Who's the Better Creative: ChatGPT Plus vs Sora
Recently, the affiliate community has been actively discussing the potential of neural networks for creative production. Some affiliates and teams have already integrated the paid versions of ChatGPT and Sora into their funnels.
Today we put two top neural networks through a stress test. We'll compare how they handle creatives and whether they can be used for full-scale ad campaigns.
What You Need to Know About ChatGPT and Sora
In April 2025, ChatGPT significantly updated its image generator, but free access is only available to users on the paid plan at $20 per month. Free plan users were given the ability to generate a limited number of images per day.
The chatbot also received a built-in editor that allows you to modify specific parts of an image — for example, a character's hairstyle or a person's face. Edits can be made either through the visual editor or by describing the changes in text.
Public access to the Sora video generation neural network had been anticipated since February 2024, but even now only ChatGPT Plus subscribers can properly test it — with a video length limit of 10 seconds. Generating longer videos requires the $200/month plan, which still caps output at just 20 seconds.
Our stress test was inspired by public experiments from Ukrainian affiliate bloggers who are actively testing ChatGPT-generated creatives across various verticals. The author of the "Де депи, Лебовські!?" channel even burned a 61% ROI running gambling traffic to Australia.
Affiliates love that they no longer have to wait for a designer to deliver a batch of images — they can quickly test approaches on the fly. For example, pull creatives from a SPY service and ask ChatGPT to adapt them for a different GEO.
E-commerce
To test creative quality for e-commerce, we asked the neural networks to generate images for selling the Ecoflow station — and the results genuinely surprised us. Both handled it well — ChatGPT on the left, Sora on the right.
Sora didn't add any unnecessary elements, while ChatGPT inserted a random logo. That said, everything can be fixed with additional prompts if needed.
Dating
For the dating vertical, we asked the networks to generate an ad for a white-hat dating app based on a reference image.
ChatGPT delivered a solid result with no text errors — something the previous version of the model struggled with badly. Sora handled the task almost perfectly, though there are minor artifacts on the faces.
iGaming
Based on affiliate tests, neural networks perform exceptionally well with slots and gambling-themed content in general. Designers are already using them to generate ready-made UI kits for slots, cutting down significantly on manual work.
We prompted the networks to create creatives for the popular crash slot Chicken Road with Ukrainian visual elements. We also asked them to incorporate banking app interfaces, but ChatGPT once again inserted strange logos. Sora did a decent job, though there are some unrecognizable symbols on the money.
Finance
For the finance niche, we asked the networks to generate creatives for a fictional payment service in the vertical's standard visual style. We also included visual elements that communicate the offer's value proposition.
The ChatGPT character resembles Zuckerberg, but the network refused to add his actual face due to its safety policy. Sora produced a more realistic creative, but once again struggled with logos.
Lead Gen
In lead gen, creatives typically carry minimal informational load. The key is for the image to clearly communicate the offer's message.
We asked the networks to come up with the concept independently to see how creatively they could operate within acceptable boundaries. Once again, both delivered decent results.
Crypto
Despite the gray-area reputation that follows crypto, white-hat ads do appear in ad networks. Based on preliminary tests, neural networks also handle crypto creatives quite well.
The prompt was straightforward — create an ad for a new crypto exchange with 0% fees and no KYC. In this case, Sora produced the higher-quality result, though for some reason it defaulted to English despite the prompt being written in Ukrainian.
Key takeaways:
- Neural networks generate images at a significantly higher quality than a year ago. Text errors are nearly nonexistent.
- A clear prompt is essential for a precise result.
- In most cases, at least 2 generations are needed to produce a ready-to-use creative.
- Using references will dramatically improve the output.
- AI-generated creatives can be used in real campaigns, but results are unpredictable.
The good news is that for $20 per month you can use ChatGPT Plus together with Sora and generate creatives across both neural networks at once. As affiliate case studies show, AI-generated creatives are already capable of driving profit.
