Food images sell on specificity, not volume of keywords. A buyer searching a stock library almost never types the word food; they type a dish, a preparation state, a surface, a viewpoint and sometimes a mood, and the files that surface are the ones carrying all five. Keywording a food photo well means describing those layers in order of importance, then stopping before you start padding.
Food is also one of the most crowded categories on every major agency, which cuts both ways. Demand is genuinely deep, because restaurants, delivery apps, recipe publishers, grocery brands and meal-kit companies all license continuously. But a plate of pasta described as pasta, food, meal, delicious is competing against tens of thousands of identically described plates of pasta and will lose to any file that says more.
What a buyer actually types
Commercial food searches tend to arrive with a job attached. A recipe site wants a hero shot with room for a headline. A delivery app wants a dish that reads instantly at thumbnail size. A nutrition publisher wants an ingredient photographed honestly rather than styled into an advert. Each of those is a different query, and the difference shows up in words like overhead, copy space, rustic, close-up and homemade rather than in the name of the dish.
The other shift worth keywording for is toward food that looks real. Buyers have moved away from hyper-polished plates toward natural light, wooden and linen surfaces, visible crumbs, half-eaten portions and hands in frame. If your image has those qualities, say so explicitly, because natural, rustic, imperfect and homemade are terms buyers now type on purpose.
The five layers every food file needs
Work through these in order and put the earliest layers first in your keyword list, because on most agencies the opening terms carry the most weight.
1. The subject, named as precisely as you can defend
Sourdough bread beats bread. Carbonara beats pasta. Include the specific name and the broader category, since buyers search at both levels, but never claim a dish you cannot see. If the shot is generic short pasta in tomato sauce, describe it as that rather than labelling it a regional speciality, because a reviewer comparing keywords to the image can bounce the file for it.
2. Preparation and state
Raw, sliced, chopped, kneaded, baking, grilled, steaming, freshly baked, cooling. This layer is where a lot of commercial intent lives, because an article about meal prep needs chopped vegetables specifically, not vegetables. Cooking verbs also unlock process searches, which are less contested than finished-plate searches.
3. Setting, surface and styling
Wooden table, marble countertop, cast iron pan, ceramic bowl, linen napkin, restaurant kitchen, home kitchen, outdoor picnic. Buyers match food to the environment of the brand they are designing for, so the surface is frequently the deciding factor between two similar dishes. Include the light too when it is distinctive: natural window light and moody dark food photography are both searched by name.
4. Composition and usable space
Flat lay, overhead view, top view, close-up, macro, 45 degree angle, copy space, text space, banner, vertical. Designers search these constantly because layout constraints come before aesthetics in most briefs. Copy space in particular is worth its slot in the top ten: a competent food photo with deliberate empty area is far more licensable than a tighter, prettier crop with nowhere to put a headline.
5. Concept and audience
Healthy eating, comfort food, vegan, gluten free, plant based, family dinner, street food, meal prep, sustainable. These are the terms a marketing team searches when it has a message rather than a menu. Dietary claims need care: label a dish vegan only if nothing visible contradicts it, since a stray parmesan shaving turns a useful keyword into a rejection reason.
Where the agencies disagree
One keyword set rarely travels cleanly between agencies, and food files hit the differences quickly because the descriptive vocabulary is long.
Adobe Stock accepts up to 49 keywords per submission but steers contributors toward roughly 25 to 35 accurate terms, and its own guidance is explicit that the first ten carry the greatest weight in search. It also asks you to use each keyword once. That makes ordering a real editorial decision rather than an afterthought: the dish, the preparation and the composition term belong in the opening block, and the mood words belong further down.
Shutterstock requires a minimum of seven keywords and allows up to 50, and treats the description as prose rather than a list. A comma-separated pile of terms in the description field is a documented rejection trigger there, so the same metadata that satisfies a CSV column for one agency has to become a readable sentence for another. Descriptions must be in English and long enough to read as a sentence rather than a label.
Four ways to do the keywording, honestly compared
By hand
Best for: small, high-value food sets where you know the cuisine and the client. Cost: your time. Limitation: it does not scale, and consistency drifts across a long session, which is how the same dish ends up keyworded three different ways in one portfolio. It remains the most accurate method for anything unusual, regional or dietary-sensitive, because you are the one who knows what is actually in the bowl.
Desktop keywording tools such as Xpiks and StockSubmitter
Best for: contributors who want local control and keyword presets they can reuse across a shoot. These are mature tools with strong batch editing, and StockSubmitter in particular is built around submitting to many agencies from one desktop. Pricing follows a desktop or licence model rather than a share of your royalties. Limitation: they run on your machine, so throughput is tied to it, and the keyword suggestions lean on dictionaries and your own presets rather than reading the picture.
Generic AI image taggers and vision APIs
Best for: getting a first pass of literal object labels very cheaply at large volume. They will reliably tell you there is a bowl, a spoon and a tomato. Limitation, and it is the important one for food: they are trained to describe, not to sell. They rarely produce copy space, flat lay, comfort food or meal prep, they do not know any agency's field limits, and they will happily suggest a brand name that gets the file refused.
Managed pipelines such as Wirestock and BlackBox
Best for: contributors who would rather hand over the whole distribution problem and not think about metadata or agency accounts at all. They genuinely remove work, and for someone shooting food occasionally that convenience can be worth it. Limitation: these models take a share of royalties, which is a permanent cost on every future licence of a file you shot once. For a growing food library that compounds against you.
Stock-specific metadata tools such as Rastock AI
Best for: libraries large enough that consistency matters and per-agency rules have become the bottleneck. The distinction from a generic tagger is that the output is shaped by each destination: field limits, keyword caps, banned and mandatory terms per agency, and delivery over FTP or SFTP to 10+ agencies, with a rollback if a batch goes out wrong. Pricing is a flat subscription with no revenue share and no ownership claim, and IPTC and CSV export stay open. Limitation: it is metadata and delivery, not a managed service, so you still keep your own agency accounts and make the editorial calls.
When to use which
Under a few hundred food files and one or two agencies, keyword by hand or with a desktop tool. The overhead of anything else is not repaid, and your own knowledge of the cuisine is the most valuable input you have.
Between roughly a thousand and several thousand files across three or more agencies, the per-agency mapping is what breaks first, not the keyword ideas. This is where a stock-specific tool earns its place, and where a generic tagger tends to disappoint, because the failure is about rules rather than recognition.
If you shoot food rarely and never want to touch a contributor portal, a managed pipeline is a reasonable trade. Just price the royalty share over the lifetime of the files rather than over this month, because food images have long tails and a well-keyworded dish can license for years.
Mistakes that bury or reject food files
Visible branding is the quiet killer in food photography, because packaging, bottle labels and logo-embossed cutlery sneak into frame constantly. Keywording the brand name makes it worse, not better. Second, padding with near-synonyms such as tasty, yummy, delicious and appetising in one file adds nothing to search and reads as stuffing to a reviewer. Third, reusing one keyword set across a whole shoot flattens the variation that made the individual frames worth licensing.
The general principles behind all of this are covered in our guide to choosing keywords for stock photos, and if you are deciding what to run your food library through rather than how to word it, we compared the options in best tools for food stock content.
The honest summary is that keywording food well is a writing problem before it is a software problem. Get the five layers right on twenty files by hand and you will understand what any tool should be producing for the next two thousand. If that volume is where you are heading, Rastock AI has a 14-day free trial with no card required, which is enough to run one real food shoot through it and compare the output against the set you would have written yourself.
Frequently asked questions
How many keywords should a food stock photo have?
Aim for roughly 25 to 35 accurate terms rather than filling the maximum. Adobe Stock accepts up to 49 per submission but its own guidance points at that narrower range, and Shutterstock requires at least seven while allowing up to 50. Food images can justify a fuller set than a simple product shot because there are genuinely five layers to describe, but every term still has to be defensible against the picture.
Which keywords should come first?
Put the specific dish name, its broader category and the preparation state in the opening block, followed by the composition term if the image has usable copy space. On Adobe Stock the first ten keywords carry the most weight in search, so ordering is part of the metadata rather than a cosmetic choice. Mood and concept words such as comfort food or healthy eating belong further down the list.
Can AI keyword food photos accurately?
It depends on what kind of AI. Generic vision models identify objects well and will reliably name the visible ingredients, but they describe rather than sell, so they tend to miss commercial terms like copy space, flat lay and meal prep, and they do not know any agency's field limits. Tools built for stock contributors add those layers and the per-agency rules, which is the part that determines whether a file is findable.
Should I use the same keywords on every agency?
The keyword ideas can be shared, but the formatting cannot. Adobe caps titles at 70 characters in its CSV and allows up to 49 keywords, while Shutterstock expects the description to read as a proper sentence and treats a comma-separated keyword list there as a rejection trigger. Copying one metadata set across agencies without adapting it is the most common reason a well-shot food file gets refused on the second platform.
Why do my food photos get rejected for keywords?
The two usual causes are irrelevance and stuffing. Reviewers compare keywords against the image, so naming a regional dish you cannot actually identify in frame, or claiming a dietary property such as vegan when something visible contradicts it, is grounds for refusal. Piling up near-synonyms like tasty, yummy and delicious in one file reads as manipulation. Brand names on visible packaging cause a separate problem and should be removed from the frame, not described.
Does copy space really matter for food images?
It matters more in food than in most categories, because so much food licensing goes into recipe headers, menu banners, delivery app promotions and social templates that all need room for text. A technically weaker image with deliberate empty space frequently outsells a tighter, prettier crop that leaves a designer nowhere to put a headline. Keyword it explicitly, and include the viewpoint terms such as overhead or flat lay alongside it.