OpenAi, makers of ChatGPT, have launched Shopping Research. Now you can have conversations with ChatGPT about suitable products for you to buy. Going beyond a traditional short search term such as “gifts for 4 year old girl”, you can now ask a more detailed question such as “please recommend a gift for a 4 year old girl who enjoys arts and crafts and her favourite colour is blue”.

And obviously from there, a whole conversation can follow about various other caveats or likes/dislikes. ChatGPT can’t know all the products in the world, but it can know more than you, and it can help save a lot of time in hunting for the right thing – or can it? I decided to put it to the test.

I read a study this week by YotPo and their findings mentioned the household name of Revlon not being surfaced by ChatGPT, and referenced their lack of structured data being a potential contributor to that. Well structured data is something I’ve been harking on about for months (years, but especially the last few months – what was a “nice to have” for SEO is now essential for GEO). So yesterday I broke down what it is and how to get it here. The story stood out to me because I used to do PR for Revlon – if you want a throw back to a 21 year old me dressed as She-ra you can see my Insta post about it here.

Real world test 1: ChatGPT Shopping Research for a bathroom vanity unit

Now, this might seem obscure, but when we were renovating our house, and building an ensuite, I needed to choose a new vanity unit… you know, a piece of furniture that your sink (basin) sits on. I wanted quite a wide one, but I only wanted 1 sink. I don’t need 2 sinks to clean – my husband and I are quite capable of doing our teeth over 1 sink at the same time.

Now, this was surprisingly hard to find. All of the vanity units of 120cm wide that I could find had holes for 2 sinks (because you need a hole in the top for the drainage to flow down). OR they had one sink to the side, and the hole was there – but I wanted 1 sink sat in the middle.

I searched for hours. And considering I had a poorly baby at the time it was quite stressful as the builders really needed me to make a decision and move things along.

Eventually I found what I was looking for, ordered it and loved it. And still love it. Our bathroom has even been featured on Forbes Home! (You can see it here if you like an interiors account on Insta.)

So, just because this has stuck in my head as a real world shopping problem (in a First World Problem sense), I thought I’d see – fast forward 4 years – if ChatGPT could have helped make it quicker.

I started with the prompt: “I’m looking for a wooden vanity unit for a bathroom, 120cm wide, but for only 1 sink. And the sink should be in the middle. Can you help me find one available to buy online?”

I then got a loading graphic of a word like “thinking…” (as usual) and then got this:

I clicked “Get started” and got this:

I clicked to “Preview and rate” and when I dismissed the first option (shown above on the left), it asked me why it wasn’t a good choice and I was able to type in my own reason – which was that I didn’t want an integrated basin.

However… it then showed me 10 more options, all of which had integrated basins. And any time it asked me why I didn’t like them, I gave them the same answer. Yet they kept coming. After 10 (bad) suggestions I got this screen:

I didn’t hold out much home to be honest, after 10 attempts of it seemingly ignoring my requests! But then it surprised my by coming back with this:

Hurrah! That was everything I’d asked for! I didn’t like it, but I hadn’t really talked about what I liked (if you see what I actually bought on Insta you’ll see it has drawers and doesn’t float – although floating furniture does make your room feel bigger as you can see more floor).

BUT there was a slight snag and that was the price tag. I think I paid around £300 for the one I finally bought, although I saw lots of expensive ones during my search. So I asked ChatGPT if it could find one for £500 or less…

I guess it struggled to find anything because it came back with…

BUT then it did! A range of options for under £500 – I won’t bother with a screenshot because the image didn’t load. And even a shelf for £140 it said could be a budget option. And then a comparison table (I’ve told you before, ChatGPT lovse a table!)

I didn’t love any of the options as much as I love the one I did get 4 years ago – but it certainly came up with more options that I did. The flaw in this test, is that we’re searching 4 years apart, and maybe there are more available now. Maybe I wasn’t alone in my searching back then. But as a research tool, it did it’s job well, even though I nearly gave up when it seemed like it was ignoring my requests. So they might want to tweak that part of the journey as people may drop out at that point.

Whoever controls the beginning of a shopping journey controls the money.

There’s been a saying in ecommerce since 2010 – “Whoever controls the beginning of a shopping journey controls the money.” And now OpenAI are certainly controlling a lot of the beginnings. But how are they going to monetise it? They can’t join an affiliate programme for every merchant on the planet, and they can’t just highlight people who pay them because users expect good advice, not sponsored advice. I’m going to delve into this more next week.