How To Use ChatGPT For CRO Purposes

The buzz around AI didn’t start in 2023. Rather, it can be traced back 72 years ago.

Let me take you on a trip down memory lane…

In 1951, the first movie about AI, titled “The Day the Earth Stood Still,” was released. It was based on Harry Bates’ book “Farewell to the Master,” and featured an AI character named Gort. This movie sparked an interest in AI among science fiction and movie enthusiasts worldwide.

Since then, more movies featuring AI have been released, capturing the interest of movie and science fiction lovers worldwide.

Today, hundreds of AI tools are being introduced on a daily basis. Small and big companies have started using AI to improve their efficiency/productivity. 

The wave of AI has also had an impact on the field of conversion rate optimization. 

In order to provide insight into the utilization of AI tools such as ChatGPT for CRO, I consulted with three experts in the industry:

• Johann Van Tonder CEO AWA Digital.
• Khalid Saleh, CEO of InvespCRO.
• Shawn David Systems Engineer.

In this article, you will discover the specifics of ChatGPT, its features, and how it can be utilized for CRO purposes. 

By the way, we have a podcast version of how you can use ChatGPT for CRO. Take a listen to the CRO Live Hour podcast…

With that said, let’s get started.

What Is A Large Language Model?

A large language model (LLM) is a deep learning algorithm capable of recognizing, summarizing, translating, predicting, and generating text and other content by leveraging knowledge acquired from vast datasets.

These models, which rely on transformer architectures, have proven to be highly effective in a wide range of applications beyond language learning, including protein analysis and software development.

Apart from enhancing natural language processing tasks such as chatbots, translation, and AI assistants, LLMs have been deployed in various industries, including healthcare, software development, and many others.

What Is ChatGPT?

ChatGPT is an AI tool built on a  large language model (LLM) developed by a team of researchers at OpenAI and has the capability to summarize text, generate text, allows you to have human-like conversations, and much more with the chatbot.

ChatGPT can answer questions and assist you with tasks like composing emails, essays, and code.

It launched on November 30th, 2022, and 5 days later, it had its first 1 million users, making it the fastest-growing app in the world.

In 2 months, it hit 100 million users.

Below is a graphic of how many months it took some of the popular apps to get to 100 million users.

Features Of ChatGPT.

In order to know how to use ChatGPT, it is important to have a comprehensive understanding of its features. Here are the features of ChatGPT: 

1. Advanced Reasoning.

Since ChatGPT is built on a large language model that has been trained with a lot of information, it can follow complex instructions in natural language and solve difficult problems with accuracy.

The amazing thing here is GPT4, the newer model, does this better than GPT3 and 3.5.

2. Creative Collaborator.

The newer model GPT4 is more creative and collaborative than the older models 3 and 3.5.

It can generate;

• Songs, screenplays, or learn a particular style.

Here’s an example below where ChatGPT rewrites a popular manga;

3. Accepts Visual Inputs.

GPT4, unlike the other models, accepts visual input and can generate captions and classifications and even analyze these inputs for you.

See this example;

The input was – what can I make with these ingredients?

Note: This particular feature isn’t publicly available yet. It’s something they’re working on.

4. Analyzes and Produces Longer Text.

Older models of GPT couldn’t handle much text (as prompts) or analyze a large block of text. You’d usually see this error message.

Now, GPT4 has the ability to handle longer texts (over 25,000 words of text, allowing for use cases like long-form content creation, extended conversations, and document search and analysis).

See what Stefan Georgi (#1 copywriter in the world) had to say about how he and his team make use of longer prompts in ChatGPT;

See an example of a long text analysis here;

Will AI Replace Conversion Optimization Specialists?

I chatted with experts who are neck deep in the weeds with CRO and AI to know their thoughts about AI replacing CRO roles.

Khalid Saleh, CEO of InvespCRO, put forth this perspective;

‘ChatGPT replacing conversion optimization specialists depends on the type of CRO you’re dealing with. The less sophisticated you’re; eventually AI and automations are going to replace you. The more sophisticated you are, the longer the road ahead for AI to replace you.

Take, for instance, copywriting. Copywriting is an amazing gift; some have it, others do not, and yet ChatGPT can help with marketing copy, and most times, it’s easy to tell when a copy is written by a human or AI.

Now, the top copywriters aren’t scared that AI will replace them; rather, they see AI tools as a way to get their jobs done faster and make them better at the same time, 70% of copywriters might be replaced. This ties back into my statement that it depends on the type of CRO specialist and their level of sophistication.’

Shawn David, a Systems Engineer, adds an interesting angle;

‘Any sort of data manipulation, correlation, and classification job will be replaced by a specific or general AI. How fast is this going to happen? No one will be able to tell you that, and that’s because for 20 years, as humans, we’ve been working on general AI. We’ve only gotten to the point where now the AI can iterate on its algorithm. So it’s not a question of if AI will replace CRO experts because it already has. With Google Optimize being sunset, if you don’t have the ability to pay for a better tool or even hire a CRO agency or specialist, just giving AutoGPT access to the internet, it can already be a CRO (tho not a good one, but better than someone that doesn’t know what they’re doing and that’s because AutoGPT can already do data manipulation and correlation.

Take the instance of a user test, you get 10 videos back, and the CRO specialist has to sit down, watch each video, transcribe, and pull out themes. GPT does that now, and there’s absolutely no need for a human to do that anymore. 

It’s not a question of f AI will replace CRO specialists but if you want it to. My answer is absolutely not, and my reasoning is that best practices are not best practices when it comes to this, and that’s all you’re going to get. What you’re going to get is what’s happened in the previous, and that’s because it’s not general AI.

 So at first, when you spin this up, all that will happen is that AI  will look at the past. And it will say, All right, what can I find before 2021? Or if you hook it up to the Internet, what can I find right now? Great. That doesn’t help. That doesn’t help, right? Like, that’s fine. That’s a great system. And it’s an amazing magic trick. And it gives the appearance that you’re doing something. But you’re not. All you’re doing is tagging what previously worked in different places and hoping it works indoors. Still, you’re not going to be able to understand like your customer, and you’re not going to get the voice of customer, you’re not going to get pain points, you’re not going to get any of that. This is how AI is going to test it without a hypothesis. Suppose there’s a carousel; remove the carousel, you’re gonna get that type of response. AI doesn’t take into consideration that people are in a consideration set and that you need to convince them, maybe it’s some product you’re selling, that they need a sales page for that you need copy that will convince them. The robot doesn’t have the context of being able to pull in what you’re looking at. They only have what happened in the past and what you feed it in. You’re not going to get good CRO from the bot unless you’ve done voice of customer research and you figured all this stuff out, and then you can feed it and now it has the content. Oh, yeah. That’s a different story. Yeah. But if you already have that, then you don’t need to chat GPT because you already have a system that works.’

Johann Van Tonder brings a different perspective;

‘The answer to this question isn’t binary. AI can do certain things better than humans, but there are other things humans are better at than AI, and I don’t see that changing, certainly not fast.

Back in the early days of eCommerce, one of the refrains we often repeat in conversion optimization is that people buy from people they don’t buy from websites. That fundamental truth applies to AI as well. You know, people want to have human interaction. So I don’t think AI will completely overtake everything. But you’ve got to look at the ability of AI to do certain things a lot better than others. And so I think it’s different for each role. So within CRO, there are multiple roles, right? So there’s a strategist, there is an analyst there, the dev, the data analyst. But within each of those roles, certain tasks can be done better by AI. And so I think it’s not a question of whether AI will replace a role. It’s how much of a role will be replaced by AI. And I’ve made this controversial statement on LinkedIn, which I stick by. And it is, if you’re not willing to do one of the following two things, you’ll be out of a job. And those two things are work with AI or work for AI. And work with AI, I think is that’s really the thing to talk about.’

During the conversation with the experts, I asked how folks working in conversion optimization can use AI to stay longer in their roles.

Johann put forth this perspective;

‘ The best that all of us can do is to embrace it and to find ways to use it in your job and in your life. That’s easier said than done. I told my team to go away and figure out how they can disrupt their lives and how they can disrupt their jobs with AI and report back to me. That’s easier said than done. That’s an easy thing to say to somebody. But that the best answer I have is. I don’t know what the future looks like, and nobody knows what the future looks like, not even the people who are developing open AI. All of them can’t predict what it looks like. When nobody knows, nobody can give you the perfect answer to that.  The best thing we can do right now is to embrace AI into our lives to see how we can use it, to play with it and get stuff wrong, and, you know, mess it up and figure things out. And that’s the best you can do right now.

Shawn answers directly;

The way that I use GPT, and I think it’s amazing, is I imagine it’s like Jarvis from Ironman, so Ironman in and of itself, he’s an intelligent, brilliant man. But he has this AI that does the monkey work for him. It’s like, if you’re in Excel, for instance, and you have a bunch of stuff that you have to do, you have a couple of pivot tables, and V lookups, and H lookups and a bunch of stuff that connects. Don’t do that; give it to ChatGPT and tell it to do it. And then you’re done. If you have a couple of like concatenation, where you have a database, and you have to like build emails off of things, or you have to write A/B test formulas and similar formulas as a CRO, or you want to learn how to scrape the internet and find people on LinkedIn well go to GPT, ask it, how to create a Python script, how to install Python, and it will walk you through these things.’

Unique Ways CRO Specialists Are Making Use Of ChatGPT

ChatGPT has some features that can come in handy for a CRO specialist and any team member, and we look at each feature separately in this section.

1. It can act as a copywriter.

There’s almost no page on a website where you won’t find copy. Copy plays a crucial part in conversions from the landing page to the product pages. To maximize the use of ChatGPT here, you can have it write product descriptions, sections of a landing page, and anywhere text is needed.

Here’s an example of ChatGPT writing a production description for a fictitious eCommerce website that deals in vintage candles.

Pro Tip: The more detailed your prompt, the better the output.

Here’s another example where ChatGPT writes the headline for a Black Friday landing page offer;

See another example;

2. It can help with Hypothesis creation:

A hypothesis is a crucial step in any A/B testing process. Skipping this step can set the test up for failure.

The thing is, before hypothesis creation, you need to identify a problem (conversion issue) that you’re creating the hypothesis for. 

How does ChatGPT come in? It’s simple.

When you give it the right inputs, it gives you different hypotheses variations.

Khalid gives a simple prompt to get this done easily;

‘Hey ChatGPT, here’s the problem I’m trying to solve; here are some things I’ve learned about my site visitors,  here are some observations about my site visitors from previous ab testing, and here’s an initial hypothesis, and you let ChatGPT give you some other variations of your hypothesis.’

Note: the line where the prompt says ‘problem I’m trying to solve’ list out the issue.

Here’s an example;

Khalid adds a note of caution;

‘The hypothesis ChatGPT shares with you is only as good as the input (details) you share with it, so you want to make sure your prompts are as clear as possible.’

3. It can code A/B test scripts.

Before an A/B test is launched, the variation(approved design) needs to be coded. Now, in an agency or in-house CRO team, many tests are being prepped simultaneously, so some tests might be delayed in launching, except the team has a lot of front-end developers, but that’s not always true the case. 

Here’s how ChatGPT can help;

You feed ChatGPT code direction and what you’re trying to put together, and it writes it out. All the developer needs to do is come in and clean it up before the test is launched.

Khalid shares his thoughts;

‘ChatGPT has programming languages written into it, and most A/B tests use front-end code to manipulate a webpage. As a conversion specialist, you can have ChatGPT write most of your A/ B test code. Why most and not all? It’s because ChatGPT isn’t 100% accurate. It can code up 80% of your tests, which is a win.’

Here’s an example;

Here’s the continuation of the code;

4. It can act as a QA specialist.

Quality assurance folks are those guys that review A/B test codes and ensure they look good and work right on several devices; also, they ensure there are no issues in the testing environment before a test is launched.

Considering the same scenario where a team is occupied with many tasks, ChatGPT can come in and help review AB test codes.

See this example;

5. It can help with design concepts.

Designs are a big part of the conversion optimization process. When working with multiple clients and across different CRO roadmaps, coming up with new design concepts can be difficult, and ChatGPT can be of help.

See this example below;

6. As a Personal Knowledge System:

User research, A/B  test research, and other types of research are being conducted daily by teams and organizations that prioritize CRO.

This means the CRO specialist has to keep up with a lot of new information that’s coming out, but that’s not entirely possible, and that’s because a lot of details can be forgotten/ slip through the cracks.

That’s where ChatGPT comes in.

According to Johann;

‘I’ve taken all my notes going back 10 years. Some of them are handwritten, and I’ve scanned them. I also took all my digital notes on Evernote. Combined, I’ve got 10 years of notes which include journal articles and includes web articles that include newspaper articles that include books on Kindle, etc. I put all those highlights from books, and all those notes, into a database using obsidian. I made a Link between Obsidian and ChatGPT. So it’s like Wikipedia, and I’ve been using that for a while. And every day, this database grows. That’s what a lot of people would call the second brain, you know, but it’s a personal knowledge system.

Instead of referencing the entire internet, the chatbot only references my notes and my readings. And so I can ask that chatbot anything about any of my notes, going back 10 years, any of my readings, any of my highlights. I’m constantly finding new ways to use it. But already, you know, I could go in there. And I could say, what if I wanted to know something about a subject, it could be any subject.  And  I can say to the Chatbot? What can you find in my notes about this particular topic? It goes ahead to connect the dots and pulls out the information.’

Johann also added;

And we’re just starting to build this now in my agency. It’s not built yet. We’ve got a lot of data, content, and frameworks, so that’s a wealth of knowledge and information. And now I’m building something, a bot similar to that, that doesn’t care about the internet. It doesn’t care what anyone else has written; it cares only about our internal content. Our research, methodology, and many meeting notes going back all the years relating to how we do things, how we think about things, our knowledge, our results, the meaning of those results, and all the discussions we had. Imagine when you have a chatbot that references just that institutional knowledge; how incredibly powerful that is. A practical example of the use of this bot agency-wide could be training for new employees on our methodology, and it could assist with expert insight in writing content, etc.’

Final Thoughts

ChatGPT will remain an important tool for CRO specialists looking to stay ahead of the curve and deliver outstanding results for their clients. 

By incorporating ChatGPT into their workflows, CRO specialists can tap into its potential to accelerate their tasks while maintaining quality.

Any CRO specialist who refuses to incorporate AI into their roles will be out of a job soon.

Johann is very clear about this;

If you’re not willing to do one of the following two things, you’ll be out of a job. And those two things are work with AI or work for AI.

Khalid agrees with Johann;

The less sophisticated you’re; eventually AI and automations are going to replace you. The more sophisticated you are, the longer the road ahead for AI to replace you.

If there’s anything ChatGPT won’t do, it is to replace the need to interact and connect with people.

Johann puts it best.

Humans are social creatures and always will be. AI won’t replace that.

So work on your social and networking habits. It’s definitely going to come in handy.