A photo nobody can find might as well not exist. You can shoot the best frame of your life, but if it lands in a library of fifty thousand images with no keywords attached, the only way anyone ever sees it again is by scrolling past it on the right day. Keywording is the work that makes a photo findable: it is how you, a client, a search engine, or a stock buyer locates one specific image out of thousands.
Most photographers treat keywording as a chore to rush through or skip entirely. That is a mistake. Good keywords are the difference between an archive you can search in seconds and a black hole you avoid opening. This guide covers how to choose keywords that work, how many to use, how to structure them from specific to general, and how to keyword large batches without losing your mind.
Keywords are descriptive words attached to a photo that describe what is in it and what it is about. They live in the image's metadata, specifically the IPTC keywords field, alongside the title, caption, and copyright. If you want the full picture of where this data lives and how it travels with a file, start with what photo metadata is. Keywords are one part of that larger system, and they are the part that does the heavy lifting for discovery.
Folders force every photo into exactly one place. A picture of your daughter blowing out birthday candles at the beach has to go in "Family" or "Beach" or "Birthdays," and whichever you pick, it is missing from the other two. Six months later you remember the beach but file it under birthdays, and you cannot find it. Folders are a single hierarchy, and real photos belong to many categories at once.
Keywords solve this. The same photo can carry "beach," "birthday," "daughter," "candles," "sunset," and "celebration" all at once. Search any one of those terms and the photo surfaces. You are no longer guessing which folder you chose. You describe the image and it appears. For an archive of any real size, this is the difference between a usable library and a junk drawer.
Pick a photo you shot a year ago. Now imagine you need to find it and all you remember is the subject, the place, and roughly what was happening. Type those words into your library's search. If the photo comes up, your keywording works. If it does not, no folder structure will save you. Search is only as good as the words you put in.
Different systems read your keywords in different ways, and knowing how each one works tells you what to write.
Tools like Lightroom Classic, Capture One, and digiKam index your keywords into a searchable database. When you type a term, the catalog returns every image carrying that keyword, instantly, across folders and drives. Smart collections take this further: a rule like "keyword contains bird" auto-populates a collection with every bird photo you ever tag, forever, with no manual sorting.
Agencies like Adobe Stock, Shutterstock, and Getty match buyer searches against your keyword field. A buyer searching "happy senior couple gardening" gets results ranked by keyword relevance. If your photo of exactly that scene is missing the words "senior," "gardening," or "happy," it never appears, and you never make the sale. Keywording is not optional here. It is the entire discovery mechanism.
On sites like Flickr, tags directly feed both the site's internal search and, in many cases, what general web search engines can read about the image. The same logic carries to your own website, where keywords inform alt text and captions. The mechanics of writing for discovery are covered in depth in the image SEO guide and, for the specific platform, in Flickr SEO and tags.
The goal is to predict the words a searcher would actually type, then make sure those words are on the photo. A reliable way to generate them is the who, what, where, when, why method. Walk through each question and write down what you see.
| Question | What it captures | Example keywords |
|---|---|---|
| Who | Subjects and people | woman, fisherman, two people, crowd |
| What | Objects and actions | fishing rod, casting, river, sunrise |
| Where | Location and setting | Scotland, Highlands, riverbank, outdoors |
| When | Time and season | dawn, autumn, golden hour, misty |
| Why | Mood and concept | solitude, patience, escape, tradition |
Order matters, both for how you think and for how some platforms weight terms. Start with the most specific accurate word, then add broader ones. For a bird photo, that means "great blue heron" first, then "heron," then "wading bird," then "bird," then "wildlife." The specific term catches the expert searcher who knows exactly what they want. The general term catches the buyer browsing for "bird" who would never find you otherwise. You want both. Skip the specific terms and you blend into ten thousand generic bird photos. Skip the general terms and the casual searcher never reaches you.
People search for the same thing with different words. Someone looking for a picture of a "sofa" and someone looking for a "couch" want the same image, so include both. Add "kid," "child," and "youngster" where they fit. Watch plurals too: a search for "mountains" may not match a keyword of "mountain" on every system, so include the form your audience is likely to type. The cost of an extra accurate synonym is near zero, and the upside is a search you would otherwise have missed.
Literal keywords describe what is physically in the frame: "businessman," "laptop," "coffee cup," "office." Conceptual keywords describe what the image means or communicates: "productivity," "remote work," "deadline," "ambition." Buyers and editors search heavily on concepts because they are looking for an idea to illustrate, not just an object. A photo of a single tree on an empty plain is literally "tree" and "grassland," but conceptually it is "solitude," "isolation," and "minimalism," and those concept words are what an art director searches. Tag both layers.
There is no magic number, but there is a healthy range. For most purposes, somewhere between fifteen and thirty relevant keywords per image is the sweet spot. Stock agencies often allow up to fifty, and many photographers use the full allowance, but only when every term genuinely describes the photo. The principle is simple: include every keyword that is accurate and useful, and stop the moment you start reaching.
Both failures come from the same misunderstanding: treating keyword count as the goal. The goal is accuracy. Too few keywords leaves money and discoverability on the table. Stuffing in irrelevant terms is worse than too few, because it actively damages you. Many stock platforms track the rate at which buyers click your image and immediately bounce, and a photo that shows up for searches it does not match gets flagged. A spammed keyword field can drag down the ranking of your entire portfolio, not just the one bad image.
The single biggest threat to a searchable archive is inconsistency. If you tag some photos "USA," others "U.S.A.," and others "United States," then no single search finds all of them. Multiply that across hundreds of terms and your archive quietly fractures into incompatible pieces.
A controlled vocabulary fixes this. It is simply a fixed, agreed list of the exact keywords you will use, so you pick from the list rather than inventing a new spelling every time. Decide once that you write "United States," not "USA," and that you write "color," not "colour," and then never deviate. Lightroom Classic, Capture One, and similar tools let you import a keyword list and choose from it, which makes consistency the easy path instead of the disciplined one.
A hierarchy organizes your controlled vocabulary into a tree, where specific terms nest under broader parents. For example:
Animals > Birds > Raptors > Eagles > Bald Eagle
The payoff is automatic parent keywording. When you tag a photo "Bald Eagle," a hierarchy-aware catalog includes "Eagles," "Raptors," "Birds," and "Animals" on export without you typing them. You tag the one specific term you know, and the structure fills in every broader category for free. This is the cleanest way to get specific-to-general coverage at scale.
Adding trending or high-volume words that do not describe the photo. It backfires: buyers bounce, platforms penalize, and your reputation as a reliable contributor takes the hit.
Mixing "NYC," "New York City," and "Manhattan" without a controlled vocabulary. No single search ever returns the full set. Pick one form and stick to it.
Applying one shoot's keyword set to every image regardless of content. The wide shot and the tight portrait need different terms. Shared keywords are fine for location and event; content keywords must be specific to each frame.
You know the model's name and that it was her birthday, but the photo just shows a smiling woman with a cake. Tag the visible scene, because that is what strangers search. Private context belongs in the caption, not the keywords.
Listing only literal objects and ignoring the idea the image conveys. Editors and art directors search on concepts, so a photo with no conceptual keywords misses the highest-value searches.
Thorough keywording is genuinely slow. Writing twenty to thirty accurate keywords for a single image takes a few minutes when you do it well, and across a few hundred images that becomes a full day of work. This is exactly why most photographers cut corners, and why their archives end up under-tagged. The cure is to change the task from writing keywords to reviewing them.
AI vision tools look at each image and propose a comprehensive keyword set: subjects, objects, setting, colors, lighting, and often the mood and concepts as well. The model does not get bored on image two hundred, so the last photo in a batch gets the same thorough treatment as the first. It catches obvious terms that a tired human skips and frequently surfaces synonyms and concepts you would not have thought to add. For the mechanics of running this across a large set, see the batch metadata workflow.
AI is strong on what is visibly in the frame and weak on what it cannot know. It cannot reliably tell a great blue heron from a grey heron, name the exact national park, or know that the building is a specific landmark rather than a generic office tower. It also does not know your private context. Review every batch for three things: correct identifications, especially species, places, and named subjects; removal of any term that is close but not quite right; and addition of the facts only you hold. Specialized subjects raise the stakes, which is why wildlife keywording needs a careful human pass on every scientific name.
Picture a single image: an elderly man in a wool sweater sits alone on a weathered wooden dock at dawn, holding a mug, looking out over a still lake wrapped in mist, with autumn trees on the far shore. Here is how the who, what, where, when, why method turns that into a real keyword set, ordered specific to general.
Who: elderly man, senior man, old man, one person, fisherman
What: wooden dock, jetty, coffee mug, wool sweater, sitting, looking away, lake, calm water, reflection, mist, fog
Where: lakeside, lakeshore, outdoors, rural, nature, wilderness
When: dawn, early morning, sunrise, autumn, fall
Why (concept and mood): solitude, peace, calm, reflection, contemplation, retirement, stillness, escape, mindfulness
That is roughly thirty-five keywords, every one of them true to the image. Notice what is happening: the literal terms ("dock," "mug," "lake") catch people searching for objects, the synonyms ("elderly man," "senior man," "old man") catch the same searcher's different wording, and the concept terms ("solitude," "retirement," "mindfulness") catch the art director who needs an image to illustrate calm or aging well. Notice also what is absent: there is no "happy," because his expression is contemplative not joyful, and there is no "summer," because the trees say autumn. Accuracy is the filter. If a term is not visibly true, it does not go on.
PhotoScanr analyzes each image and generates accurate, specific-to-general keywords along with titles, captions, and descriptions, ready to review and export
Try PhotoScanr FreeFree to use . No sign-up required . Instant results