Try AI Keywording in Lightroom Mode
Hierarchical keywords, IPTC fields, and XMP sidecars ready for catalog import
Try PhotoScanr FreeFree to use • No sign-up required • Instant results
Lightroom Keyword Strategy for 2026
By Duncan Rawlinson · Updated
A Lightroom catalog without a keyword strategy is a hard drive with thumbnails. You can scroll through it, but you cannot find anything. After ten years of shooting, that scroll becomes physically impossible.
A keyword strategy is the answer, and 2026 is a good year to revisit yours. Adobe split Lightroom into two products with very different keyword behavior. AI tools can now generate keywords that are good enough to be the first pass on tens of thousands of frames. Stock and editorial buyers continue to expect IPTC compliance. Smart catalogs and AI search overlap in ways that were not true two years ago.
This guide is the working keyword strategy that holds up across all of that. It assumes you care about being able to find your work in five years, not just this month.
Lightroom Classic vs Lightroom Cloud
The two products handle keywords differently, and any strategy has to start by acknowledging which one you are in.
Lightroom Classic
Classic supports a full hierarchical keyword tree. You can nest categories arbitrarily deep, control synonyms, mark keywords as private so they do not export, and define export options on a per keyword basis. The keyword list lives in the catalog, and it is the most powerful keyword system in any consumer photo tool.
Classic also writes keywords to the file or to an XMP sidecar, depending on the format. This is what makes Classic the right home for serious archive work.
Lightroom Cloud
The cloud version uses a flat keyword list with no hierarchy. It compensates with Adobe Sensei, which auto tags content based on what is in the image. The auto tags are useful for casual search but not visible as keywords in the conventional sense.
If you live in Lightroom Cloud and want IPTC keywords that travel with the file to other tools, you need to add them yourself. The cloud product is fine for personal libraries. It is a frustrating fit for stock or editorial workflows.
Most working photographers stay in Classic specifically for the keyword system. The rest of this guide assumes Classic, although the controlled vocabulary and AI ideas apply to either product.
Hierarchical Keywords Done Right
Hierarchical keywords are Classic's superpower. They let you tag a photo with one specific term and have all the parent terms apply automatically. Tag a photo with Kyoto and Lightroom can also export Travel and Japan, depending on how you have configured the parents.
The pipe character is how hierarchical keywords are expressed in plain text and in import files. Travel|Japan|Kyoto means Kyoto is a child of Japan, which is a child of Travel. When you import a keyword list with this syntax, Lightroom builds the tree for you.
A Sensible Top Level Structure
- Places: Continent, Country, Region, City, Specific Location
- People: Family, Friends, Clients, Public Figures
- Subjects: Animals, Architecture, Vehicles, Food, Sports, etc.
- Events: Weddings, Concerts, Sports, Personal events
- Concepts: Mood, Theme, Color, Season, Time of Day
- Production: Stock Ready, Portfolio, Client Delivered, Archive
The Production branch is the one most photographers skip and later regret. Tagging photos with their lifecycle status is what lets you find every photo you have ever delivered to a specific client, or every frame ready for stock submission.
Controlled Vocabularies
A controlled vocabulary is just a fixed list of approved terms that you commit to using. Without one, you end up with bird, birds, Bird, avian, and aves all describing the same thing, none of which surface in the same search.
You do not have to build a controlled vocabulary from scratch. The Controlled Vocabulary Keyword Catalog by David Riecks is the most established commercial option. The Lightroom Queen sells a starter hierarchy. For wildlife, the IOC bird list is the standard reference. For botanical work, the Royal Horticultural Society plant database is a common source. Pick one, import it as a starting tree, and build from there.
For AI generated keywords, a controlled vocabulary is even more important. The model will produce variations that are technically synonyms but not the same string. Standardizing on your vocabulary on the way in saves you from having to deduplicate later.
IPTC Keywords vs Catalog Keywords
This distinction trips up new Lightroom users, and it matters more in 2026 than it did before because AI tools work directly with IPTC.
Catalog Keywords
The Lightroom Classic catalog stores your full keyword tree, including hierarchy, synonyms, and export rules. This information lives in the catalog database. It is rich, and it is what you see in the Keyword List panel.
IPTC Keywords in the File
When you export a photo, Lightroom writes a flat list of keywords into the IPTC fields. Hierarchy is preserved as separate keyword entries based on your export settings. Synonyms are included if you have flagged them for export. Private keywords are stripped.
The IPTC keyword list is what other applications read. Stock platforms, Photo Mechanic, Bridge, and any web service that ingests your photos all see the IPTC keywords, not the Lightroom hierarchy.
XMP Sidecars
For RAW files, the safest place for keywords is an XMP sidecar that lives alongside the file. Lightroom can write keyword changes back to XMP automatically. This is what lets a future you, or a different application, recover all your work even if the catalog is lost.
For more on the file level metadata picture, see the photo metadata for backups guide.
AI Workflow With PhotoScanr
The realistic 2026 workflow is to let AI do the first pass and use your time on review and curation. PhotoScanr's Lightroom platform mode is built for this.
Style Preferences for Lightroom Output
Open style preferences and add a few rules. Tell PhotoScanr to use hierarchical keywords with pipe syntax. Tell it to follow your controlled vocabulary if you have one. Specify whether you want IPTC headline and caption populated, and what tone the caption should take. The model will respect these on every photo in the batch.
Batch Sizes
The Pro tier processes twenty five photos per batch up to one hundred per day. The Studio tier processes one hundred per batch up to six hundred per day. Studio is the older tier name renamed in version 1.22.0, with double the per day quota and double the batch size, which matches what most working photographers actually need.
ZIP Export
When the batch finishes, ZIP export gives you a single download that includes XMP sidecars or a CSV of metadata you can apply with a Lightroom plugin or directly with ExifTool. This is the cleanest way to bring AI keywords into the catalog without touching anything by hand.
Grounding for Identification
Turn grounding on for any work where specific identification matters. Wildlife, plants, products, landmarks. Grounding cross checks visual identification against external sources, which is the difference between a keyword that says "bird" and one that says "Cooper's Hawk." For more on this in a wildlife context, see the wildlife keywording guide.
When AI Is Wrong, and How to Override
AI keywording is fast and mostly right, but it is not infallible. The places it tends to go wrong are predictable, and a good workflow assumes review rather than blind trust.
Closely Related Species
Vision models sometimes confuse species that look alike. Cheetahs and leopards. Different warbler species. Ravens and crows. Spot check anything where misidentification would be embarrassing or wrong, especially if the photo is heading to a stock platform or a scientific archive.
Proper Nouns
Specific buildings, lesser known landmarks, regional dishes, and local cultural objects can be misidentified or generalized. Grounding helps a lot. So does setting style preferences with the location and trip context before you upload, which gives the model the right frame of reference.
People
PhotoScanr does not identify private individuals by name, by design. If you photograph the same family across years, you are still going to use Lightroom's People view or your own face recognition workflow for that. Treat AI keywords as the descriptive layer, not the personal one.
The override workflow is straightforward. Import the AI keywords. Skim through your collection in grid view at one second per photo. Anywhere a keyword looks wrong, fix it in Lightroom directly. Because Lightroom writes back to XMP, the fix sticks.
Catalog Maintenance
A keyword strategy is only as good as the maintenance habit behind it.
Quarterly Cleanup
Once a quarter, open the keyword list and look for orphan keywords with one or two photos, near duplicates, and inconsistent capitalization. Merge or delete as needed. The list stays clean, search stays useful.
Write to XMP on Import
Set Lightroom to write metadata changes to XMP automatically. The catalog is not a backup. The XMP sidecars are. If your catalog dies, your keywording survives.
Keep a Master Vocabulary File
Export your keyword list to a text file periodically. This is your controlled vocabulary in source form, and it is the file you import into a new catalog or share with a collaborator.
The Working Strategy
A 2026 Lightroom keyword strategy looks like this. Stay in Classic if you care about hierarchy and IPTC. Build or import a controlled vocabulary. Structure your keyword tree around Places, People, Subjects, Events, Concepts, and Production status. Use AI for the first pass with PhotoScanr in Lightroom mode, ZIP export, and style preferences that match your vocabulary. Review for known failure cases. Write to XMP on every change. Maintain quarterly.
That is it. The result is a catalog that finds anything in seconds, exports clean files for any platform, and survives software changes, hard drive failures, and the passage of time.
See pricing for the Pro and Studio tier limits, or open the homepage and try a small batch to see how the Lightroom output looks against your existing catalog.