Where to find them
Open a deck and go to the Images tab. The Relevance and Date buttons sit next to Upload Images, directly above the search field. Relevance is selected by default. Switching between them re-runs your current search — you don't need to retype it.
Date: name and tag matching
Date is the original search behavior. It looks at exactly two fields:
The image name
The tags applied to the image
Your query is matched literally against those fields. If the text doesn't appear in the name or a tag, the image isn't returned. Results that do match are ordered by upload date, newest first.
Because matching is literal, the words you use have to be the words that were entered. A search for "crowd" will not return an image tagged "fans," and a search for "arena" will not return one named stadium_exterior.jpg. Images with generic file names and no tags — DSC_04821.jpg, IMG_2291.jpg — are not reachable by content in this mode.
Relevance: weighted, multi-signal scoring
Relevance evaluates your query against several sources of information about each image rather than one. Each source is weighted according to how strong a signal it is, and the weighted results combine into a single relevance score. Images are then ordered by that score.
Image metadata
The image name and the file details captured at upload. This is the same information Date search uses. It still counts toward the score — it is one input among several rather than the whole test.
Tags
Tags your team has applied. These are a deliberate human signal: someone decided the image is a "hero shot" or belongs to "2026 Season." Tags carry meaningful weight in the score, and they're where team-specific vocabulary — campaign names, seasons, approval status — lives. No other signal covers that.
AI descriptions
A generated description of what is depicted in the image: subjects, setting, action, mood, dominant colors, and visible on-image text. This signal covers content that was never typed into a name or tag field, which is what makes an untagged IMG_2291.jpg reachable through a query like "empty theater seats from the back row."
How the weighting plays out
Because the signals are weighted rather than applied as pass/fail filters, a strong match on one signal can rank above a weak match on another, and an image matching on several signals at once ranks above one matching on a single signal. The result is a ranked list rather than a set of literal matches — every query returns something, ordered by how closely it scored.
Side by side
| Relevance | Date |
Looks at | Image metadata, tags, and AI descriptions | Image name and tags |
Matching | Weighted score across all signals | Literal text match |
Result order | Highest relevance score first | Most recently uploaded first |
Untagged, generically named images | Reachable through their description | Not reachable by content |
Query style that fits | A description of the image | A known name or tag |
Which to use when
Relevance fits when you know what the image looks like but not what it's called — "fans celebrating in the stands," "aerial view of the ballpark at sunset," "player portrait on a dark background."
Date fits when chronology or an exact string is what you're after:
You just uploaded a batch and want to see it at the top of the grid.
You're reviewing what's been added to the library recently.
You know the exact file name or tag and want only literal matches.
Writing queries for each mode
Mode | Query | What happens |
Relevance | wide aerial view of the stadium at sunset | Descriptive phrasing gives the scoring more signals to weigh. |
Relevance | team logo on a white background | Composition and color are part of the description signal. |
Date | headshot2026 | Returns images whose name or tag contains that exact string. |
Date | 2026 Season | Returns images carrying that tag, newest first. |
In Relevance, a phrase of roughly three to eight descriptive words gives the scoring the most to work with. Including the setting and mood — "indoor," "night," "crowded," "dark background" — and any text visible in the image adds signals it can weigh. Tagging remains worth doing in both modes: it's a weighted input to Relevance and the only searchable field besides the name in Date.
FAQ
Do I need to re-tag or rename anything?
No. Both modes work with your library as it stands. Existing names and tags continue to count in Relevance scoring.
Will Relevance return everything Date would have?
Relevance considers the name and tag fields Date uses, so literal matches still surface — though they appear in score order rather than date order. Use Date when you want strict matching and chronological ordering.
My Relevance search returned images that don't look related. Why?
Relevance returns a ranked list rather than cutting off at an exact-match threshold, so lower-scoring images appear further down. Adding more descriptive words changes the ranking.
Does this apply to files, videos, and audio?
The Relevance and Date toggle applies to the Images tab.
Can I still search by tag?
Yes. Tags are searchable in both modes — as a literal match in Date, and as a weighted signal in Relevance.
Still can't find what you're looking for? Reach out to your DIGIDECK account team and we'll help track it down.