Say Cheese to’s New AI Shopping Buddy!

Who knew a picture could replace a thousand search terms?

Bas Jaburg
3 min readOct 28, 2023

If you are like me, getting inspiration from sites like Pinterest or trying to find something similar on (a Dutch platform where people sell their stuff), often leads to frustration and endless scrolling.


Because unless sellers are descriptive wizards, finding that exact piece becomes a needle-in-a-haystack situation.

So the strategy is as follows:

  1. You start with way too many results
  2. You narrow by distance
  3. You set a price range

“Aaah, thats more like it: only 155 results!”

“But hang on, …!”

A simple search like “modern couch” on can unleash a tidal wave of over 10,000 results, with the ideal piece potentially lurking just beyond our set filters.

Too many words also does not help

Why is this? The reason is simple: the results are solely and completely based on what the seller entered as a title and description. If he/she does not know how to create a proper description (or is just too busy), the item will not show up.

This makes matching random and the results often unpredictable.

No one likes this! The buyer, the seller and definitely not (or its advertisers).

So how about the following solution?

A picture says 1,000 words (or tags)

I have redesigned the matching process in such a way that…

  1. the buyer can perform a search
  2. the seller can create an ad …

all based on an image!

1. Searching by image

So instead of typing in a description, you upload an image. The system finds out what is you are looking for and show the tags and corresponding results.

You can even modify these details to get better results.

2. Placing a new ad: Let’s sell this room!

We could take this even a notch further: how about selling 7 items at the same time?

How about selling everything?

The system recognises these different items, so it gives the option to create one at the time, or all at once.

7 objects to sell

Then it just asks to create a nice and proper description of your object and Bob’s your uncle.

Since the description is made up of the tags it found in the image, finding a match for something like this should be much easier now.

Another benefit: multi-linguality.

Since is not only used by Dutch natives, this also solves an additional problem: how to post ads in another language. If you are French, should you post in French (or Dutch or English)? The same applies to searching.

If a ‘universal’ tagging system based on images would be implemented, this problem is resolved immediately.

So can this be done?

The short answer: Yes.

And to be clear: I have already built it!

In my previous article, I explained how to create a Chrome extension. That is exactly what I did!

Stay tuned!