The battle of the GPT’s – GPT-4 v GPT-4o – 3 head-to-head tests and what it means for marketers and online business owners
Here's what we have for you today
I got access to OpenAI’s new “smarter and faster” GPT-4o late yesterday afternoon.
And boy, does this feel like a massive upgrade.
Throughout 2024, I’ve mainly been using Claude and Gemini.
I had become allergic to the voice and tone of GPT-4 content.
And, annoyed at how it could no longer perform specific tasks it once could.
But I have a feeling I will be using GPT-4o more and more:
Here are 3 head to head tests I gave GPT-4 and GPT-4o.
Test 1 – Analysis of 878 responses to a multi-question survey
Client feedback is a goldmine for copywriting and offer creation.
GPT-4, even when using Advanced Data Analysis, always struggled with analyzing large volumes of data. It also provided feedback that was inconsistent with the data or based on a small sample. GPT-4o was the clear winner in this head-to-head test, providing precise, accurate, and comprehensive feedback on the data.
1-0 to GPT-4o. On to the next test.
Test 2 – Create content similar to previous content
I uploaded my last 15 posts into GPT-4 and GPT-4o and asked them both to draft a new post on drafting Facebook Ad Headlines.
GPT-4 responded with its usual verbose and over-hyped language style. GPT-4o, on the other hand, sounded natural, followed my language style, and even mirrored my approach to copywriting from some of my earlier posts.
This was a big win for GPT-4o. With the score at 2-0, we move on to the last test.
Test 3 – Draft a Facebook ad based on a framework
For this test, I provided minimum instructions or guardrails in the prompts.
I loaded up the landing page copy and asked them both to draft ads to maximise outbound clickthrough rate and on page conversion rate. This was the most interesting test. GPT-4o possibly had a slight edge but it was very close.
It is worth noting that outputs from both models were flagged as 100% AI created in my detection tool.
ChatGPT-4o is a big leap forward, possibly even bigger than the leap from 3.5 to 4.
Based on my tests, it looks like the more information you give GPT-4o, the better it will perform.
And this makes it the perfect language model to use for your CustomGPT.
Load up all your previous content, marketing copy, client feedback and even scripts from your course and membership videos into a CustomGPT.
Then get, your CustomGPT to create new content using GPT-4o as the base model.
I think you’ll be surprised at the results.