The Dos and Don’ts of Using AI Images in Digital Marketing

Whether you’ve seen it on John Oliver’s Last Week Tonight or simply existed on the internet in the last year or so, AI, and especially text-to-image AI image generators, is shattering conceptions of what is and is not possible for image creation. What better realm than digital marketing and digital PR especially to harness this incredible creative power for brands? 

While it’s not quite the Wild West of AI, there are several best practices to keep in mind as you start to flex your new image creation skills: here’s how to use AI-generated art in digital PR.

Different Text-to-Image AI Platforms

There are a huge variety of text-to-image AI platforms out there, but two of the largest and most popular are AI image generators DALL-E 2 and Midjourney. 

DALL-E, a product of OpenAI, the same company that brought ChatGPT to the world, works on OpenAI’s site platform and will generate images based on text prompts. Each prompt generates four images that reflect possible generations from the text. Each month, you get a finite amount of free image generations before you are prompted to subscribe or let time pass; the images are high quality and lack most of the flaws that have come to be associated with AI-created images, like extra limbs or slightly askew faces.

Midjourney AI is an independent platform that interfaces through popular webchat server Discord; while initially a bit steeper in its learning curve and trickier to set up, Midjourney is incredibly powerful and has been responsible for some of the more recent incidents of credible pictures making the news, like the Pope in a white puffer coat. As of this writing, its free trial has been suspended, but the most basic plan is $10 per month and allows for 200 hi-res images each month. Its placement on Discord makes it highly collaborative, which is a great benefit for the growing community working to learn how to use Midjourney AI image creation. 

When to use AI images- and when not to

Let me be clear: AI is not a replacement for all of your graphics, particularly those that include text or are purely infographics. Think of it as an augmentation of a campaign, or the campaign’s focus, as opposed to expository material for the campaign. No platform is particularly adept at creating legible (or even decipherable) text, and any logos you might ask it to produce won’t fare well in the AI image generation process. 

What AI CAN do, however, is imagine and depict things not currently achievable in reality: football players as farm animals, or what Chicago’s skyline will look like 100 years from now when the mushrooms take over. This can be of huge benefit as a focus to a content campaign, or it can be concept art to accompany data it might be loosely depicting. It’s also great as a jumping off point for ideation and inspiration in your next campaign. You can brainstorm campaign ideas using image AI. What these platforms CAN’T be, however, is exactingly precise.

Think about how your campaigns could benefit from imagery: do they need a focal point, or do they need an image to illustrate concepts in play? This is a great opportunity to use AI.

Here’s an example of how to use Midjourney in digital marketing and digital PR: say you ran a survey on America’s favorite burger toppings, and found that the ideal burger- per Americans- is a double cheeseburger with pickles, onion rings, ketchup, lettuce, and bacon. All you might want to do is feed those characteristics into Midjourney with instructions to have the burger sitting atop a podium and wearing a gold medal, and boom: you have concept art for survey data that focuses the reader and illustrates what your data says. This becomes an eminently sharable piece of digital PR information that journalists can use in stories and can also circulate on social media with ease. 

The greatest power of AI image generation for digital PR comes in its ability to instantly create impossible or far-off realities: while so many digital PR campaigns are grounded in consumer trends, perhaps looking at future predictions, composite images, or even fanciful reimaginings of popular landmarks could be to your benefit. These campaigns are a great way to shake up your own creative energy and showcase your client’s future-mindedness. How can you marry trends today, and consumer insights, with visions of impossible realities? That should definitely get your creative juices flowing. 

Careful Prompt Construction can Save You Time and Frustration

The key to getting what you want and need out of image-generative AI is really all about the prompt. Without learning how to speak the language of the generator you’re using, you’ll find yourself limited to uncanny valley imagery or pictures that aren’t at all what you envisioned. While these engines are powerful, they do require learning best practices for prompts.

Some notes:

• Be specific: Instead of keeping it vague, describe as much detail as you can of the image you have in mind: what’s the lighting? What’s the perspective? Is this a specific artistic style, or is it a photo of something? Some of the best prompts have strings of adjectives that convey mood, lighting, style, even the type of camera lens used to “shoot” the “photo” – the more words you can pack in to lend specificity to the prompt, the more you’ll also be able to scientifically tweak until you get what you want
• Be clear: These image generators don’t always know English in a conventional sense- try “a flock of birds” vs “birds” and see the difference. Similarly, keeping descriptions simple, and focal points to one subject will make it more likely that the image will turn out to your liking
• Collaborate: Platforms like Midjourney encourage collaboration among users as a way to figure out how to better use prompts together. The Discord server channel #prompts has lots of helpful use case scenarios and will compare what tweaking various elements of a prompt can result in. Work with your team to tweak or workshop prompts to reliably achieve the type of image you’re looking for– and then write it down! 

More specifics come with each generator and its specific platform- prompts that work one way in DALL-E might not achieve the same result in Midjourney, and vice/versa.

Campaigns, Copyright, and Careful Creativity

A brief note on copyright: unedited images produced by AI are not under copyright, according to the most recent court rulings. Therefore, if copyright is a priority for you, be sure to edit the image and add an element of human creativity to the basis of AI art. 

Furthermore, there are risks of using image AI in marketing. AI campaigns will be tricky: public sentiment is mixed on AI-generated text, and latest developments signal that Google will prioritize content created by humans as opposed to AI. While using AI imagery in your campaigns is great, be sure to include human elements and the human context to anything – text, image, etc — that might be created. AI is a powerful assistant and an incredibly capable visualizer of things real and unreal, but it is no substitute for a human. 

There’s also increasing uses of deepfakes in the public sphere as a result of AI- from Donald Trump’s perp walk to new “releases” from (AI versions of) Drake, the public is already being misled by AI; it’s important to be clear and explicit that images that are remotely photorealistic should be disclosed as AI.

Conclusion

AI image generation is a powerful tool in the belt of any digital PR pro. Using AI images in digital PR can provide campaign focal points as well as art that can accompany a data-driven campaign. From future landscapes to composite imagery exactly to the standards of the average American, AI art is a boon to anyone looking to add just a bit more creativity to their work. While all the benefits of image AI in marketing have yet to be seen because it’s a new field… there’s huge opportunity– and huge competition– to execute some truly unique campaigns. 

Read more from our blog