A meta tag on a picture is not one thing. It is three separate systems that happen to share a name: the metadata written inside the file, the metadata you supply alongside it in a spreadsheet, and the HTML tags that describe the image on a web page. Stock agencies read the first two. Search engines read the third. Most contributors only ever fill in one of them.
That gap is where downloads leak away. A file with perfect keywords typed into an agency portal carries nothing when you re-upload it somewhere else. A portfolio page with beautiful images and empty alt attributes is invisible to image search. Knowing which layer does what is the difference between filling fields and actually being found.
The three layers of picture metadata
Before touching any field, it helps to know where it lives and how far it travels.
- Embedded metadata travels inside the file itself. Copy the JPEG to a new drive, email it, upload it to a different agency - the fields come along. This is EXIF, IPTC and XMP.
- Sidecar and CSV metadata sits next to the file. A spreadsheet row or an .xmp file matched by filename. It is how agencies that ignore embedded fields still get your titles and keywords.
- Web page meta tags describe the image in HTML: the alt attribute, Open Graph tags, structured data. They never touch the file and only matter when the picture is published on a page you control.
If you sell through agencies, layers one and two are your income. If you also run a portfolio site or a print shop, layer three is where image search finds you. They are not interchangeable, and filling one does not fill the others.
Embedded metadata: EXIF vs IPTC vs XMP
EXIF is what the camera writes. Shutter speed, aperture, ISO, focal length, lens model, capture time, and often GPS coordinates. You almost never edit it by hand, and agencies rarely use it for search. It matters for two reasons: reviewers sometimes check it to confirm a photo is a photograph, and GPS data can expose a location you did not intend to publish.
IPTC is what the human writes. It is the descriptive layer: Title, Description or Caption, Keywords, Creator, Copyright Notice, Credit Line, City and Country. This is the standard that stock agencies were built around, and it is the one that pays.
XMP is the modern container Adobe introduced to hold all of it. In practice most software writes IPTC values into XMP fields and keeps a legacy IPTC copy for older readers, which is why the same keyword can appear twice when you inspect a file. That duplication is normal and harmless - problems start when the two copies disagree because one tool updated only half of them.
If you want the mechanics of how these standards are written and read, the guide to IPTC, XMP and embedded keywords goes deeper than this reference does. Here the goal is coverage: knowing every field exists so none of them stays empty by accident.
The embedded fields stock agencies actually read
Out of the dozens of fields the IPTC standard defines, a small handful carry almost all the weight in stock. These are the ones worth filling on every single file.
- Title, also called Object Name or Document Title. A short descriptive sentence, not a filename. Adobe Stock advises keeping titles brief - under about 70 characters - and that is a sensible ceiling everywhere.
- Description or Caption. Longer context: who, what, where, and the likely use case. Some agencies show it to buyers, some only index it, but a filled caption never hurts.
- Keywords. The field that decides whether your file appears in a search at all. Adobe Stock accepts up to 49 per submission; other agencies set their own ceilings, and going over usually means the extras are silently dropped.
- Creator and Copyright Notice. Your attribution, travelling with the file. Agencies overwrite their own credit line, but these fields protect you everywhere else the image ends up.
- City, State and Country. Underused and genuinely valuable for editorial and travel content, where buyers search by place before they search by subject.
The practical payoff of writing these into the file rather than typing them into a portal is portability. Adobe Stock reads embedded IPTC and XMP when you deliver by FTP, which means a correctly tagged batch can go from your drive to review without you opening a browser form. The same file, unchanged, then carries the same metadata to the next agency.
When the file cannot carry it: CSV and sidecars
Not every agency reads embedded fields, and not every format holds them reliably. Video containers and some vector formats are inconsistent about what survives an export. That is what the second layer is for.
A CSV maps filenames to metadata in columns. Each agency defines its own column names and order, so the same portfolio needs a differently shaped sheet for each destination - the Shutterstock CSV template is a good example of how specific those requirements get. Column headers that do not match are the single most common reason a bulk upload lands with empty titles.
An .xmp sidecar is the other option, and the one RAW workflows depend on. Because a RAW file should not be rewritten, editors keep the metadata in a matching file next to it. It works well as long as the pair stays together - rename one without the other and the link breaks silently.
Meta tags for pictures on a web page
The third layer has nothing to do with the file and everything to do with the HTML around it. If you publish a portfolio, a blog or a print shop, this is what image search reads.
- The alt attribute is the one that matters most. It describes the image for screen readers and for search engines, in a plain sentence. It is not a keyword list, and treating it like one is read as spam.
- Open Graph and Twitter card tags control the thumbnail when someone shares the page. They do not affect ranking, but they decide whether a shared link looks like a portfolio or like a broken box.
- Structured data, specifically the ImageObject type, lets you declare creator, license and content URL in a form search engines parse directly. For anyone licensing their own work, this is the field that states the terms machine-readably.
The filename itself counts here too - descriptive-file-name.jpg outperforms DSC_4821.jpg, and the surrounding caption text gives search engines context the image alone cannot. The full treatment of that side is in the guide to image keywords for SEO.
Mistakes that quietly cost downloads
- Exporting with metadata stripped. Most editors offer a privacy option that removes everything except copyright. Turn it on once and forget, and an entire batch ships blank.
- Filling the keyword ceiling with near-duplicates. Twenty variations of the same idea do not widen your reach; they dilute relevance and invite moderation flags.
- Letting the title default to the filename. A title of IMG_2043 is a wasted field on the one line buyers actually read.
- Reusing one description across a whole shoot. Agencies treat near-identical metadata across many files as a similar-content signal, and it can hold up the entire batch in review.
- Skipping the AI disclosure. Where an agency requires generative content to be declared, that declaration is metadata like any other - and missing it is a policy problem, not a search problem.
- Leaving GPS coordinates in personal work. EXIF location data is precise. For a landscape it is useful; for anything shot at home it is not something to publish.
Filling these fields at volume
None of this is hard on one image. It becomes hard at four hundred, which is where most portfolios actually live. Doing it by hand means either a metadata template that makes every file look the same, or an evening per batch. Both cost more than they look like they do.
The alternative is generating the descriptive fields from the image itself and writing them straight into IPTC and XMP, then producing whatever CSV shape each destination expects. That is what Rastock does across the pipeline: metadata, delivery to 10+ agencies over FTP and SFTP, and per-agency status tracking in one place - with banned and mandatory keyword rules so the output stays inside each platform's policy rather than merely being fast.
The control side matters as much as the speed. Metadata you cannot review, override or roll back is metadata you do not own. Keywords should be editable before delivery, IPTC and CSV export should be free to take elsewhere, and the files should stay yours - no revenue share, no claim on the work.
If you want to see what your own files look like once every field is filled properly, there is a 14-day free trial with no card required.
Frequently asked questions
What is a meta tag on a picture?
It is any structured field that travels with an image rather than living in the visible pixels. In practice that covers three separate systems: EXIF written by the camera, IPTC and XMP written by you or your software, and HTML attributes written into a web page. Different consumers read different layers, so a field that matters to a stock agency can be invisible to a search engine.
Do stock agencies read EXIF data?
Agencies read EXIF for technical and moderation purposes rather than for search. Capture date, camera model and colour space help reviewers judge whether a file meets submission standards. Discoverability comes from the descriptive fields instead, which is why a photo with flawless EXIF and an empty keyword field still ends up effectively unfindable after publication.
Which matters more, IPTC or XMP?
XMP is the more modern container and is what most current software writes, but you rarely have to choose. Adobe applications and most metadata tools write both, keeping the legacy IPTC block and the XMP packet in sync. The failure worth avoiding is not picking the wrong standard; it is exporting through a tool that writes neither of them.
Why did my keywords disappear after I edited a photo?
Almost always an export setting. Save for Web and several batch resize presets strip metadata by default to keep file size down, so keywords written before the edit do not survive it. Check the metadata option in your export dialog, and verify one finished file before pushing a whole batch to an agency.
Do HTML meta tags help a photo rank on an agency site?
No. Alt text, og:image and schema markup apply to pages you control, such as your own portfolio site. Inside an agency the only inputs are the fields that agency stores for your asset, supplied through embedded metadata or a submission form. The two systems never read each other, which is why both are worth filling in separately.
Can I add metadata to thousands of files at once?
Yes, and at any real portfolio size it is the only workable approach. Bulk metadata editors and AI keywording tools write IPTC and XMP across an entire folder in one pass, then export a CSV for agencies that prefer their own template. The part worth keeping under your control is review, since a bad field written once becomes a bad field written ten thousand times.