A photo tag is the keyword text embedded inside an image file — the words a stock agency's search engine reads to decide whether your picture matches a buyer's query. Get the photo tag wrong, or lose it during export, and a technically perfect image sits invisible in a catalog of millions of files. This guide covers the three metadata standards that actually carry those tags — EXIF, IPTC, and XMP — and how to keep keywords intact from your editing software to the agency's server.
What a "photo tag" actually is: EXIF, IPTC, and XMP explained
Three separate metadata standards live inside a typical stock photo, and only two of them matter for keywording. EXIF is camera-generated technical data — shutter speed, ISO, GPS, camera model — written automatically at the moment of capture, with no input from you. IPTC is the older, purpose-built standard for exactly what stock buyers search on: IPTC metadata is descriptive and administrative information that you or an organization adds manually (copyright, keywords, creator name). XMP came later — XMP (Extensible Metadata Platform) was published by Adobe as an XML-based wrapper designed to hold IPTC and EXIF fields plus custom data, edit history, and star ratings in one flexible container.
In practice the two you write on purpose are IPTC and XMP, and they usually mirror each other. A capture date can appear as EXIF DateTimeOriginal, IPTC DateCreated, and XMP xmp:CreateDate all at once, and a photographer's name can be EXIF Artist, IPTC By-line, and XMP dc:creator. Usually they agree, but sometimes they do not, because one app updated XMP but left the old EXIF value in place, and when values conflict, which one wins depends on the app reading the file. Adobe tools generally trust XMP first, then fall back to EXIF and IPTC — which matters because that's the software most contributors use to prep files.
Where your keywords actually live — and where they get lost
The most common way a photo tag disappears isn't a bug — it's a setting. In Lightroom, if there are keywords applied to the photo and you choose a metadata option that includes them, those keywords will be written as a flat list to the IPTC Keywords field on export. But that only happens if you tell it to. In the Keyword List panel, right-clicking a keyword and selecting "Edit Keyword Tag" reveals export behavior, and if the "Include on Export" checkbox is unchecked for a keyword, it will not be included in exported images, regardless of the "All Metadata" selection. A contributor can spend twenty minutes keywording a batch, export with the wrong preset, and upload files with an empty keyword field without ever seeing an error.
A second, less obvious trap is the sidecar file. Sidecar .xmp files add another wrinkle: the edits and ratings you see in Lightroom may not be embedded in the JPEG at all, they may live only in the sidecar, and if you hand someone the image file alone, that metadata never travels. If you're moving files between machines, cloud drives, or tools that only read embedded data — including most agency FTP intake systems — a keyword sitting in a sidecar is invisible to them.
It's also worth knowing where photo tags reliably survive and where they don't. Instagram strips all EXIF data on upload, and downloaded images contain zero metadata, because Instagram processes every photo through its own pipeline of resize, compress, and convert, and metadata doesn't survive the process. Facebook also strips EXIF completely, Twitter/X strips all EXIF on upload, and LinkedIn strips EXIF data from uploaded images as well. None of this affects a direct stock agency upload, but it's a reminder that "the metadata is in the file" and "the metadata will still be there after this passes through some other app" are two different claims.
How each stock agency reads your embedded photo tags
Agencies don't all treat embedded IPTC/XMP the same way. Adobe Stock will read metadata straight from the file: you can fill fields one image at a time in the portal, or write directly into the file's IPTC fields beforehand, since Adobe reads embedded IPTC/XMP data on upload. Adobe's own guidance is to keyword generously but not carelessly — including keywords, up to 49 per content submission, helps customers discover your content, using relevant nouns, adjectives, and concepts specific to your content. It also offers a prompt for accuracy: include accurate titles and keywords that reflect the subject and setting of your content.
Shutterstock's own documentation frames metadata as one connected system: description/title, keyword, category, and location information are collectively called "metadata," and metadata drives Shutterstock's search and discovery tools. On volume, at least 7 keywords that capture details and context in addition to the description are required per submission, though the practical ceiling is much higher than the minimum — most experienced contributors go well beyond seven for real visibility.
Getty Images and iStock work differently from both. They don't score whatever text you type — Getty and iStock do not run on free-text keywords; every term is matched against Getty's controlled vocabulary, a fixed, curated list of approved terms, and any keyword that is not on the list does nothing. That's why a photo tag that ranks well on Adobe Stock can be functionally invisible on iStock if the exact phrasing isn't in their vocabulary. It's also worth knowing that Getty's model is largely a one-shot submission: once a file has been accepted, contributors can no longer add keywords to them, and if a file is lacking essential keywords, the contributor has to open a support ticket and the keyword team decides whether to add them. Get the photo tag right before you submit — there's no easy self-service fix afterward.
This is where Rastock's metadata generation is built around the fact that IPTC, XMP, and the CSV each agency expects aren't the same output. Writing coherent, policy-aware tags once and letting them route to Adobe Stock's field limits, Shutterstock's structure, and a Getty-friendly vocabulary shape saves the trap of assuming one keyword list works everywhere unchanged.
Writing photo tags that survive export, upload, and re-edits
The safest habit is to write your photo tag into the file's IPTC/XMP fields before you ever open an agency's upload form, not after. You can fill metadata one image at a time in the portal, or write it into the file's IPTC fields before uploading — the IPTC route is faster and your metadata survives even if you re-upload or migrate to another agency. That last point matters more than it sounds: metadata typed only into a web form lives in that platform's database, not in your file, so if you ever pull the image back out, resubmit it elsewhere, or rebuild your portfolio on a new machine, the keywords are gone unless they were embedded.
If you're exporting from Lightroom or similar software, check the export dialog's metadata option deliberately rather than trusting the default. Options like "All Except Camera & Camera Raw Info" will include things like keywords, title, caption, star rating, and color labels, but a narrower preset chosen for a different job (say, stripping location for privacy) can silently drop keywords too if it wasn't built with stock submission in mind. For catalog-wide edits, changes made in Lightroom Classic must be written to XMP for other applications to recognize them, and metadata is only synced this way if the catalog setting is enabled.
Doing this by hand across two or three agencies is manageable. Doing it across the ten-plus agencies a serious contributor might sell on — each with a different keyword ceiling, a different controlled-vocabulary quirk, and a different CSV template — is where the process breaks down. Rastock's bulk FTP upload and status tracking exists specifically for that stage: one metadata pass, embedded IPTC/XMP plus a free CSV/IPTC export, distributed and tracked per agency without a revenue share or any claim on the files themselves. If your workflow is a handful of files a month on a single platform, that overhead isn't worth paying for — see the pricing page and compare it honestly against just doing it in Lightroom.
Common photo tag mistakes that quietly kill discoverability
Keyword padding is the most common self-inflicted wound. Shutterstock's own position is direct: search still works with single entries — repeating a word in different phrasings does not stack the odds, it just risks a rejection for keyword spam. A photo tag list should describe what's actually in the frame, not every synonym you can think of.
Vague or list-style titles are the second. Reviewers and search algorithms both penalize titles that read like a keyword dump rather than a sentence, and the most common metadata rejections include a title that is too short or generic, keywords that don't match the image content, or a category that doesn't match the main subject. Related to that, forgetting that export settings can strip keywords silently is a recurring frustration in contributor communities — support threads routinely walk through checking that "Include All Metadata" is selected and that keywords are set to export in the Keywording panel before assuming a batch upload will carry them. For a deeper look at how metadata mistakes turn into agency rejections, see our guide on common stock photo metadata rejection reasons.
Finally, treating every agency's keyword field the same is a mistake baked into a lot of workflows. A tag list tuned for Adobe Stock's free-text search won't automatically perform on Getty's controlled vocabulary, and a Shutterstock-optimized description format won't match what an editorial buyer expects elsewhere. For a breakdown of exactly how many keywords each major platform actually rewards, see our post on how many keywords you should use for stock photos.
Frequently asked questions
What's the actual difference between IPTC and XMP for keywords?
IPTC is the older, fixed-structure standard where keywords, captions, and copyright are stored in a defined set of fields. XMP is Adobe's newer XML-based wrapper that can hold the same IPTC data plus edit history and custom fields. Most software writes to both simultaneously, so for stock keywording you generally don't need to choose one — you need both to stay in sync.
Do I need to embed a photo tag if I fill in an agency's web form anyway?
Not strictly for that one upload, but embedding is what makes the keyword portable. Metadata typed only into a platform's form lives in that platform's database; if you ever re-upload the file, move to another agency, or rebuild your library, form-only keywords are gone while embedded IPTC/XMP tags travel with the file.
Will Instagram or Facebook strip my stock photo's embedded keywords?
Yes, for any copy shared on those platforms. Instagram and Facebook strip most EXIF and IPTC data, including camera details and location, from the public-facing image. This doesn't affect files you upload directly to a stock agency, but it means you can't rely on a social-media copy retaining your metadata for any other purpose.
Why do my Getty/iStock keywords not seem to work even though they're relevant?
Getty and iStock match submitted terms against a fixed controlled vocabulary rather than scoring free text. A perfectly accurate keyword that isn't on their approved list contributes nothing to search. Once a file is accepted, you also can't self-edit keywords — you'd need to file a support request to add missing ones.
Why did my keywords disappear after exporting from Lightroom?
Usually an export setting, not a bug. Lightroom only writes keywords to the file if a metadata-inclusive option is chosen in the export dialog and the keyword's own 'Include on Export' setting is checked. A narrower metadata preset, or a keyword marked not to export, will leave the field empty in the exported file even though it shows correctly in the catalog.
How many keywords should I actually embed in a photo tag?
It depends on the agency, but relevance always beats volume. Shutterstock requires at least seven keywords and allows up to fifty; Adobe Stock allows up to forty-nine but prioritizes accuracy over padding; Getty/iStock cares more about matching their controlled vocabulary than hitting a specific count. Describe what's genuinely in the frame rather than maximizing the field.