Generative AI has done some amazing things in the last couple of years. And there’s a lot that it can help you do in marketing your products on Amazon. But should you use AI to optimize your listings?
The answer is a little more complicated than a yes or no.
The short answer is that generative AI can be useful in building listings. But the better at optimizing your listings without AI you are, the better you’ll be at optimizing your listings with AI.
And that’s primarily because AI is useful as a tool, not a solution. Let me explain what I mean.
Understanding AI
The important thing to understand Artificial Intelligence is that it doesn’t come up with information out of nowhere. AI is trained on data using a model that allows you to access that information using a query.
We call this Natural Language Processing (NLP). This allows you to use normal language to access the data. So when you tell Craiyon “draw me a picture of a fish holding a hammer”, you get images like this.
That drawing is based on images that already exist. Craiyon is just accessing widely available images to give you what it thinks you’re asking for. But the information already existed.
Similarly, if you use writing AI, like ChatGPT, you’re just accessing information that already exists in the dataset. You’re not creating anything new–you’re just rearranging things.
Out-Dated AI
Most AI is trained on a dataset from the past. For example, the GPT 4 model (the most recent OpenAI writing model that a lot of tools are built with) was trained with a dataset from September 2021.
This means that any information you get from tools that use the GPT 4 is at least that old.
There’s not an effective model yet that is recursive (adds data to the dataset it pulls from). So all you’re doing is pulling and manipulating data from September 2021 and before.
That’s not to say that writing AI is useless. But it might change how you use it.
Image Limitations
But let’s say that you’re not interested in using the writing AI tools. What if you’re using image-based generative AI?
Well, there are limitations with that too.
For example, you’re still going to want to have actual product photos when possible. Real product photos convert better than renders or generated images. They’re also more accurate, which leads to a better experience.
So that more or less rules out using AI for generating photography (both studio photos and in-use photos).
But what about infographics?
You could take a cut-out photo of the product, put it on an interesting background and then just add informative text, right?
Well, sort of.
The best tool we’ve found for something like this is CreatorKit, which does this exact thing. Except, it doesn’t add the text callouts. You would still have to do that part on your own. But you could use CreatorKit to replace the blank background with an engaging one.
And that would totally work. You get a few of these images, add some callouts, features, or benefits, and you have a pretty great infographic.
But then again, you could also get some stock photography from Adobe Shutterstock, or some other stock imagery marketplace and do the same thing.
The only difference is that, although it may be a little more challenging or costly, the stock photography route would look better because the photos are, well, real. And real images are always going to convert better than generated ones.
Video Limitations
The same things are pretty true for video as well. However, you might be able to leverage an AI video editor (CreatorKit actually has one) to do some effective work.
These video editors take the footage you have and the outline you give it and compiles a decent video.
With the right clips and script, you could probably make a decent Amazon Video Ad that converts customers.
But that’s probably the full extent of what you’ll be able to do effectively with AI. You’ll still need to film the clips used, you’ll still need an effective script, and you’ll need a general idea of what you want the video to look like.
Making AI Effective
These limitations (with generative written and visual AI) are common across all the AI tools you’ll find. They might be good for some things, but for other things, they fall laughably short.
So what, just give up on AI altogether?
Of course not. What these AI tools are good or not good at is just part of the story. Because there are things that AI is really good at. Just not everything. And you should leverage what AI is good at to augment what you do that AI is not good at.
This means that knowing how to use the tool is just as important as the tool itself. The best way to use AI is to pair it with human work.
And that’s because humans and computers are good at fundamentally different things.
You could use a keyword generator to produce a list of keywords, feed those keywords to an AI writer, and then put those bullet points on your Amazon listing–but you and I both know that listing probably wouldn’t perform very well.
That’s because you know your customers better than the AI does. You know your brand better than an algorithm. You know your product better than a computer. And no tool is going to change that.
BUT! If you’ve ever made an Amazon listing before, you know there’s a tedious process to try and come up with effective bullet points, product descriptions, and callouts for your listings.
I’ve personally been writing Amazon listing content for years, but I still get “blank page syndrome”.
Writing listings all alone can be difficult and awkward. And it can take an absolutely stupid amount of time to come up with anything good.
So relying just on AI is a bad idea, and writing it all yourself is slow and boring.
The solution is something in the middle. Something that leverages the strengths of both artificial and human intelligence.
Pairing Humans and AI
Sorry for the long lead up. Here’s what we recommend for using AI to optimize your Amazon listings: Use AI as a jumping off point for listing optimization. Then, edit and refine on your own for the best results.
There are a number of ways you can do this, both when initially creating a listing as well as refining a listing over time. Let’s go over a couple of examples here.
Titles, Bullet Points, Product Descriptions, A+ Content
There are a number of tools you can use for producing written listing content with AI. ZonGuru and JungleScout both have wonderful and beautiful tools, but our favorite so far is Perci because of how involved you are in the process of keyword utilization. You could even just use ChatGPT if you’re willing to feed it all of the parameters and do a little bit of the grunt work yourself.
But that kind of defeats the purpose.
Regardless of the tool you use, here’s the process:
- Find a bank of keywords (all three tools mentioned above have them) that you want to leverage in titles, bullet points, product descriptions, and written A+ Content.
- Have the tool generate the first draft of the written content. All of the tools can do this pretty quickly and intuitively.
- Read through the content, figure out what makes sense to keep and what makes sense to change.
- Generate more drafts with the AI as necessary and continue to revise.
This might sound like a tedious back and forth, but it actually goes pretty smoothly. You could run through 5-10 title iterations in about 5-10 minutes. It goes pretty quickly, and because all the tools are trained on the GPT-4 model, you can just have it make iteration after iteration.
As a result, you end up with keyword rich content that doesn’t sound clunky or awkward. And you got there by leverage the best parts of AI and human intelligence.
Images
This process is also true for generating images.
- Leverage CreatorKit and other generative imagery tools to get an idea of the aesthetic you want.
- Then give your AI mockups to an actual photographer or graphic designer. Let them do the final renditions.
The most expensive part of any creative process is the iteration process. It’s difficult for non-creative people and creative people to have clear communication regarding what the project outcomes should be.
You think you’ve really explained to the graphic designer what you’re going for.
But the infographics are nothing like you had hoped.
So you have them re-do them. Which is frustrating for them and expensive for you.
But if you can create a bunch of mockups (maybe even with your product in it), you can say “I like the colors from image 1, the background of image 2, and the water effect in image 3. Can you make something like that using the following callouts?” (Of course, you used AI to generate the callouts–so they’re really good.)
Trust me, the graphic designer will thank you–and so will your bank account.
Video
Okay, I’m going to level with you here–it probably won’t be helpful to have AI generate video mockups for you. Even with Stable Diffusion, it’s hard to get good video for product marketing. But you’re welcome to give it a try.
Instead, if you’re low on budget, we recommend shooting product video yourself (you would be surprised what you can do with a phone), and then using an AI video editor to create a video ad with lots of block text. With a little bit of effort, you can create a pretty high converting ad.
With the sales from that ad, you might be able to hire a professional to do video for you.
AI Video is the hardest kind of AI to do well. So even one day when we can have all of our images and content created by AI (and have it turn out better than what we could have done ourselves), odds are we’ll still be a long way off from creating high quality videos with AI alone.
A/B Testing Written Content
Amazon has an A/B testing tool (Manage Your Experiments) that allows you to split test titles and other attributes of a listing. This is probably the best use-case for AI.
Head back over to your AI listing creation tool and churn out a few titles that you think look really, really good. Then, run an experiment with one of them on Amazon.
After you have enough data, keep the best converting title, then do it again.
You can do this for titles, bullets, and A+ content–all of which AI can help out with.
The Importance of an Expert
The biggest downside to AI is that it doesn’t know when it produces crap. So you could use AI and have a really awful listing. This means that no matter what you do with AI–it’s not a replacement for actually knowing what you’re doing.
So it’s important you hire an expert–or at least do as much of your own research as possible. Amazon’s a big place with a lot of things to learn and know. It’s almost impossible to do it on your own.
Fortunately, this is where BLAZON can really help. We’re a little bit biased, for sure. But we think we’re pretty good at this Amazon thing. We know a good listing from a bad one. And we think we could help you out.
Reach out–even if you just want us to look at your listings and give you our thoughts. We’d be happy to help.
Because any bloke can use AI. But it takes an expert to use it well.