As generative Artificial Intelligence (AI) continues to rise, we’ve been exploring ways to adopt these technologies to enhance efficiency while maintaining authenticity on social media.
The role of generative AI
In the world of social media, generative AI tools are like having an on-demand assistant. The tools learn from you and can provide ideas, insights and even create content based on prompts, images and information provided.
While other forms of AI can analyse data and tell you what it sees, generative AI can take that data and create something entirely new.
Testing generative AI
Our team has integrated generative AI into our workflows using OpenAI’s ChatGPT for tasks such as checking spelling, grammar and punctuation, summarising media releases, and writing post copy. We’ve also tried using Canva’s Magic Write to edit headlines for our social media tiles by selecting options like “More fun” or “Summarise.”
These tools have provided rapid, on-demand feedback for the team. We’ve found it especially useful for providing an initial check over work before it’s peer reviewed by other team members or sent for approval.
We’ve also used AI image generators such as Canva Magic, Adobe Firefly and Midjourney. These tools convert text prompts into images.
To assess the ease of use of these platforms and test the effectiveness of AI-generated images on the NSW Government social media channels, we ran A/B tests on Facebook and Instagram. These tests were designed to assess customer engagement and feedback on using AI images on social media including comparisons between AI and stock images, and between AI photorealistic and AI graphic images.
We used a common prompt, which was “A flat lay of summer school holiday activity items including water sports at the beach, going for walks in National Parks, outdoor activities like picnics at parks or going for a swim in a rock pool,” to generate images for these tests. After several rounds of edits a series of images were created and then used for the A/B tests.
Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus. Vivamus elementum semper nisi. Aenean vulputate eleifend tellus. Aenean leo ligula, porttitor eu, consequat vitae, eleifend ac, enim. Aliquam lorem ante, dapibus in, viverra quis, feugiat a, tellus. Phasellus viverra nulla ut metus varius laoreet. Quisque rutrum. Aenean imperdiet. Etiam ultricies nisi vel augue. Curabitur ullamcorper ultricies nisi. Nam eget dui. Etiam rhoncus. Maecenas tempus, tellus eget condimentum rhoncus, sem quam semper libero, sit amet adipiscing sem neque sed ipsumLorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus. Vivamus elementum semper nisi. Aenean vulputate eleifend tellus. Aenean leo ligula, porttitor eu, consequat vitae, eleifend ac, enim. Aliquam lorem ante, dapibus in, viverra quis, feugiat a, tellus. Phasellus viverra nulla ut metus varius laoreet. Quisque rutrum. Aenean imperdiet. Etiam ultricies nisi vel augue. Curabitur ullamcorper ultricies nisi. Nam eget dui. Etiam rhoncus. Maecenas tempus, tellus eget condimentum rhoncus, sem quam semper libero, sit amet adipiscing sem neque sed ipsum
A/B test findings
The results of Test 1 (stock vs AI photorealistic images) were similar and indicated that both strategies could work, but further testing with more distinct creatives would help form a better understanding on what format is more engaging.
In Test 2, the results showed that AI graphic images outperformed AI photorealistic images on both Facebook and Instagram. Based on the winning ad set, we could consider making the graphic images bright and colourful using ‘colourful’ in the prompt.
Lessons learned and ethical considerations
These tests helped us figure out which tools are easiest to use and meet our needs for privacy and copyright. We learned that being clear and specific with what we ask the AI to do is also crucial, and uploading a reference image can help us meet branding guidelines.
Despite its benefits, using AI isn't without its faults. It doesn't always get local slang or understand some descriptions, and making images look realistic is difficult. This shows why it's important to have people check over AI's work to make sure it's suitable and doesn't have any biases.
AI can be a big help when we need to create and review content, especially when we have limited resources. However, it’s important to balance its use with ethical considerations and human oversight to ensure quality and authenticity.
Are you considering using AI on social? Our AI guidelines are here.