I evaluated the image generation capabilities of ChatGPT against those of Nano Banana, and only one succeeded in producing useful stock photos.

      One of my personal hobbies is to pose the same question to two different chatbots and evaluate which one gives the better answer. Similar to how a professor might quiz students (a practice I sometimes do as a volunteer), I assess whether the bot is fabricating information or if it delivers a smart and coherent response that is genuinely helpful.

      “Intelligence” and “coherence” are not terms I would typically associate with the image outputs from chatbots. Often referred to as AI slop, these bizarre creations — marked by blurred backgrounds, overly polished humans with styled hair, and occasionally extra fingers — are prevalent on social media and even appear as feature images in articles like this one.

      Stock images have already suffered from a poor reputation. A video featuring Emilia Clarke remains one of my favorite parodies of stock photos, and it still brings a smile to my face despite being released long before the era of AI slop.

      Times have evolved. OpenAI has recently enhanced the ChatGPT image generation model to rival Google Gemini and Nano Banana Pro.

      Unexpectedly, the images no longer exhibit that characteristic blur, and in my experiments, they often appear much more realistic than in the past.

      I decided to use identical prompts for both chatbots to see which one would yield the most useful results. I'm referring to images that could be utilized in advertising campaigns or articles, something suitable for sharing on social media alongside a clever caption.

      ChatGPT impressed me with its speed — none of its stock images resembled low-quality computer-generated output. However, Nano Banana Pro holds the advantage overall because it consistently produced more lifelike images in my trials. Below are the prompts I used and the images created by both bots, allowing you to draw your own conclusions. (Images from Nano Banana are on the left.)

      1. Individual listening to a podcast

      AI-generated podcast image. Gemini / Nano Banana is on the left John Brandon / Digital Trends

      Prompt used: Create a stock image of someone listening to a podcast in a stylish office

      I requested both bots to generate an image of a person listening to a podcast in an office setting. Nano Banana emerged as the clear winner here, even though the Nano Banana image features a slight background blur that might be noticeable. Nevertheless, AI is improving in avoiding absolute perfection. Honestly, I would find it challenging to deduce that the woman in the Nano Banana image isn’t real — her hair appears somewhat tousled, as one might expect in real life. In contrast, the podcast listener in the ChatGPT image seems off — the photo comes across as overly polished.

      2. Design a promotional image for my book

      AI-generated image of someone reading a book. Gemini / Nano Banana is on the left John Brandon / Digital Trends

      Prompt used: Showcase this book cover in the hands of someone genuinely interested in reading it

      For this test, I uploaded the cover of my book and requested both bots to create an image of someone engaged in reading it. Nano Banana ventured a bit too far with an image that depicts someone displaying excessive enthusiasm. Meanwhile, the ChatGPT image has an otherworldly feel (and not in a good way) that lacks realism. Everything about the ChatGPT image feels overly perfect — from the lighting to the pose — and it still exudes a CGI quality.

      3. Generate a stock image for an article

      AI-generated image illustrating AI image creation. Gemini / Nano Banana is on the left John Brandon / Digital Trends

      Prompt used: Create a stock image to accompany an article on AI image generation

      This time, Nano Banana didn’t win every instance. I requested both bots to produce an image for this article. ChatGPT delivered a somewhat useful image, though it’s still a bit too fake. In contrast, Nano Banana Pro created a convoluted image that layers a brain and a paintbrush on an Apple iPad, which ultimately serves no practical purpose.

      4. Leadership challenges

      Leadership challenges. Gemini / Nano Banana is on the left John Brandon / Digital Trends

      Prompt used: Depict someone grappling with a leadership goal, ensuring it looks very realistic

      One of the most frequently seen stock photos features someone struggling to reach a goal — whether it involves climbing a mountain, raising their arms to the sky, or leaping across a ravine. These stock images resonate since they reflect common experiences — we all face challenges in our pursuits. I would argue that both Nano Banana Pro and ChatGPT yielded below-par results here. The stock images they created depicted someone frustrated in an office surrounded by crumpled papers. Nevertheless, Nano Banana retains a slightly more authentic appearance.

      5. Emilia Clarke smiling at the camera

      Emilia Clarke AI-generated stock image. Gemini / Nano Banana is on the left John Brandon / Digital Trends

      Prompt used: Produce a new stock image of Emilia Clarke smiling at the camera in an office setting

      For one last test, I asked both bots to generate a new image of Emilia Clarke smiling at the camera within an office. I was