On every stock agency, search is the storefront. Buyers don't browse — they type a phrase, scan the first two rows, and license something within minutes. Keywords decide whether your photo is in those rows or on page fourteen. This guide covers how agency search actually ranks images, how many keywords each platform wants, why order matters, and a repeatable method for keywording stock photos that works whether you tag by hand or generate metadata with AI.
How stock agency search actually works
Every major agency ranks search results using roughly the same ingredients: how well your title and keywords match the query, how early the matching terms appear in your keyword list, and how the file has performed historically — clicks, downloads and licensing revenue.
You can't control performance history on day one. Metadata is the one ranking ingredient that is 100% in your hands, and it compounds: a file that matches the right queries early starts collecting downloads, and those downloads then lift it further. Weak keywords don't just cost you today's search — they starve the file of the history it needs to rank tomorrow.
The 3-layer keyword method
Strong keyword lists aren't brainstormed — they're built in three passes, from the most literal to the most conceptual:
Layer 1 — Subject (what is literally in the frame)
The nouns a buyer would use to find exactly this image: golden retriever, puppy, autumn leaves. These go first. If someone typed only your first three keywords, they should get an accurate picture of the file.
Layer 2 — Context (where, when, how)
Environment, season, time of day, action and composition: park, golden hour, running, close-up, copy space. Context terms are what separate your file from ten thousand near-identical subjects.
Layer 3 — Concept (what the image means)
The ideas a designer is actually shopping for: happiness, companionship, carefree, new beginnings. Concept keywords are where commercial buyers live — a marketing team searches “teamwork”, not “five people at table”.
Work through the layers in order and you'll naturally produce 25–40 relevant terms with the strongest ones first — which is exactly what agency ranking rewards.
How many keywords per platform?
Limits and ranking behavior differ by agency. The "sweet spot" below is where relevance stays high without padding:
| Platform | Max keywords | Sweet spot | Order matters? |
|---|---|---|---|
| Adobe Stock | 49 | 25–35 | Yes — first 10 weighted most |
| Shutterstock | 50 (min 7) | 25–40 | Low — relevance matters more |
| Getty / iStock | 50 | 15–30 | Mapped to controlled vocabulary |
| Dreamstime | 80 | 30–50 | Yes — early terms weighted |
| Freepik | 50 | 20–30 | Low |
The universal rule: stop when you run out of true things to say. Padding a 28-keyword image to 50 with near-misses dilutes every strong term above it. Platform-specific rules are covered in our dedicated guides for Adobe Stock, Shutterstock and Getty / iStock.
Keyword order: the first 10 decide the most
Adobe Stock openly weights the first 10 keywords most heavily; Dreamstime behaves similarly. In practice that means your keyword list needs an opening line-up, not just a pool: the subject, the most specific two-word phrases a buyer would type, and the single strongest concept — all inside the first ten positions.
A useful test: cover everything after keyword 10. If what remains wouldn't let a stranger guess the image, your best terms are buried too deep.
The 7 categories every keyword list should cover
When a list feels thin, it's almost always missing whole categories rather than individual words. Run through these seven:
Subject
golden retriever, puppy, dog
Action
running, playing, jumping
Environment
park, outdoors, autumn forest
Concepts & emotions
joy, freedom, companionship
Composition & technique
close-up, copy space, shallow depth of field
Light & color
golden hour, warm tones, backlit
Season & occasion
autumn, fall, weekend
If people are recognizable in the frame, add demographic terms buyers filter by (woman, man, senior, family) — but only with a model release on file, since those terms imply one exists.
Five keywording mistakes that hurt ranking
Keyword spam
Adding terms that aren't in the image to chase search volume. Agencies actively penalize it, reviewers reject for it, and buyers who land on the wrong file never come back to your portfolio.
Trademarks and brand names
“iPhone”, “Nike”, “Lego” in keywords is one of the most common rejection reasons — even when the product is genuinely in the photo. Describe generically: smartphone, sneakers, building blocks.
One keyword block for the whole batch
Copy-pasting the same 40 keywords across 100 different images fragments relevance and looks like spam to reviewers. Every file needs keywords for its own content.
Stopping at the literal
Tagging only what is visible and skipping concepts. The buyer with a budget searches for the idea — “work-life balance”, “sustainability”, “digital nomad” — far more often than for the objects.
Ignoring keyword order
On Adobe Stock and Dreamstime the first positions carry the most ranking weight. A perfect keyword hidden at position 45 is doing almost nothing.
The first two aren't just ranking problems — they're rejection problems. Our guide to metadata mistakes that get stock photos rejected covers what reviewers actually look for.
Manual vs. AI keywording: an honest comparison
Manual keywording produces excellent results on a handful of images and collapses at scale: by image 80 of a shoot, fatigue means fewer keywords, sloppier order and copy-paste blocks. AI keywording inverts the trade-off — perfectly consistent coverage of all three layers across the whole batch, with occasional misses on abstract or highly technical subjects that you correct by hand.
The workflow that wins in 2026 is hybrid: generate metadata for the entire batch automatically, then spend your attention on the 10% of images the AI could plausibly misread and on verifying the first-10 order of your best files. Our complete AI keywording workflow walks through this step by step, including the export format each agency needs.
Frequently asked questions
How many keywords should I use for stock photos?
Use 25-40 genuinely relevant keywords. Adobe Stock allows up to 49, Shutterstock up to 50, Getty/iStock up to 50 and Dreamstime up to 80 - but filling every slot with weak terms dilutes relevance. Stop when you run out of words that truly describe the image.
Does keyword order matter on Adobe Stock?
Yes. Adobe Stock weights the first 10 keywords most heavily in search ranking, so lead with the subject and the most specific terms a buyer would type. Order matters less on Shutterstock, but leading with the strongest terms never hurts.
Should I use the same keywords on every agency?
The core keyword set can be shared, but each agency has different limits, ranking behavior and formats. Adobe Stock rewards keyword order, Getty/iStock maps terms to a controlled vocabulary, and Shutterstock requires at least 7 keywords. Export in each platform's own format instead of pasting one list everywhere.
Is it okay to use AI to keyword stock photos?
Yes - all major agencies accept AI-assisted metadata as long as the keywords accurately describe the image. You remain responsible for accuracy, so review the output, especially for abstract images or niche technical subjects.
Why are my stock photos not selling even with many keywords?
Usually because keywords are generic rather than wrong. 'Woman, computer, office' competes with millions of files; 'remote onboarding call, home office, freelancer' competes with thousands. Specific buyer-intent terms in the first positions matter more than total keyword count.
Summary: the keywording checklist
- ✅Build keywords in three layers: subject → context → concept
- ✅Aim for 25–40 genuinely relevant terms; never pad to the limit
- ✅Put the subject and the most specific buyer phrases in the first 10 positions
- ✅Cover all 7 categories: subject, action, environment, concepts, composition, light, season
- ✅No trademarks, no spam, no copy-pasted blocks across a batch
- ✅Export in each platform's own format so keywords survive the upload