No, you’re not gaslighting yourself. You know the “blue bottle on the beach" shot is there. You just saw it on Friday before you logged off. 

But the filename’s a string of camera digits and the person who exported it left in March. So you start the hunt: Drive, Slack, an old Dropbox subfolder, a message to whoever might remember.

Before you know it, you’ve just wasted 30 minutes looking for an image. Not to mention you’re this close to scheduling an appointment with your doctor to see about memory issues.

Air gets rid of that hunt (and the expensive doctor’s bill). Instead of limiting you to searching by filenames and manually applied tags, you can use Air to find an image based on the way you remember it.

Just a simple “blue bottle on the beach” and you have your sanity. But how does it work exactly?

In this article, you’ll learn about five ways you can search for an image on Air and how natural-language search pulls those signals together.

First, let’s set the record straight. Air’s AI image search for creative asset libraries isn’t the same thing as a reverse image search (like Google’s). 

Google, for example, takes an image and finds where it appears across the public web. AI image search finds assets you already own based on their visual content. Everything below refers to AI image search in a digital asset management platform (DAM), like Air.

Search runs against your workspace

Air analyzes and indexes the assets uploaded to your workspace. Results only ever return your own files. There's no web crawling and no external database. When you search, you're searching your library and nothing else. Search won't generate new images for you, either, because it only finds the ones your team has already uploaded.

So if an archive still lives in Dropbox, Google Drive, Box, or SharePoint, it won't show up in results — unless you import it into Air first.

That’s pretty easy to set up, though. If your creative files currently live in Google Drive, Dropbox, Box, or SharePoint, Air can connect to them and pull the files in directly. That means you don’t have to download everything to your computer and re-upload it. Think of it as a bulk import.

Candid was in that spot before Air, with assets split across Dropbox, Google Drive, servers, and hard drives. Just over a year after switching, their workspace held 90,000 assets, comprising 5.5 TB of storage, all in one searchable place. And their Lead Brand Designer went from spending up to 20% of her week finding assets for people to around 2%.

Find images in your library without tags or filenames

Air's Visual Recognition runs the moment a file uploads, generating Smart Tags automatically. Concepts, objects, and colors get indexed without anyone touching a metadata field. You can find images in your library without tags or filenames because the system already knows what's in each frame.

Most libraries only stay searchable as long as one person keeps them organized. The instant that person gets busy or leaves, there’s chaos. Either you tie down your creative team with mind-numbing admin work, or you risk your future selves’ wrath because the file you need next quarter will be buried under ‘Final_v3_REAL’. 

The Infatuation knows that trade-off well. Every new freelancer got a Google Doc of in-house naming conventions, yet files still landed in one sprawling Dropbox with names like “IMG10872.jpg.” Finding a specific shot meant spending hours manually sifting through files and scanning thumbnails.

After Air, they saved time on both ends: organizing content to begin with and finding it later.

“A platform like Air that puts visual assets first as opposed to traditional archival file management, for a person like me who thinks image-first, it allows us to be more creative, more effective, more efficient with the way we create content, and more powerful in the way we publish and distribute content.” — Nick Bilardelo VP, Marketing and Creative

Not everyone remembers the same shot the same way. One person thinks of it as a “blue bottle on the beach,” while someone else thinks “Malibu shoot with the ocean in the background.”

That’s the kind of different thinking you need on a creative team. But a normal digital asset management system doesn’t account for that. Air does.

Air reads five separate signals from every asset so you can search images by what’s in them, and each one maps to a different way people actually remember a shot. 

Those five signals are:

1. What’s in it

2. What it looks like

3.  What color it is

4. Who’s in it

5. What it says

Here's how each one works:

What's in it: Smart Tags and AI Summaries

Air’s Creative Intelligence generates Smart Tags for objects, concepts, and scenes when you upload them. For example, The Infatuation could upload a photograph of a burger and fries and automatically generate the tags: burger, fries, food, plate.

There are also AI Summaries. These give a short synopsis of longer assets, so you can skim a result set quickly instead of opening each file to check what it is. That speeds up the moment right after search, when you're scanning ten near-matches for the right one.

And for brand-specific needs, Custom Object Recognition can be trained to detect your own products or logos. That turns a proprietary SKU or a brand mark into a searchable object, so you can pull every shot featuring a specific item.

What it looks like: visual similarity search for brand assets

Sometimes the shot you want is described best by another shot. Drag a reference image into search and Air returns visually similar assets from across the workspace. This is visual similarity search for brand assets, and it's the fastest route when words fail you.

This is especially helpful when you have one approved hero image and you need every variant from the same shoot: the alternate crops, the other angles, the outtakes that didn't make the first cut. Instead of remembering which board or shoot folder holds them, you let the hero image do the searching.

Results cross folder and board boundaries, so a matching frame that got filed in the wrong place still surfaces. 

What color it is: filtering by palette

Color is often the fastest way to isolate assets from a campaign or season, especially when naming was inconsistent. Air's color filter lets you search by the tones an image contains, and the range is adjustable. You can pin it to one exact hex value or widen it to a family of nearby shades.

A strict filter finds the one asset in a precise brand red. A looser one gathers everything in the warm-red neighborhood, useful when a season's palette drifts across shoots.

Color pairs well with the other signals rather than working alone. Filter to product shots first, then narrow to the assets matching a seasonal campaign palette. Two clicks and you've isolated a slice of the library that would've taken a folder-by-folder crawl to assemble.

Who's in it: Facial Recognition

Air detects faces and lets you tag, filter, and organize content by the specific people appearing in it. If you’re managing talent, spokespeople, athletes, or executives, this can potentially save you hours every time.

But how do results stay reliable as the library grows into the tens of thousands of files? Face groups are admin-validated. This means an admin confirms and names each person. You're not trusting the machine to guess who someone is. 

One point of setup to know is that Facial Recognition is enabled with Air’s team rather than switched on by default, and the same goes for smart search and visual search if you can’t see them in your workspace yet. Video Facial Recognition, for finding video of specific people, is coming soon (at the time of writing).

What it says: text inside the image

Optical Character Recognition (OCR) indexes text inside images, PDFs, and videos, and makes on-image copy searchable. If words are printed, stamped, or burned into an asset, you can find the asset by those words.

For example, you can type a phrase printed on a product label and find the packaging shot. Or search the headline burned into a social graphic and pull the exact asset. No one had to transcribe that text into a tag first. 

The same approach works for documents and video: 

• Document Intelligence indexes full slide decks and PDFs, so a phrase buried on slide 14 is searchable.

• Video Intelligence makes spoken words searchable through time-synced transcripts, supporting 100-plus transcript languages.

The technology that makes Air find the shot you want when you type “the one with the bottle by the water, sunny, from last summer” is natural-language search. 

 You type a phrase describing the shot you remember, then tighten the result set with filters rather than rephrasing the query over and over.

Here’s what goes into that:

Type the phrase and refine with filters

Conversational Search accepts plain-English queries like "surfer photo at sunset" instead of requiring an exact keyword match. Behind that phrase, Air compares the meaning of what you typed with what it has already read from each file, like its Smart Tags and any text inside the image, so you can describe what you remember instead of guessing a filename.

Just know that this isn’t the same as having a back-and-forth chat with an assistant that asks follow-up questions.

Instead, you narrow the results step by step. 

Progressive refinement looks like this in practice:

1. Start with a description: "model holding product outdoors."

2. Stack a Type filter to show only images, not videos or documents.

3. Add a color filter for the campaign's palette.

4. Add a person filter for the specific talent.

5. Add a metadata filter for the region or channel you're working on.

Each step shrinks the set until you're looking at a handful of assets instead of hundreds. One note on scope: some metadata filters, such as Camera, Creator, Copyright, Imported Keywords, and Date taken, are only available on Air’s Enterprise plan.

Results that carry approval and version context

Finding an image is only half the job. Finding the image you're actually cleared to use is the other half. That's why Air surfaces the context behind each asset too.

That context lives in Custom Fields. You can store values like these against each asset:

• Campaign

• Channel

• Region

• Product line

• Owner

• Usage rights

• Approval status

Those values are visible and filterable right in results. So you can filter to "approved, US, spring campaign" and trust that what comes back is safe to ship.

Version Stacking keeps the result set clean. New iterations stack on the original, so search returns one asset with its current version on top instead of five near-identical files fighting for attention.

Visual Annotations sit with each asset as the record of creative decisions, the pinned and timestamped feedback that explains why a version looks the way it does. This makes it easier to find the asset you can use.

Search on shared boards

Board share links are searchable too. An external partner or stakeholder can find assets within what they've been given access to.

For privacy reasons, this also means the filter set available on a shared link is narrower than in the workspace. It depends on the link’s permissions. A collaborator can find things, but not with every filter an admin has inside the workspace.

Secure Share Links also carry password protection, expiration dates, and email-gating. Making a board searchable doesn't loosen who can see it or for how long.

Results that carry approval and version context

Finding an image is only half the job. Finding the image you're actually cleared to use is the other half. That's why Air surfaces the context behind each asset too.

That context lives in Custom Fields. You can store values like these against each asset:

• Campaign

• Channel

• Region

• Product line

• Owner

• Usage rights

• Approval status

Those values are visible and filterable right in results. So you can filter to "approved, US, spring campaign" and trust that what comes back is safe to ship.

Version Stacking keeps the result set clean. New iterations stack on the original, so search returns one asset with its current version on top instead of five near-identical files fighting for attention.

Visual Annotations sit with each asset as the record of creative decisions, the pinned and timestamped feedback that explains why a version looks the way it does. This makes it easier to find the asset you can use.

Search on shared boards

Board share links are searchable too. An external partner or stakeholder can find assets within what they've been given access to.

For privacy reasons, this also means the filter set available on a shared link is narrower than in the workspace. It depends on the link’s permissions. A collaborator can find things, but not with every filter an admin has inside the workspace.

Secure Share Links also carry password protection, expiration dates, and email-gating. Making a board searchable doesn't loosen who can see it or for how long.

In most libraries, search and organization are two separate jobs. Someone tags and files things (organization), and then you hunt through the result (search). In Air, the same AI analysis that makes an asset findable also organizes the library around it. 

Every asset becomes searchable the moment you upload it. Air also applies smart metadata automatically so you don’t have to waste time tagging manually.

From there, you can also adapt an asset without leaving Air, which is where a DAM with AI goes beyond a traditional DAM. Adapt or reframe it in Canvas, or run one of your team’s Skills on it, keeping the brand kit and the full version history right beside it. Or use simple prompts to fine-tune lighting or color, extend the image to a new size, or upscale it for sharper detail. 

If your archive still lives across Drive and Dropbox, start an Air workspace, import it with the native cloud tools, and run a search against it. The first "we definitely have that somewhere" you resolve in ten seconds tends to make the case on its own.

Is AI image search in Air the same as reverse image search?

No. Reverse image search, like Google Lens, finds where an image appears across the public web. AI image search in Air runs against your own workspace and returns only assets you've uploaded, matched on what they contain rather than where they appear online.

Do I need to tag assets manually for AI image search to work?

No. Air's Visual Recognition runs on upload and generates Smart Tags automatically for objects, concepts, and colors. 

Can I search my library by dragging in a reference image?

Yes. Drag a reference image into search and Air returns visually similar assets from across your workspace. It matches composition and visual content rather than filenames, so results cross folder and board boundaries to surface every variant of a shot.

How does Facial Recognition get set up in Air?

Facial Recognition is part of Air’s AI Features, so you get it switched on by contacting your customer success manager or Air support. Once it’s on, Air detects faces and suggests who they are, and workspace admins validate those suggestions by confirming and naming each person. That admin step keeps people search reliable as the library grows.

Can I search for text that appears inside an image or PDF?

Yes. OCR indexes text inside images, PDFs, and videos, so on-image copy is searchable. You can find a packaging shot by a phrase printed on the label or a social asset by the headline burned into it.

Are assets on a shared board link searchable for external collaborators?

Yes, board share links are searchable, so partners can find assets within what they've been given access to. The available filters are narrower than in the full workspace and depend on the link's permissions, while password, expiration, and email-gating controls keep distribution locked down.