Reference Glossary
Keyword tagging
Attaching descriptive terms to an asset — either free-text or drawn from a controlled vocabulary — so it can be found through search, as distinct from structured fields like date or creator.
Why it matters in a DAM
In image- and video-heavy DAM libraries, keyword tags are often the primary way an asset actually gets found, because structured metadata fields can't describe visual content like 'beach,' 'product close-up,' or 'diverse team' the way a search query does. This is also where AI auto-tagging services (Google Vision, AWS Rekognition, Clarifai, and similar) have become common in DAM platforms, generating candidate keywords automatically from image content to reduce manual tagging load.
A worked example
Common mistake
Publishing AI-generated tags straight to the live index without human review. Generic or wrong tags — mislabeling a product's color, or tagging everything with broad terms like 'person' and 'object' — flood search results and bury the precise assets a user actually needs.
Keyword tagging is the retrieval layer that structured metadata alone often can’t provide. A field for ‘Date Created’ or ‘Photographer’ tells a search system exact, unambiguous facts about a file, but it says nothing about what’s actually depicted — a red dress, a mountain landscape, a specific product SKU shown in a lifestyle setting. Keywords fill that gap, whether entered by a cataloger from a controlled vocabulary (a fixed taxonomy the organization maintains) or contributed as free-text tags by whoever uploads the asset.
The trade-off between the two approaches is well established in information science: controlled vocabularies keep tagging consistent (everyone uses ‘automobile,’ not a mix of ‘car,’ ‘auto,’ and ‘vehicle’) but require maintenance and slow down tagging; free-text folksonomy tagging is faster and captures language users actually search with, but fragments over time as synonyms accumulate uncontrolled. Most enterprise DAMs use a hybrid — a controlled list for core categories, with free-text keywords allowed alongside it.
The more recent shift is AI-assisted tagging: image-recognition services can auto-generate a set of candidate keywords the moment an asset is ingested, dramatically reducing the manual burden on catalogers. The tradeoff is accuracy — computer vision models reliably identify generic categories (person, outdoor, vehicle) but routinely miss brand-specific or context-specific labels a human cataloger would catch (the exact product SKU, a company-specific campaign name, a misread studio background as ‘outdoor’). Most mature DAM workflows treat AI tags as a suggested first pass requiring human approval before they’re searchable, rather than publishing them directly.
Frequently asked
What is keyword tagging?
Attaching descriptive terms to an asset — either free-text or drawn from a controlled vocabulary — so it can be found through search, as distinct from structured fields like date or creator.
Why are keyword tags often the primary way assets get found in image and video libraries?
Because structured metadata fields can't describe visual content like 'beach,' 'product close-up,' or 'diverse team' the way a search query does — tags fill the descriptive gap that structured fields leave open.
What role do AI tools play in keyword tagging today?
AI auto-tagging services like Google Vision, AWS Rekognition, and Clarifai have become common in DAM platforms, generating candidate keywords automatically from image content to reduce manual tagging load.
What's the risk of publishing AI-generated tags without review?
Publishing them straight to the live index without human review lets generic or wrong tags — mislabeling a product's color, or tagging everything with broad terms like 'person' and 'object' — flood search results and bury the precise assets a user actually needs.
How does keyword tagging relate to a controlled vocabulary?
Tags can be either free-text or drawn from a controlled vocabulary — using a controlled vocabulary for tagging avoids the fragmentation that free-text tagging causes when different people describe the same thing with different words.
Should AI-generated tags go live automatically?
No — they should be reviewed by a human before being published to the live search index, since unreviewed generic or incorrect AI tags can degrade search quality for the entire library.