, healthcare advice, senior healthcare, clinic visit. This gives search breadth without repetition and without blurring the meaning.

And here is an immediate example of how easily this goes wrong. The tag telehealth visit suggests itself for such a frame — it sounds fitting, the subject matches. But telehealth means a remote appointment, while in the photo the doctor and patient are physically together. The tag contradicts the content of the frame, and a buyer looking for an online consultation leaves with nothing. This is exactly the case where a plausible phrase turns out to be irrelevant — and irrelevance is precisely what Adobe penalises in search placement. Check not only how a phrase sounds, but whether it argues with what is visible in the frame.

The logic is identical for vectors and illustrations. If it is a template background or an ornament, seamless pattern is appropriate. For video, slow motion or motion graphics may be useful — but only when they genuinely describe the format and the content, not when they are there for the count.

What Matters Most Not to Forget

Adobe Stock allows a maximum of 49 keywords, and the first 10 carry the greatest weight in search, so keywords should be arranged by relevance: most important first, secondary later. Shutterstock requires 7 to 50 keywords and separately penalises repetitive or irrelevant wording. Alamy adds its own 10 supertags on top of that, while Getty and iStock accept only terms from a controlled vocabulary. In every case the best strategy is the same: less noise, more precision. Twenty strong, honest keywords beat fifty random ones.

One more thing that often gets overlooked: on Adobe the keyword order is edited manually, by dragging, and that is worth using — order is exactly what helps a buyer find your work faster. Any automated suggestion, wherever it comes from, remains an assistant and not the author of your metadata: the final word on what is actually in the frame always belongs to the contributor.

Conclusion

Compound keywords are neither good nor evil. They are simply a powerful instrument that only starts working in the hands of a contributor who can think like a buyer. If a phrase is natural, precise and genuinely describes the image, it helps. If it is assembled for the illusion of SEO, it gets in the way. On microstock the winner is not whoever added more words, but whoever managed to name the file honestly, clearly and with commercial precision. This is exactly the approach we aim for in MetaBrain: we build attribution so that it stays clean, relevant and useful for real search on platforms such as Adobe Stock and Shutterstock.