
Metadata feels optional right up until an editor asks for a caption, a client requests a photo from two years ago, or five jobs on your drive all contain a file called DSC_1042. The fix is not more metadata. It is one sweep, run once, after the picks exist, on the frames that survived.
Two neighboring jobs are covered in full elsewhere on this site, and this article deliberately does not repeat them. How to build a file name that sorts correctly, field by field, is in the photo file naming system. How to design a keyword vocabulary you will still understand in three years, without losing a weekend to it, is in the keywording guide.
This one is about the pass itself: what you do in a single sitting once the shortlist exists, and in what order.
The order matters for one blunt reason. Metadata effort scales with the number of files you apply it to. Typical ranges put a wedding at 2,000 to 4,000 raw frames and 400 to 800 delivered images, so a keyword or a caption applied before selection costs you several times what the same work costs afterward, and most of that spending lands on frames nobody will ever open again. Do the sweep on the keepers and the same standard becomes affordable.
There is one exception, and it is the reason this article is not called keywording after the cull. Renaming happens at ingest, before anything else. A file name is an identifier, and identifiers have to exist before you start pointing at things. Everything else waits for the picks.
Six steps. Each one applies to fewer files than the one before, which is what keeps the whole thing to a single sitting.
| Step | Applies to | Why it sits here |
|---|---|---|
| 1. Rename at ingest | Every file off the card | Names must be stable before anything references them |
| 2. Pick your keepers | Every file | Everything below only makes sense once this exists |
| 3. Batch the global fields | All keepers, in one action | Copyright and creator are true for every frame, so type them once |
| 4. Job level keywords | All keepers, in one action | Location, event type and client are also true for all of them |
| 5. Captions and specific tags | A few dozen frames | Only images that will be published, licensed or reused earn a sentence |
| 6. Privacy check, then archive | The export, not the originals | The last chance to catch what should not travel |
Steps three and four are the ones people do the hard way, one image at a time, and they are the entire reason this sweep has a reputation for being tedious. They are batch operations. Step five is where your judgment actually earns something, and it should be the only part that feels like typing.
A useful gate before you start: your markers should already be settled, because a sweep run over an unclear shortlist just spreads the confusion into the files. If the picks are still ambiguous, sort that first, and our three way on ratings, flags and labels covers which marker answers which question.
Copyright notice, creator, website, contact, job title, broad location. None of that varies across a shoot, so none of it should be typed twice.
In Lightroom Classic this is a metadata preset. Adobe's route is the Metadata panel, then the Preset menu, then Edit Presets. You fill in the groups you care about, Basic Info, IPTC Content, IPTC Copyright and the rest, then choose Save Current Settings as New Preset from the Preset menu, name it and click Create. Editing an existing one is the same route, followed by Update Preset. Adobe documents both paths.
Build one preset per kind of work rather than one for everything. Wedding, commercial and personal projects usually differ in contact details, usage language or credit line, and a single preset that almost fits is how a wrong copyright year ends up on four hundred delivered files.
Adobe is explicit about synchronizing metadata across a selection: "The metadata you select to synchronize overwrites existing metadata in the selected photos." So sync the fields you mean, not all of them. A blanket sync is how a caption you wrote earlier quietly disappears under a blank one.
Job level keywords go on in the same breath, and this is the part of keywording that genuinely is a batch job: country, region, city, event type, client or job code. That is five or six terms applied to the whole keeper set in one action. Anything more specific than that belongs to individual frames, and the discipline for deciding what deserves a specific term is in the keywording guide rather than here.
Keywords are for retrieval. A caption is for reading. They are stored separately, they serve different people, and using one as a substitute for the other is the most common metadata mistake in editorial work.
A keyword list says: Brooklyn; wedding; ceremony; vows; grandmother. A caption says: "Maya Chen and Jordan Ellis leave their ceremony at The Foundry in Long Island City on June 14, 2026." An editor can publish the second one. Nobody can publish the first.
Captions need full names, correct spellings, a place and a date, which is exactly why they do not scale and should not be attempted on a whole gallery. Write them for the frames that are heading somewhere public: an editorial submission, a licensing library, a nonprofit's annual report, a venue or vendor who will credit you, your own portfolio. On a wedding that is a few dozen images at most.
Two habits make captions cheap. Collect the names and the venue's legal spelling while you are on the job, not from memory a week later, and keep them in the same place as the rest of your job notes. And write the caption once, into the file, rather than into an email, so the sentence travels with the photograph when somebody forwards it. That second habit is what turns a delivered gallery into an asset you can still license in three years.
Metadata travels with exports, which is a feature until it is not. A file headed to a public gallery does not need your home address in the contact field, a child's full name in the file name, or coordinates precise enough to put a stranger at a family's front door.
Lightroom Classic's Export dialog handles most of this in the Metadata section, and the options are worth knowing by name rather than by position. Adobe's published choices are "Copyright Only", "Copyright & Contact Info Only", "All Except Camera Raw & Camera Info", and "All". There is a separate checkbox, "Remove Location Info", which strips GPS data whatever else you chose. A fourth control, "Write Keywords As Lightroom Classic Hierarchy", writes parent and child terms joined by a pipe character, which matters if the receiving system reads keywords literally.
| Where the file is going | A sensible starting point |
|---|---|
| Client gallery | Copyright and contact info, location info stripped |
| Social media and blog | Copyright only, location info stripped |
| Editorial or licensing | All, including caption and keywords, location kept if the story needs it |
| Your own archive | All, nothing stripped, this is the copy that has to answer questions later |
Two things stay out of the file entirely. Model and property release status belongs in your business records, tracked against the job, not typed into an image field where it will be half filled in and unreliable. Private client notes belong in the same place. The rule of thumb is that anything you would not want read aloud should live in the system you control, not in a JPEG you have handed to two hundred wedding guests.
Newborn and family work deserves an explicit pass rather than a default. Keep children's names out of public facing file names, strip coordinates from anything going to social, and ask rather than assume when a family has been quiet about what they want shared. The delivery mechanics around that sit in our client gallery delivery guide.
Metadata that only exists inside one catalog database is one corrupted file away from never having existed. Three habits fix that, and all three are set once.
Write it out to the files. In Lightroom Classic, Catalog Settings has a Metadata tab with an option called "Automatically Write Changes Into XMP". Turn it on and your fields are written into the file or its sidecar as you work, which means another program, another catalog or a future you can read them without your database. It costs a little disk activity and buys you independence from a single application.
Test one export and read it back. Export a single delivered JPEG, open it in a different program, and check that the file name, copyright, caption and keywords all arrived. Do this once per export preset and once again after any software update that touches export. Most people discover their metadata does not survive delivery at the exact moment they need it to.
Leave a readable note beside the archive. A short text file in the job's root folder with shoot date, client, what was delivered, usage terms and the gallery link answers most of the questions that arrive years later, and it is readable on any machine without your catalog, your software or your memory. The longer term storage question, what you keep and for how long, is covered in our photo archiving guide and in whether to keep rejected raw files at all.
Then test the whole thing the only way that counts. Once a quarter, pick a past job and find one specific photograph using nothing but your file names, folders and catalog fields. If it takes more than a few minutes, the standard needs adjusting while the lesson is fresh. Retrieval is the honest measure of whether any of this is working, and it also tells you whether your metadata fields are earning their keep.
Everything above takes twenty minutes on a wedding once the picks exist. The reason it so often does not happen is that photographers reach the end of selection with nothing left in the tank, and the metadata sweep is the first thing dropped when the evening runs out.
That first pass is the part Kepla is being built to take off you. Kepla checks every photo from a shoot against clear criteria, sharp focus, open eyes, expression, near identical frames, and hands you a shortlist to approve. From a 2,000 photo wedding that is roughly 800 picks, in minutes. It never deletes, moves or renames a single file. Every original stays exactly where it is, and anything it did not pick is one tap away.
The relevant part for this sweep is what it does to your naming and metadata. Kepla does not rename anything, so the names you set at ingest are the names you keep, and nothing downstream has to reconcile two versions of a file. You approve the shortlist on your iPhone or iPad on the way home, open the same shoot on the Mac when a frame needs a proper look, and arrive at the metadata pass with the selection already made and some energy left to spend on it.
The picking app is still being built, for iPhone, iPad and Mac. The booking page is live today and free while we build, and if your backlog already has jobs sitting unnamed and unkeyworded, start with the triage order in clearing a photo backlog rather than at the oldest job.
After, with one exception. Renaming happens at ingest, because file names are identifiers and everything else refers to them. Keywords, copyright fields and captions should wait until you know which photos you are keeping. A wedding might produce several thousand raw frames and a few hundred delivered images, so metadata applied before selection costs several times what the same work costs afterward.
It is a saved set of fields applied to a whole selection in one action. Build it from the Metadata panel, then the Preset menu, then Edit Presets, and save it with Save Current Settings as New Preset. Put in only what is true for every frame of that kind of job: copyright notice, creator, website, contact and credit line. Keep separate presets for weddings, commercial work and personal projects.
Lightroom Classic's Export dialog has a Metadata section with a checkbox called Remove Location Info, which strips GPS data regardless of which include option you chose. The include menu itself offers Copyright Only, Copyright and Contact Info Only, All Except Camera Raw and Camera Info, and All. For anything public facing, especially family and newborn work, strip the location and check one exported file before you trust the preset.
Keywords are retrieval terms, short and repeated across many images, meant for filtering and search. A caption is a readable sentence with full names, a place and a date, meant for a person who will publish or file the photograph. An editor can use a caption directly. A list of tags is no substitute, and captions do not scale, so write them only for images heading somewhere public.
It is worth it. Lightroom Classic has a Catalog Settings option under the Metadata tab called Automatically Write Changes Into XMP. With it on, your fields are written into the file or its sidecar as you work, so another program or a rebuilt catalog can still read them. Without it, the work lives only in one database, which is a single point of failure for years of tagging.
Once a quarter, pick an old job and try to find one specific photograph using only your file names, folders and catalog fields. If it takes more than a few minutes, change the naming or keyword standard while the lesson is fresh. Retrieval is the only honest test. A system that looks tidy but cannot answer a real request is costing you time without paying anything back.
Kepla for Mac clears the obvious misses from a card, names the reason on every frame it sets aside, and leaves the choosing to you. Nothing is ever deleted, moved or renamed. Free through the private preview · the first hundred photographers keep it at $99 a year.