Your photo library is invisible (and AI can fix that)

Go and open your company’s photo folder. Search for “kitchen with an island” or “the team at last year’s conference”.

We’ll wait.

Nothing, right? Because the photos are called IMG_4831.jpg and DSC_0092.jpg, and the only search engine that works on them is whoever took them, if they still work there.

We genuinely think the photo library is the most valuable thing most businesses own that nobody can use.

Why we care about this

Partly because we’ve watched people lose hours to it. Marketing needs “a bright lounge, landscape, no people in it” and someone ends up scrolling through four thousand thumbnails with a cup of tea going cold.

Partly because the best photos never get used. If nobody can find the great shot, the okay one ends up on the brochure.

And partly because it hurts the website. Images without proper descriptions and alt text are invisible to Google Images and to AI search tools. They’re also a problem for anyone using a screen reader, which is something we should all care about more than we do.

What the AI does

The vision models available now are properly good at looking at a picture and saying what’s in it. What kind of room, what materials, what the light’s like, indoors or out, whether there are people, even the season. We use that to tag every image, write useful alt text, and make the whole library searchable. So you type “pool, sunset, no people” and that’s what you get.

The photo library is the most valuable thing most businesses own that nobody can use.

The fixed list is the secret

If you let AI describe photos however it likes, it’ll call one a “lounge”, the next a “living room” and the third a “sitting area”. Which is fine for a person reading it, and hopeless for searching. So we sit down with the client first and agree a fixed list of tags in their own words, and the AI has to choose from that list. That consistency is what makes the search actually work.

A person still gets the last word

The AI tags thousands of images in a few hours. Then someone on the client’s team runs through them on a review screen we built for speed, approving or fixing tags in batches. The AI does the mind-numbing part. People do the judgement.