Writing stock titles and keywords that hold up

A stock title has two audiences and they want different things. The search index wants an accurate description of the content. The buyer scanning a grid of thumbnails wants to know, in about a second, whether this image solves their problem.

A title that satisfies both is a plain, specific sentence about what is in the frame and what it conveys. It is not a slogan, not a filename, and not a keyword list with the commas removed.

• Lead with the subject, then the action, then the setting. Search snippets truncate, and the first words are the ones that survive.
• Be specific where specificity is real. “Golden retriever puppy” beats “dog”. Do not invent detail you cannot see.
• Include the concept if there is one. Buyers search for ideas — teamwork, isolation, growth — as often as for objects.
• Write a sentence, not a list. Most agencies present the title as the caption; it should read as English.
• Mind the length. Adobe Stock allows up to 200 characters, which is a ceiling rather than a goal. Somewhere in the region of 80 to 150 usually says everything true without padding.

The metadata generator reads the image itself and drafts this, capped to the platform limits. Treat the result as a draft — it describes what the model sees, and only you know what the shoot was actually about.

Work through the image in layers. Each layer catches searches the others miss:

1. Literal subjects. Everything visibly in frame — people, objects, animals, place.
2. Actions. What is happening: running, cooking, presenting, waiting.
3. Setting and time. Indoors, kitchen, autumn, night, studio.
4. Concepts and mood. Freedom, deadline, calm, growth. Often the most valuable and most frequently forgotten layer.
5. Technical and compositional. Copy space, close-up, overhead, shallow depth of field, isolated on white. Buyers filter on these.
6. People attributes, if visible and accurate. Age range, number of people, apparent activity — never assumptions about identity.

• Anything not in the image. Irrelevant keywords generate views that do not convert, and a poor view-to-download ratio is not a signal you want attached to your work. Some agencies penalise it directly.
• Trademarks and brand names. A common rejection reason, and a legal risk regardless.
• Artists' and celebrities' names. Not permitted, and not helpful.
• Duplicates and near-duplicates. Singular and plural forms of the same word usually occupy two slots for one concept.
• Multi-word phrases where the agency expects single terms. Rules differ; check the platform you are submitting to.

Several agencies treat the first several keywords as the most significant, and some display only the first handful in the interface. Ordering by relevance rather than by the order they occurred to you is therefore worth the minute it takes.

A reasonable ordering: primary subject, primary concept, action, setting, then the long tail of supporting detail. The metadata generator returns keywords already ordered by relevance, which mostly removes this chore.

The failure mode with generated metadata is not that it is wrong; it is that nobody reads it. A short review pass:

1. Read the title aloud. If it does not describe the picture, fix the title first — the keywords usually follow from it.
2. Scan the keyword list for anything you cannot point at in the frame, and delete it.
3. Add the two or three things only you know: the location, the season, the concept you shot it for.
4. Check the counts against the platform limits before export.

The title and keyword ranker does the mechanical part of this — it scores clarity, coverage and commercial intent, flags weak, generic and duplicate keywords, and proposes stronger titles. For a whole shoot, the bulk generator runs the folder and exports one CSV, which you should still read before submitting.

It is worth being blunt about the limits. No tool on this site — and no tool anywhere that lacks access to an agency's internal data — can tell you search volume, competition, how an image will rank, whether it will be accepted, or what it will earn.

A score from the ranker is an assessment of the text you supplied: is it specific, does it cover the obvious searches, does it read as something a buyer would type. That is genuinely useful, and it is not a forecast. Any tool presenting confident numbers about future performance is generating them.

Related reading

What actually gets embedded in the file — and why a keyword typed into an upload form is not the same as one written into the image.