Skip to main content

ChatGPT & Paid Media Mythbusting

“ChatGPT is a powerful tool that has the potential to revolutionize the paid media industry, and in this blog post, we will explore how it can be used to improve performance and drive results.”

Well, that’s what the chatbot told me to say anyway.

But how much of this is fact and how much is fiction? Is ChatGPT the future, or is it just hype? How worried (or excited) should PPC marketers be about this tool? 

Let’s dive in!

Methodology

There have been a lot of articles over the past month about how to use ChatGPT for Paid Media. Some of these posts have even been written entirely by ChatGPT…which makes it difficult to know just how relevant those use cases are. So we decided to test it for ourselves. 

Our first step was to give ChatGPT the following prompt – “Pretend you work for a digital marketing agency. Write an outline for a blog about “How ChatGPT Will Impact Paid Media”. The blog should not be longer than 1000 words.” 

However, we’re not mindlessly following these recommendations. Instead, we wanted to compare the AI’s answer to reality. This gives us a better understanding of how useful the tool actually is in its current state, strengths and weaknesses, and what the future might look like.

Below are some of the most common use cases for ChatGPT based on their answer to our original prompt, as well as various articles and threads. 

Disclaimer: All of the text snippets below that are in quotation marks indicate that they were written by OpenAI’s large language model, ChatGPT. Additionally:

content disclaimer from chatgpt

ChatGPT Content Disclaimer

“How ChatGPT Could Impact Paid Media”

“Keyword Research”

AI Claim: “ChatGPT’s natural language processing capabilities can be used to perform accurate and efficient keyword research, which can lead to more effective ad targeting and higher conversion rates.”

Reality: After testing various prompts, we found that the terms provided by ChatGPT were mostly relevant. In fact, the majority of them are in our campaigns already, which is why they were redacted in the screenshots below. 

  • However, the monthly search volume and average costs-per-click are not accurate compared to actual performance or Google Ads’ Keyword Planner. This is likely due to the fact that ChatGPT was trained with a dataset that only goes up until 2021.

Final Thought: ChatGPT can give you a solid list of keywords to start researching, but you should still run them through the Google Ads Keyword Planner (or whatever keyword research tool you use).

Examples:

chatgpt keyword research table

Asking ChatGPT to do keyword research for a client

screenshot of chatgpt keywords in google ads keyword planner

Volume & CPCs in Google Ads’ Keyword Planner for the same terms that ChatGPT provided above

screenshot of chatgpt keyword research

Testing if ChatGPT can refer to specific URLs to do keyword research

“Ad Copy”

AI Claim: “ChatGPT can be used to generate compelling ad copy that resonates with target audiences and drives higher click-through rates.”

Reality: Some copy is relevant, but some is not and just sounds weird. We also found that the tool works best when you give it parameters and a reference point. 

  • This could include character counts, audience/persona information, keywords, previous top-performing ads, competitors’ ad copy, etc.

Final Thought: ChatGPT is a good starting point for brainstorming ad copy ideas, but outputs need to be tweaked before you consider running them in an actual ad. You should also make sure to have your client review the variations before anything goes live, which ensures alignment with their brand’s messaging guidelines.

Examples:

screenshot of ad copy ideas from chatgpt

Ad copy ideas from ChatGPT

screenshot of ad copy ideas from chatgpt

Using ChatGPT to build on the previous output (above) and get more useful ideas

“Optimization”

AI Claim: “ChatGPT can be used to optimize ad campaigns in real-time by analyzing data and making adjustments as needed.”

Reality: Knucklepuck tested this a few different ways using fake data. 

  1. We provided account-level performance metrics and asked how we could improve CPA (screenshot 1). 
    • The recommendations we got back were not very specific, more like general suggestions. I would not consider that “real-time optimization.”
  2. We asked general questions about Quality Score and how to improve various factors (screenshots 2 and 3). 
    • Again, the recommendations we received were accurate but not specific. 
  3. We provided keyword-level performance metrics and asked which term we should pause (screenshot 4). 
    • Honestly, this response was pretty solid. It wasn’t the most difficult decision to make, but the bot’s explanation/reasoning was sound.

Final Thought: ChatGPT isn’t going to solve all of our account optimization dilemmas. Claiming that the tool can be used for real-time optimization is a stretch. 

  • However, I was pleasantly surprised with the overall accuracy of the outputs, and it gives you a great starting point. The responses could be used to put together a general framework, but then it’s on the advertiser to create a more customized action plan. 

Examples:

screenshot of chatgpt suggestions for google ads account optimization

Account-level optimization recommendations

screenshot of chatgpt suggestions for google ads quality score optimization

Quality Score optimization recommendations

screenshot of chatgpt suggestions for optimizing expected CTR in google ads

Expected CTR optimization recommendations

screenshot of chatgpt suggestions for google ads keyword optimization

Keyword optimization recommendations

“Audience Segmentation”

AI Claim: “ChatGPT can be used to identify patterns in customer behavior and segment audiences for more effective ad targeting.”

Reality: You can’t have full confidence in any “patterns in customer behavior” that ChatGPT identifies, as it’s only trained on data through 2021. 

  • In terms of segmentation – as is the case with the other tactics above – you get some decent, general insights into how you could target (or avoid) different audiences. It largely depends on the prompt you use, and you should double check with your client before moving forward with its recommendations.

Final Thoughts: Audience segmentation can be done to an extent with ChatGPT. You are still going to need to tweak, customize, and review those strategies with your client though. 

  • Additionally, some of the longer-term behavioral patterns that ChatGPT identifies may still be accurate (depending on the industry you’re in). But you will be perpetually behind the times if you rely on ChatGPT for more recent/specific consumer trends.

Examples:

screenshot of chatgpt audience segmentation recommendations

Asking for segmentation advice based on existing personas

screenshot of chatgpt audience segmentation recommendations and insight into consumer behavior patterns

Asking for segmentation advice and insight into consumer behavior patterns based on existing personas

screenshot of chatgpt audience segmentation recommendations and insight into consumer behavior patterns

Asking for segmentation advice and insight into consumer behavior patterns with no existing personas

screenshot of chatgpt audience segmentation recommendations and insight into consumer behavior patterns

Asking for insight into consumer behavior and advice on positioning

Additional Use Cases & Tips

There are a number of other potential uses for ChatGPT that we came across in our research. This includes:

  • Strategic Planning – In general, ChatGPT can help provide guidance on strategy. So if you’re ever stuck on how to move forward with a new initiative, this can be a good starting point. 
    • However, there may be some key components missing. For example: in the first screenshot below, there isn’t any mention of regular keyword research (only competitor and negative keyword research). 
    • You can also refine your prompt by asking for more specific ad formats and platforms, like in the second screenshot below.
screenshot of chatgpt paid media strategy recommendations

Asking for general strategic recommendations

screenshot of chatgpt paid media strategy recommendations

Asking for more specific strategic recommendations

  • Code and Scripts – The fact that ChatGPT can generate code is appealing to marketers without technical knowledge or experience. However, we didn’t test applying this script, and it may require some changes before being fully functional. 

screenshot of google ads script code provided by chatgpt

Additionally, the best way to get meaningful outputs from generative AI tools is to provide context and be specific in your prompts. For example: 

  • Character counts (“headlines can not exceed 30 characters”) 
  • The number of results you want generated (“draft 4 descriptions”)
  • The format you want the response (“present the results in a spreadsheet”)
  • Leaning into roleplaying (“pretend you’re a PPC Specialist”)
  • Asking to use a specific tone or voice (“use a clear, concise, and informal writing style”)
    • You can take this one step further by providing an example of your previous work (“analyze the paragraph below for voice and tone. write a new blog post about AI in the same style”)
  • Outlining information about your audience personas (“the audience you’re writing for is digital marketers who are hesitant to use LinkedIn”)
  • Adding phrases like “prioritize uncommon ideas” to get more unique results

Other Helpful Threads

Conclusion

I was skeptical about ChatGPT when we started researching this article. It seemed like a shiny new toy that everyone would stop talking about after a week or two. 

  • There are also concerns about AI in general, especially around misinformation, plagiarism, and having bland/un-unique content (including its potential effects on SEO). Additionally, humans are biased creatures, and therefore any data that is fed to machines (and its outputs) will also have biases. 

But honestly, I was impressed after testing the tool. In its current state, ChatGPT is very useful as a starting point for Paid Media tasks. 

The next time you’re struggling to think of ad copy or keywords for a new initiative, why not “ChatGPT it”? It’s worth testing how you can add the tool to your workflow – just make sure you’re aware of the limitations. It would be a mistake to aimlessly follow and implement every output that it gives you. 

Lastly, I don’t believe that PPC marketers will be out of a job because of generative AI tools. I actually think that the opposite is true: If everything around you is created by a machine, having a human touch will be even more important in the future.

 


Resources

Can We Help?

If you have an idea, a project or a challenge, we’d love to hear about it.