So I really had fun creating this article. I think it solves a need a lot of SEO teams have, especially lean ones. Oftentimes as SEOs we’re extremely focused on net new content but not images. Whether we’re performing content refreshes or undergoing full scale content operations campaigns, we’re oftentimes 95% lasered-in on the text of a page, but not rich media assets. Oftentimes, that falls outside of our purview.
Well the good news is that with AI, we can at least offer to put rich media assets back into our scope. Especially now that it’s gotten better and better to the point of being able to create some extremely high quality things.
In this article, I’m going to show SEOs how they can create images at scale for existing site assets. Of course, you could teak this process to whatever fits your needs of large scale image creation by adjusting the prompt or the Screaming Frog configuration.
Of course, if you’re interesting in having us run this process for you, please feel free to reach out.
Let’s get into it.
Introducing Nano Banana Pro
So one launch that’s been huge in the AI-community is Nano Banana Pro. Recently Google announced their latest image generation model was coming to Gemini 3. The model has been exceptionally well received, especially for doing things that a lot of image generation models have struggled with.
For instance, AI has definitely struggled with placing words in text. However, Nano Banana Pro is exceptional at this. This Reddit thread has a few examples of how sophisticated in can be.
And the good news is that there’s an API directly for it. The Gemini API has been updated with Gemini 3, which includes direct access to Nano Banana Pro. The modal name is “gemini-3-pro-image-preview” and that’s what you’ll need to reference later.
Fortunately, Screaming Frog already has the capabilities of connecting to Gemini + image generation. You just need to configure it to the latest model to take advantage of it’s ability to connect to Nano Banana Pro.
How To Create Images At Scale With Screaming Frog + Nano Banana Pro
I still believe that Screaming Frog’s AI crawling is one of the most underutilized features of the SEO community. It allows SEOs to use a tool they’re familiar with as an API access point and push in data at scale. Without knowing code or running Terminal commands, SEOs still have the power to access OpenAI and Gemini’s APIs.
First you’re going to need to grab a Gemini API key. Go here to learn how to use Google AI Studio to grab yours.
Next you’ll go into Screaming Frog and navigate to Configuration > API > AI > Gemini. You’ll see a screen where you can copy/paste your API key directly.
Next navigate to “Prompt Configuration” to configure the prompt you want to send to Gemini. Here are the settings you’ll need to configure. Of course you’ll customize the prompt to your exact needs:
- Model Category: Image Generation
- Modal Name: gemini-3.1-flash-image-preview
- Prompt Target: Page Title
- Prompt: Read the page title. Then create a flat vector illustration that we can use as the hero image with 1,280 x 720 dimensions.
I’d also recommend testing your prompt a few times before you run it. Open the “Prompt Tester” (play button) in your Prompt Configuration box. You’ll then be taken to a new configuration box where you can directly test individual pages before running the crawl.
Before starting your crawl, I’d also recommend limiting it a bit. I found that Screaming Frog was more consistently able to connect to the Gemini 3 API after doing that. This is really easy, just go to Configuration > Speed. I check Limit URIs/s and change it to 2-3 maximum.
Finally, you’ll want to go ahead and start your crawl. Plug your domain into Screaming Frog and start to crawl on Spider Mode. As the crawl executes, you can go to the ride hand sidebar to see the “AI” section at the bottom. Clicking the name, you should see images start to populate in the output. Extremely cool to see.
You can then hover over an individual image. Screaming Frog will expand the image to show you
To export them to use, you’ll need to perform one last step. Go to Bulk Export > AI > Images > Open. You’ll then have a file that lists all of the images from Nano Banana Pro downloaded directly on your computer.
The Results
Honestly, these images pretty freaking good. I found that general prompts such as “create images” did lead to generic AI-looking type of images. However, if you’re more specific with what you’re asking for (flat vector illustration, 1,280 x 720 dimensions), the output dramatically improves.
For a Twilio article titled “How to Build an AI Chatbot with Facebook Messenger, OpenAI, and Twilio Programmable Messaging using Python” it PERFECTLY captured all of the entities.
To be honest, I think it’s a lot more relevant than the generic one Twilio has set for it. I’d argue replacing would be a directly improvement in terms of image relevance + integrating optimized text into the asset.
There are definitely still some misses and not every image is perfect. For instance, I’m not sure why this image on “Email Marketing” has everyone checking their email from their smartwatch. Not technically inaccurate but probably not the best representation.
Conclusion
Image generation is one of the things in SEO that we don’t spend a lot of time because frankly we don’t have the design skills. I’m literally one of the worst designers ever. However, AI enables us to create these assets on our own. AI + APIs allow us to even potentially scale that production. The system isn’t perfect but I believe running this process will at minimum lead to sites have a few more relevant/optimized images assets than they had before. For lean teams with no designers, this could be a huge breakthrough.
