Before you start: Please download this Participation Information Sheet, that explains about data collected in this tool.
The Basics
Visual Generative AI refers to a type of artificial intelligence that creates new images based on prompts—these can include text descriptions, rough sketches, or even other images. Unlike editing or filtering tools, these systems generate entirely new visuals from scratch, which can range from realistic photographs to surreal or stylised artworks.
At the core of these systems are machine learning models trained on massive datasets containing millions of images. Tools like Midjourney, DALL·E, and Stable Diffusion learn from these examples—absorbing patterns in shape, colour, composition, and style—to produce outputs that match a given prompt. Though they don’t “see” like humans do, they’re adept at blending ideas and mimicking visual styles.
Most visual generative AI tools rely on diffusion models or GANs (Generative Adversarial Networks). In simple terms, these models start with a field of noise and gradually refine it into a coherent image, guided by what the AI has learned during training. For example, if you ask for “a sunset over a futuristic city,” the AI constructs an image by drawing on relevant visual concepts from its training data.
Important Disclaimer
Not all AI-generated images are created with malicious intent. Many are used for legitimate artistic, commercial, and educational purposes.
AI-generated images are becoming increasingly realistic—and with that realism comes influence. These visuals can shape opinions, stir emotions, and even alter how we understand events, people, and social movements. From synthetic political rallies to distorted beauty standards, visual AI has real-world impact.
While these tools open up space for creativity, resistance, and expression, they can also be used to mislead, manipulate, and cause harm—especially when images are taken out of context or designed to deceive.
Understanding the potential impacts—both positive and negative—helps us engage more critically and ethically. From misinformation and deepfakes to creative resistance and therapeutic applications, the effects of AI imagery are wide-ranging.
The more visually literate we become, the more prepared we are to spot manipulation and harness these tools for good.
This section explores the wide range of consequences, both positive and negative, to help you navigate this evolving landscape with greater awareness and confidence:
Misinformation & Political Manipulation
AI-generated images can fabricate events, people, or crowds—altering how we perceive political truth.
These tools have been used to depict fictional protests, digitally insert politicians, or exaggerate public support.
This poses serious risks to trust in media and democracy.
Trauma & Conflict Imagery
Synthetic war and disaster images may be used to evoke strong emotions or simulate atrocities that never occurred. While some are intended to raise awareness, others may be used maliciously or without regard for ethical boundaries—potentially triggering distress or spreading misinformation.
Beauty Standards & Fake Bodies
AI tools are now used to create flawless, unrealistic human figures. These bodies—often shared on social media or promotional platforms—can distort our perception of normal appearance and reinforce narrow, exclusionary beauty ideals, especially for younger audiences.
Creative Resistance & Activism
Not all AI use is harmful. Artists and activists are using visual AI to challenge censorship, explore identity, and create work that speaks to marginalised experiences. When used ethically, these tools can amplify voices and foster cultural expression in new ways.
AI-generated images are made using powerful systems trained to recognise and replicate visual patterns. But understanding how they’re created can help us interpret them more critically. This section breaks down the process behind AI imagery—and shows you what to look out for when trying to spot synthetic visuals.
By exploring the tools, techniques, and tell-tale signs of image generation, you’ll gain the skills to read AI content with greater awareness and confidence.
AI-generated visuals often look convincing at first glance—but many carry subtle clues that reveal they weren’t made by a human hand. Learning to ‘read’ these images means slowing down, looking closer, and asking the right questions.
This section helps you train your eye to detect inconsistencies in detail, lighting, perspective, and realism. Whether you’re spotting a fake news image or simply trying to understand if an artwork is AI-generated, these cues can help you interpret what’s real, what’s synthetic, and why that matters.
Key cues to include:
Inconsistent lighting or multiple shadows
Repeated or melted background elements
Unusual body parts (e.g., too many fingers or twisted limbs)
Implausible object relationships or logic gaps (e.g., floating hands)
Surfaces that are overly smooth or textureless
Misspelled or warped text in signage
🧩 “Ask Yourself: Does this image seem too perfect, too chaotic, or somehow ‘off’?”
AI images don’t just appear—they’re created through a process that starts with a user prompt and ends with an entirely new image. This process combines creative input with machine learning to produce visuals based on patterns in data.
Here’s a simplified breakdown of the steps most platforms follow:
Text/Image Prompt Input – The user describes what they want to see (e.g. “a tiger made of leaves”).
Neural Network Processing – The AI interprets the prompt using training data from millions of images.
Image/Video Generation – A new image is created by blending patterns and features it has learned.
Key Platforms for making AI Images
🤖 ChatGPT 4o Images: Web and App-based (instructions via conversational text).
Explore more here.
🧠 Leonardo AI: Web and App-based (instructions via prompts and visual references).
Explore more here.
The Process
User describes what they want to see, or shares an image showing what they want things to look like
AI interprets the prompt using training data
Creates new image/video from learned patterns
Which platform have you heard of or used before?
Your answers help us understand which tools are most familiar and where more explanation might be helpful!
AI image tools give us the power to visualise things that don’t yet exist—alternate futures, powerful emotions, or complex ideas. But with that power comes responsibility. This section invites you to explore AI image generation ethically, using creativity to inspire, include, and inform—rather than mislead.
Whether you’re designing flyers, storytelling through visuals, or advocating for change, this is a space to think about how to create with care.
Before generating AI visuals, consider the following:
What is the goal of the image?
Is it to educate, persuade, provoke thought, or bring joy?
Is the image respectful?
Avoid using AI to mimic real people without consent, or reproduce harmful stereotypes.
Could the image mislead someone?
If yes, consider clearly labelling it as AI-generated or using design cues to prevent confusion.
Can you add context?
When appropriate, pair images with explanations or captions to help others interpret them responsibly.
💡 Remember: ethical AI creativity means being mindful of the impact your image might have—on individuals, on trust, and on truth.
Despite the potential harms, Al images can also be used as a powerful tool for creative activism and resistance. AI-generated visuals can be a powerful tool for public engagement—especially when time, budget, or visual materials are limited.
In this example from a vaccine confidence campaign, AI was used to:
Design inclusive workshop flyers showing diverse community settings
Create visual metaphors (e.g. “Vaccination as a Community Shield”) to explain complex ideas simply
Produce culturally relevant, audience-centred visuals that feel familiar and empowering
Engage participants in co-design by iterating on AI-generated drafts together
These images helped spark meaningful dialogue and were clearly labelled as AI-assisted. The goal wasn’t to trick the viewer, but to encourage understanding and belonging.
Now it’s your turn to experiment with AI image generation. You don’t need advanced tools—just a good idea and a thoughtful prompt.
Activity Prompt:
Ask ChatGPT (or another image tool) to:
“Generate a visual for a health campaign that builds trust.”
You might want to include:
The tone or emotion you want to evoke (e.g. calm, caring, hopeful)
The audience you’re speaking to (e.g. parents, young adults, elders)
Visual themes (e.g. community, protection, science, family)
Colours or symbols that convey safety or care
Example prompt:
“Create an image of a warm, community health setting where diverse families are learning about vaccines. The tone should feel welcoming, with natural light, posters about trust and science in the background.”
Now… have you been paying attention?
This is your chance to test what you’ve learned so far. Each image below has a story—some are real, and some are AI-generated.
Can you tell which is which?
Click an image to reveal the answer and the context.
Which of the images below are fake or real?
It was taken at the funeral of Nelson Mandela in 2013.
At the time, while there was controversy over whether or not it was disrespectful to smile and take a selfie at a state funeral, the photo itself was fact-checked as real.
It was created with MidJourney, and shows a fake, but realistic image Obama taking a selfie with famous Indian leader, Mahatma Ghandi who died in 1948. President Barack Obama was born in 1961, so they would never have met.
Whether you got them all right or were surprised by a few, the goal of this activity isn’t perfection—it’s perception.
AI-generated images are designed to look convincing. But if you look closely, you may start to spot:
Subtle inconsistencies in lighting, facial expressions, or proportions
Implausible scenes or people together in ways that don’t make sense historically
Unrealistic textures or “perfect” symmetry that feels uncanny
Emotional cues that seem exaggerated or oddly neutral
Let’s Recap
This Detect challenge brings it all together: seeing isn’t always believing—and the more visually literate you are, the more power you have to question, challenge, and choose what to trust.
Head to the Download tab to grab your Visual AI Literacy Checklist—a practical tool to help you stay sharp and share what you’ve learned with others.
So what are the tools and strategies for identifying AI-generated images and promoting visual literacy?
Visual inspection involves looking closely at certain features that often signal AI manipulation or generation.
Common signs include distorted anatomy, inconsistent lighting or reflections, nonsensical text, and repetitive or unnatural patterns. By training your eye to spot these details, you can develop a more critical and informed approach to evaluating digital images.
Look for common AI artifacts:
Using a reverse image search is simple:
These tools are accessible via both desktop and mobile devices—and some are already built into platforms you might use every day.
Here are some tools you can try:
Investigate where and how the image is being used:
Here is a summary of things for you to watch out for. You can click on the Download button to save a copy to use for later:
