Titles and descriptions pass stock moderation when they factually describe the content, contain no trademarks, brand names or personal information, use one language, avoid duplicate keywords, and stay within each agency's limits - 49 keywords on Adobe Stock, 7 to 50 on Shutterstock. Editorial submissions must additionally carry an accurate date and location.
Metadata rejections are the most frustrating kind, because nothing is wrong with the photograph. The file is technically clean, the subject is commercial, and it still comes back refused for a reason that is entirely mechanical and entirely preventable. Unlike aesthetic rejections, these rules are published, stable, and enforceable by software.
This article sets out the actual published requirements at the two largest agencies, then compares which tools enforce them before submission and which simply generate text and leave compliance to you. All rules cited below come from Adobe Stock and Shutterstock contributor documentation current in mid-2026; both revise these pages, so treat the vendor page as authoritative if it disagrees.
The published rules, agency by agency
Adobe Stock
Titles: brief and clear, ideally under 70 characters, focused on what is visually most important. Adobe specifically advises against overly technical or gear-heavy wording - a title is not the place for lens and aperture.
Keywords: up to 49 per submission, ordered by importance, with the first ten carrying the greatest influence on search ranking. Adobe recommends mixing general and specific terms, conceptual terms, location, number of people, viewpoint and model demographics.
The four mistakes Adobe names explicitly: irrelevant keywords; trademarks, brand names or personal information anywhere in the submission; more than one language in a single submission; and using the same keyword twice. Titles and keywords must also match the language configured on your contributor account, not simply be in any consistent language.
Generative AI content: Adobe refuses submissions that reference artist names whose work is still in copyright - in the prompt, the title or the keywords. This catches a large amount of AI-generated work whose creators pasted their prompt into the description field without editing it.
Shutterstock
Description/title: must be a factual sentence describing the subject matter, relevant to the content, maximum 2,048 characters, and in English. Exceptions are made for scientific Latin names of plants and animals, place names, and foreign terms in common English use.
Keywords: a minimum of 7 and a maximum of 50, in English, relevant to the content. The minimum matters - contributors who submit three keywords on a simple product shot get refused for insufficient metadata rather than for quality.
Categories and location: one category is required and a second is optional; location is optional but should be specific where provided. Shutterstock publishes definitions for every category, and mismatched categorisation is a refusal reason in its own right.
Editorial requirements: documentary editorial descriptions must include the day, month and year the content was taken and the geographical location. For content shot before 2000, the year or decade alone is sufficient. An inaccurate or missing date or location triggers the Metadata - Editorial Requirements rejection.
Also refused: trademarks and brand names in commercial metadata; spelling and grammatical errors; emojis; spam; personally identifiable information such as social security or financial numbers; support case numbers pasted into titles; and terminology that is offensive or disrespectful regarding race, ethnicity, disability, age, religion, gender or sexual orientation.
The rules that catch people out
Three of these cause disproportionate damage because they are invisible to the person submitting.
Trademarks you did not notice you wrote. A keyword list containing "jeep", "band-aid", "jacuzzi" or "velcro" contains trademarks even though all four are used generically in speech. AI taggers reproduce this habit faithfully, because the training corpus is written English. This is the most common single cause of an otherwise clean batch being refused.
Duplicates hidden by inflection. "Child" and "children", "run" and "running", "woman" and "women" are treated as repeated keywords rather than as extra coverage. Adobe's guidance to use each keyword once means the singular/plural padding habit inherited from early SEO actively hurts.
Prompt residue in AI submissions. Descriptions like "in the style of [artist], 8k, hyperrealistic, octane render" fail on two counts: the artist reference is a stated refusal reason on Adobe, and the render jargon is not a factual description of the subject. Strip prompts before submission; they are generation instructions, not metadata.
Which tools actually enforce compliance
Almost every tool in this market generates metadata. Far fewer check it against agency policy before it leaves your machine. That distinction is what this comparison is about.
PhotoTag.ai
Best for: contributors who want direct control over generated vocabulary. It supports custom context before processing and explicit blocking of terms you never want the AI to produce, which is the mechanism that maps most directly onto a trademark blacklist. Lightroom Classic integration keeps the work inside an existing editing pipeline.
Pricing model: credit packs at one credit per file, publicly listed at 2,000 for $18, 10,000 for $59 and 50,000 for $190, plus a low-cost monthly subscription. 10 free credits to test.
Main limitation: blocking is a list you maintain, not policy knowledge the tool holds. It will not tell you that Shutterstock needs a minimum of 7 keywords or that your editorial caption is missing a date. Compliance depends entirely on how good your own blacklist is.
Xpiks
Best for: manual precision. Xpiks includes spell checking and duplicate detection in its keyword editor, which addresses two named rejection reasons directly, and it works offline on a desktop. For contributors who prefer to write metadata themselves and want a safety net rather than a generator, it is the most focused option here.
Pricing model: desktop application, no revenue share, no per-file charge.
Main limitation: no per-agency policy layer. It will catch a misspelling but not a trademark, and it does not know that Adobe caps at 49 while Shutterstock caps at 50. Development has been slow in recent years.
StockSubmitter
Best for: contributors submitting to many agencies at once. It has the deepest per-agency submission logic of anything in this comparison, including agencies whose forms are not FTP-based, and its presets let you standardise safe vocabulary across a whole catalogue.
Pricing model: metadata filling and core tagging are free; paid tiers apply to submission volume. No revenue share.
Main limitation: field mapping is not the same as policy validation. It will deliver your metadata into the right fields at each agency; it will not object that the content of those fields breaks a rule. The interface is dense and much of the support material is Russian-language first.
Wirestock
Best for: contributors who want compliance to be someone else's problem. Keywording and distribution are handled for you, which means the moderation risk shifts to the platform. It also explicitly accepts AI-generated work when properly declared, and consolidates payouts into one request.
Pricing model: free to join, with a commission on royalties publicly stated at 15%. Free-tier upload limits apply.
Main limitation: you cannot fix what you cannot see. When metadata is written for you, a systematic compliance problem is invisible until acceptance rates drop, and you have no lever to correct it. The commission continues for as long as the work earns.
Rastock AI
Best for: contributors who want policy rules applied before delivery rather than discovered at review. Banned and mandatory keyword lists run across a whole batch, copyright guard flags visible trademarks, logos and recognisable faces before submission, and per-agency templates keep each destination inside its own field limits. Delivery is FTP and SFTP to 25+ agencies with IPTC and XMP embedding.
Pricing model: subscription tiered by monthly upload volume, priced by plan. No revenue share, no ownership claim, CSV export always available.
Main limitation: automated policy checks reduce rejections; they do not eliminate them. Judgement calls - whether a building is recognisable enough to need a property release, whether a caption is factually accurate - remain human decisions, and no vendor can promise otherwise. Agencies also change rules without notice, so any rule set lags reality by some amount.
The agency's own upload portal
Best for: the final word. Adobe generates suggested keywords during upload and lets you review and reorder them before submitting, which is the only suggestion source guaranteed to be aligned with the agency's own search engine. It is also free and always current.
Pricing model: free with a contributor account.
Main limitation: one agency, one batch, one browser tab. Suggested keywords are also literal rather than conceptual, and they are suggestions rather than validation - Adobe will happily let you submit a trademark it suggested nothing about.
When to use X vs Y
PhotoTag.ai vs Rastock AI. Choose PhotoTag.ai if you want a tagger and will handle submission and compliance yourself, and if credit-based pricing suits irregular volume. Choose Rastock AI if you want the policy rules and the delivery in the same place, and your volume is steady enough that a subscription is cheaper than credits.
Xpiks vs any AI generator. Choose Xpiks if you write your own metadata and want spell checking and duplicate detection as a safety net. Choose an AI generator if writing is the bottleneck. These are complementary rather than competing, and plenty of contributors generate elsewhere and inspect in Xpiks.
StockSubmitter vs Rastock AI. Choose StockSubmitter if breadth of destinations is the priority, especially agencies with non-FTP submission forms. Choose Rastock AI if the failures you actually experience are metadata refusals rather than missing destinations.
Wirestock vs doing it yourself. Choose Wirestock if repeated rejections have made contributing feel not worth the effort and you would trade a share of royalties to stop thinking about it. Choose to keep control if you intend to scale, because compliance is a skill that compounds and a commission is a cost that compounds too.
Any tool vs the agency portal. Below a few hundred files a year, the agency's own portal with its suggested keywords is sufficient and free. Third-party tooling earns its cost when you are submitting to several agencies with different field limits and cannot check each batch by hand.
A pre-submission checklist
Run this over any batch before it leaves your machine. Most of it takes seconds once it is a habit.
Is the title a factual sentence about the subject, under about 70 characters, with no gear jargon and no prompt residue? Does the keyword count sit inside each destination's limits - at least 7 for Shutterstock, no more than 49 for Adobe? Is every keyword unique, including across singular and plural forms? Is everything in one language, matching your Adobe account setting?
Then the ones people forget: no brand names, even generic-sounding ones; no artist names on AI-generated work; no personal information, case numbers or emojis; correct spelling throughout; a required category selected; and for editorial, an accurate day, month, year and location in the caption.
For the non-metadata half of the problem, our breakdown of Adobe Stock rejection reasons covers technical and quality refusals, and our guide to categorisation and release attachment covers the fields that sit alongside titles and keywords in a submission.
The honest summary
Metadata compliance is unusual in stock contributing because it is genuinely solvable. The rules are published, they change slowly, and a machine can check most of them. What software cannot do is make the judgement calls - whether a description is accurate, whether a landmark needs a release, whether a term crosses from descriptive into disrespectful. Tools that claim to remove all rejection risk are overselling; tools that catch trademarks, duplicates, language mixing and field limits before submission are removing the majority of the avoidable ones, which is a real and measurable difference in acceptance rate.
Frequently asked questions
Why was my content rejected for title or keywords when the photo was fine?
Metadata rejections are separate from quality rejections. The usual causes are irrelevant or padded keywords, a trademark or brand name in the text, duplicate keywords including singular and plural forms, more than one language in a submission, or - for editorial content - a missing or inaccurate date and location. All of these are checkable before you submit.
How many keywords do Adobe Stock and Shutterstock allow?
Adobe Stock accepts up to 49 keywords per submission and states that the first ten positions carry the greatest influence on search ranking. Shutterstock requires a minimum of 7 keywords and allows a maximum of 50. Both treat these as ceilings rather than targets, and padding a list with marginal terms is itself a refusal reason under metadata relevance rules.
Can I use brand names in keywords if the product is visible in the photo?
No, not for commercial content. Both Adobe Stock and Shutterstock refuse commercial submissions whose metadata contains trademarks, brand names or product names, regardless of what appears in the frame. This includes words that feel generic in everyday speech but remain registered marks. Editorial submissions follow different rules and require accurate identification instead.
What has to be in an editorial caption?
Shutterstock requires documentary editorial descriptions to include the day, month and year the content was taken plus the geographical location where it was shot. Content from before the year 2000 needs only the year or decade. If the date or location is missing or inaccurate, the submission is refused under Metadata - Editorial Requirements rather than for image quality.
Do I need to remove my AI prompt from the description?
Yes. Adobe Stock refuses content that references artists whose work is still in copyright in the prompt, title or keywords, and render jargon such as resolution claims or engine names is not a factual description of the subject. Write a plain sentence describing what the image shows and discard the generation instructions entirely.
Can a tool guarantee my metadata will pass moderation?
No, and any tool claiming it can is overstating. Software reliably catches mechanical failures - trademarks on a blacklist, duplicate keywords, field limits, language mixing, spelling. It cannot judge whether a description is factually accurate, whether a building is recognisable enough to need a release, or how a reviewer will read a borderline term. Expect fewer rejections, not zero.