Generating CPG Product Descriptions with Chat GPT 4.0

Creative, Design, E-Commerce, Video

Oskar Kaszubski,  Chief Growth Officer at FirstMovr, and Dave Feinleib, CEO of It’sRapid, are analyzing product descriptions generated by Chat GPT 4.0 and making recommendations on fine tuning copy to meet retailer requirements. Tune in for this 3 part series to streamline your content execution strategy.


David Feinleib

Wonderful. So, now we’ve got the same thing, but we’re gonna try it with GPT 4, which is a newer model and a more sophisticated model. Same prompt that we’re sending into it. It’s thinking. And now here with this one you can see that it’s using some bolding, it’s using some formatting here, and it’s producing a new description for us. I like this experience, the delectable blend of savory, sweet, and spicy with every bite of Doritos, spicy sweet chili tortilla chips. So it’s got the bag size, it’s nicely formatted for us. How would you think about this formatting when it comes to a retailer like Walmart or Amazon? It looks great, it’s a lot easier to read, but how does this land on the retailer website?

Oskar Kaszubski

So again, think about it as the formatting has to be adjusted versus retail requirements. So some of it is, it’s gonna depend if a specific retailer can do a basic html where you can ba you know, bolt specific, you know, pieces of content. And remember when we generated this before, when we were just messing with it, also Chat GTP 4 generated icons, right? So they’re little emojis and basically included this in the feature bullet. So it’s interesting that it didn’t do it at the moment. And you know, we are doing this live just to try to see what we’re gonna do with this. So again, the creativity of it is great, but there has to be guardrail when it comes into what specific retailer requirements in terms of the content. And, and we’re gonna see the same thing, you know, when it comes to images. Yeah. But we actually sort out things like text on images because you know, some of the retailers don’t allow text overlay, some of them do. So we’ll have to get very nuanced ones in terms of how you’re writing those prompts.

David Feinleib

Perfect. Love it. Well, before we do that, one last stop here on our product description generation, I’m gonna ask Chat GPT to rewrite this for a summer barbecue. I wanna do something that’s timely for the summer. I like what I’ve got here, but I wanna see if I can create something that’s specific to a summer promotion. So here it’s got the product title for me, and I like this. Bring the sizzle to your summer barbecue. So it’s tying it into a summertime event. It’s bringing in the backyard barbecue into the copy, talking about beach outings here. So it does a nice job rewriting this, adjusting the content for this season, this event that I wanna talk about. Barbecue essentials, the summer snacking versatility. Thoughts on this?

Oskar Kaszubski

This is where I think Chat GPT actually shines into taking the content and putting in a new context, right? My suggestion for the bullets for brands is actually use the power of five star reviews and actually look at the reviews as a source of potential inspiration for bullets. Granted that it still has to be approved through the internal legal and regulatory team, depending on what category you are in. But there could be some interesting spins in the way people actually talk about your products that you could actually include. It’s a little bit tougher to pull this off for a lot of reasons. Number one is if you actually use Chat CPT 4.0 while browsing with Bing, then basically a lot of the time Amazon is trying to block you from reading the reviews.

Oskar Kaszubski

And even if it goes through, it will only scrape the first nine reviews on Amazon because Chat GPT 4.0 is not able to actually go in into the paginated reviews. So you can also only just basically see the first nine reviews, right? But if you actually use solutions like Profitero or any way you actually scraping the data from the retailer sites and you have a copy of your bull of your reviews, you could actually feed them in into the Chat GPT to create more compelling bullets and then talk with your regulatory or legal team to make sure that they can actually get approved. But you’d be surprised that sometimes how people talk about your product is not necessarily how you think they talk about it.

David Feinleib

Yeah, yeah. Love it. So a lot of different data sources you could use, you can incorporate shopper sentiment, review data, product data, all these different things. So I’m excited about this.


Check Out the Rest of the Series


Submit a Comment

Your email address will not be published. Required fields are marked *