Guides

What Is Photo Metadata? EXIF, IPTC, and XMP Explained (2026)

What photo metadata is, in plain English: what EXIF, IPTC, and XMP each store, which fields survive export and upload, and how to fill them in at scale.

By Tagrly Team , Editorial Published July 17, 2026 7 min read

A photographer reviewing photo metadata on a laptop beside a camera, prints, and a memory card reader.

Open any photo's file info panel and you'll find dozens of fields: Exposure, Creator, Keywords, Copyright, something called XMP. Most of them are empty, nobody ever explained which ones matter, and every guide on the subject reads like a standards committee wrote it.

Quick answer: Photo metadata is text stored with an image that records how the photo was made, what it shows, and who owns it. It comes in three standards: EXIF (technical data the camera writes automatically), IPTC (caption, keywords, and copyright a person fills in), and XMP (Adobe's container that stores both, used by Lightroom). A fourth layer, catalog tags, lives in a database instead of the file, and it is the layer that makes large libraries searchable.

The three photo metadata standards, in plain English

Every metadata field you'll ever meet belongs to one of three standards. They sound interchangeable and they are not, so here is the working shorthand: EXIF is what the camera did, IPTC is what the photo is, and XMP is the box both travel in.

Diagram of the three photo metadata standards, EXIF, IPTC, and XMP, with what each one stores.

EXIF: the camera writes it

EXIF (Exchangeable Image File Format) is the technical record every digital camera and phone embeds automatically: capture date and time, camera and lens model, aperture, shutter speed, ISO, and on phones, GPS coordinates. You never fill in EXIF by hand. Its value is forensic: it answers "when was this shot, on what, and where" years after everyone has forgotten.

IPTC: a person writes it

IPTC fields come from the news industry, and the IPTC Photo Metadata Standard is still the schema every serious photo tool reads. This is the descriptive and legal layer: caption, keywords, headline, creator name, credit line, copyright notice, usage terms. Nothing here is automatic. If an IPTC field is filled in, a human or an AI put it there.

XMP: the modern container

XMP (Extensible Metadata Platform) is Adobe's metadata format, built to hold EXIF and IPTC values plus anything else, like edit history. It can be embedded inside a JPEG or written as a sidecar .xmp file next to a RAW file, which is why a Lightroom folder fills up with tiny XMP files: your keywords and edits live there, not in the RAW itself.

Where photo metadata lives, including the layer that isn't in the file

The three standards above all travel inside (or beside) the image file. There is a fourth place metadata lives, and it is the one most modern tools are built on: a database.

When a photo tool "tags" your photos, those tags usually go into the tool's own index, not into the file. Lightroom's catalog, Apple Photos' library, and connected AI catalogs like Tagrly all work this way.

Database tags are instant to search across 100,000 photos and hold richer structure than IPTC allows, like a ranked focal subject. The tradeoff is that they belong to the tool: they only follow the file if the tool exports them, so the export path is worth checking before you commit to anything.

Neither home is "correct." Embedded metadata travels with the file forever; indexed metadata searches faster and scales further. The full tradeoff is covered in our automatic photo keywords guide, which owns that decision in detail.

What survives export and upload

This is the part most photo metadata explainers skip, and it is where real workflows break.

Your own storage keeps everything. Google Drive and Dropbox sync files byte for byte, so embedded EXIF, IPTC, and XMP all survive untouched.
Editors export what you tell them to. Lightroom's export dialog can include all metadata, strip location only, or reduce to copyright only. Whatever you pick is what the delivered file carries.
Social platforms strip almost everything. Instagram, Facebook, and most messaging apps recompress images on upload and discard most embedded fields, keywords included.
Google Images reads the ownership fields. Creator, credit, and copyright IPTC fields can surface as an image credit in search results, which is why they are worth filling in even though embedded keywords do nothing for ranking. The full signal map is in our image tagging for SEO guide.

Warning. EXIF on phone photos usually includes GPS coordinates. Publish an original file shot at a client's home and you have published their address. Strip location data on export for anything public; our AI photo tagging safety guide covers this and the rest of the client-photo checklist.

Which photo metadata fields are worth filling in

You do not need most of the panel. The honest priority list:

Creator, credit, and copyright (IPTC), always. Five minutes to set once as a preset in Lightroom or Bridge, applied on import forever. This is your attribution and your legal claim.
Caption and keywords, only if something reads them. Stock agencies require them, desktop search uses them, and news wires expect them. If nothing in your workflow reads IPTC keywords, typing them by hand is effort spent where nothing looks.
Alt text does not live here at all. The description a website needs goes in the page's HTML, not the file, so writing it into IPTC and expecting your CMS to find it is a common dead end.
EXIF, leave alone. It is already correct, except for stripping GPS on public exports.

How to read and write photo metadata at scale

Reading is easy at any scale: every OS shows basic fields in file info, and ExifTool is the free utility that reads and writes every field of all three standards when you need the full picture.

Writing is the bottleneck, because the descriptive layer is the valuable one and it historically required a human. At editorial depth that is 5 to 8 hours per 1,000 photos. This is the exact gap AI tagging closed: a vision model reads each photo and produces the keywords, focal subject, and description automatically. On a working production archive of about 19,000 photos, a fast structured pass ran at roughly 8 minutes per 1,000 photos, and editorial-grade descriptions at about 22 minutes per 1,000.

A photo's empty metadata panel on the left and the same photo with AI-generated description, keywords, and focal subject on the right.

Where the AI output lands depends on the tool. Desktop keyworders like PhotoTag.ai or ON1 write straight into the files' IPTC fields; our photo metadata generator comparison covers that category head to head. Connected catalogs read photos where they already live in Drive or Dropbox and index the results in a searchable database instead; Tagrly is one example, and the wider category is surveyed in our complete guide to AI photo tagging in bulk.

Tip. Pick embedded if your files must carry their own metadata to other tools (stock submission, archives, handoffs). Pick indexed if the goal is a team finding photos fast in a big shared library. Solo stock shooters usually want the first; marketing teams usually want the second.

The bottom line

Photo metadata is three standards doing three jobs: EXIF records the capture, IPTC records the meaning and the ownership, and XMP is the container that carries them through modern tools, with catalog databases as the fourth layer that makes big libraries searchable. Fill in creator and copyright once as a preset, strip GPS from anything public, and let a vision model write the descriptive layer, because that is the part that used to cost hours per shoot. If your photos already sit in Drive or Dropbox, Tagrly's free tier will generate metadata for the first 100 photos in any folder, no credit card, which is enough to see what machine-written descriptions look like on your own work.

Frequently asked questions

What is photo metadata in simple terms?

Photo metadata is text stored alongside the pixels of an image that describes the photo: when and how it was shot, who made it, what it shows, and who owns it. Some of it is written automatically by the camera (date, settings, GPS), and some of it is written by a person or a tool (caption, keywords, copyright). It is the digital version of the note on the back of a print. Metadata is what makes a photo findable and attributable later; a file with none of it is just anonymous pixels named IMG_4827.jpg.

What is the difference between EXIF, IPTC, and XMP?

EXIF is the technical record the camera writes on its own: timestamp, camera and lens model, aperture, shutter speed, ISO, and on phones, GPS coordinates. IPTC is the descriptive layer a person fills in for publishing: caption, keywords, creator name, credit line, and copyright. XMP is Adobe's container format that can hold both, either embedded in the file or as a sidecar .xmp file next to a RAW file, which is how Lightroom stores keywords and edit history. A useful shorthand: EXIF says what the camera did, IPTC says what the photo is and whose it is, XMP is the modern box both travel in.

Does photo metadata survive when I upload a photo to social media?

Mostly no. Instagram, Facebook, and most messaging apps recompress uploads and strip most embedded metadata in the process, including keywords and usually copyright fields. That is partly deliberate privacy protection, since EXIF can contain GPS coordinates. Cloud file storage behaves differently: Google Drive and Dropbox sync the file byte for byte, so everything embedded in it survives. The practical rule is to treat embedded metadata as reliable inside your own storage and archive, and unreliable anywhere a platform re-encodes the image.

Should I remove metadata from photos before publishing them?

Remove the location data, keep the ownership data. EXIF GPS coordinates on a photo taken at a client's home or your own studio are a real privacy leak if you publish the original file, so strip location on export unless the location is public and intentional. Creator, credit, and copyright fields are worth keeping: Google Images reads those IPTC fields and can display an image credit with your name on it in search results. Most editors, including Lightroom, let you choose exactly which metadata blocks an export carries.

Can AI write photo metadata automatically?

The descriptive half, yes. A vision model can look at each photo and produce keywords, a focal-subject label, and a one-sentence description, which covers the part of metadata a human would otherwise type by hand. The technical half (EXIF) never needed a human anyway, since the camera writes it. AI-written descriptions land in one of two places: embedded into the file's IPTC fields by desktop keywording tools, or stored in a searchable catalog database by connected tools that read your Drive or Dropbox in place. On a fast tier this runs at roughly 8 minutes per 1,000 photos, against 5 to 8 hours for a person.

Try Tagrly on your own photo library

Connect your Google Drive or Dropbox folder and Tagrly will tag every photo in bulk. Search by what is actually in the image, share specific shots with clients, and never lose a photo again.

Open the live demo