Digital Innovation Guest Columns
5 mins read

ChatGPT, Deep Learning, Visuals: Top Dos and Don’ts of AI in 2023

The launch of ChatGPT has shaken publishing to its foundations, with naysayers predicting a dystopic literary future. Yet if harnessed in the right way, artificial intelligence can be the co-pilot every publisher and journalist wishes for. Tyler Bishop, CMO of Ezoic explains more…

From its use in entertainment, shopping, healthcare, and much more besides, artificial intelligence continues to be implemented everywhere in our daily lives. AI has also gained a firm foothold in marketing and publishing, but while it can be a great help, there are times when it is better not to use it.

How do you know when to use AI and when it’s best to have a human touch?

AI in 2023

Several terms have served as key technological turning points in AI: NLP (natural language processing) and stable diffusion in particular. NLP is the discipline of computer science that helps computers understand language in the same way humans use it – this is what powers ChatGPT, for example.

Stable diffusion is a new, deep-learning technological development from 2022 that converts text into images. Type in a description of the art you are trying to create (a quaint house on top of a snow-covered mountain, or you might try, black cat with a top hat, etc.) and the technology generates it. We saw big progress in this with products like Dall-E.

The culmination of these technologies hitting the news cycle repeatedly in 2023 has created a huge AI buzz on the internet. Let’s unpack what it’s really about.


#1: Do use AI in video

Video is one of the most reliable ways to use AI in content creation. Many video AI technologies work well because they’re iterating on something that the author has already written and are simply adding visuals.

A form of AI that works well for those hesitant to be on camera is using avatars (see Synthesia and Elai), created by recording actors and then processing them with AI.

If having a person in your video isn’t necessary, there are other types of AI that take written content and use it to pull files together and combine them into a .mov file, such as Flickify. This has huge potential to save publishers and creators time and energy on editing, which can be difficult to master, and often otherwise requires additional technology to do so.

#2: Do use AI for simple automation

One of the best uses for AI is anything that frees up time to be more creative. In the world of content creation, there are countless tasks that are tedious or complex and could be improved by machines.

For instance, SEO topic suggestion tools like Writesonic or NicheIQ are sophisticated and look at the data on your site to identify the content gaps you’re not currently filling, probably better and more quickly than you could do yourself.

Another example – countless tools out there like Sprout or HootSuite can automate posting to social media, an important but tedious side of digital publishing.

Bottom line: make AI do repetitive work, so that humans can create original, engaging content.

#3: Do iterate on AI’s ideas

One of the smartest ways you can use AI is to see what it comes up with and iterate on it. For instance, if you’re looking to make a logo for your new blog but have a complex idea that is hard to put into visuals, you could get inspiration for a rough draft via an AI image generator like Dall-E. Take a look at it and use it as a starting point. Use this as a mockup to give to a graphic designer.


#1: Don’t use AI to write original content

Most AI long-form content lacks flow and cohesiveness that only humans can provide. Consider a huge priority that you have as a content creator: building trust. A machine doesn’t necessarily have this same priority or capability.

In an experiment at BusinessInsider, a writer found that ChatGPT wrote a great article given her prompts – except that it generated false quotes. How do you keep a machine from lying when you can’t add a “journalistic integrity” function?

AI-generated content is considered ‘automatically generated’ content by Google and goes against Webmaster guidelines. While Google and your audiences want original content, machines can only learn when there is source material to learn from. AI can edit, add, and improve content but it will struggle to produce little more than objectively factual information. Consider the example of asking AI “how to stop a crying baby”. AI cannot have a child; nor can it imagine or experience the solutions for this. At best, it can summarize human experiences, making it, in the best-case scenario, unoriginal but an accurate summarization of content.

#2: Don’t trust every AI-generated image

AI-generated images are popular right now thanks to Lensa AI, which takes your image and generates artistic images using the selfie uploaded, and other apps, as well as Stable Diffusion, which allows you to enter a request, like “shiba inus in space,” to create that image. However, since AI cannot create original content and is instead constantly “learning” from what is available, some people have noticed that it may leverage other individuals’ art to create images, bringing with it copyright concerns. Additionally, Eurocentric beauty norms may often be used in some  AI-generated images, even if someone is not of European descent. Many AI image creators do not have enough variation to create something that is unique and accurate enough to not be an issue.

#3: Don’t forget to double-check AI’s work

You shouldn’t assume that AI won’t make mistakes just because it’s a machine. Check what AI comes up with, the same way you might check over your own work before publishing it.

It can be tricky to remember just how many elements humans are constantly weighing when we leave that process to a machine. Make sure that any of the AI you use is not only following good ethical principles, but also consistent with all copyright laws and your company’s own guidelines and requirements.


While AI makes our lives easier and better in many ways—voice assistants, fraud prevention, logistics and distribution—it is highly unlikely that AI can ever completely replace humans. People will always be necessary in some capacity because humans have something that AI can never be taught: true creativity.

If you don’t believe me, just check out these AI-generated dating app messages that received zero responses. AI is good at learning and copying, but not necessarily originality or social cues.

Tyler Bishop
CMO, Ezoic

Tyler Bishop is a digital publishing influencer and chief marketing officer of Ezoic, the artificial intelligence technology for websites to monetize content with display ads through streamlining implementation, optimization, and testing. He is an award-winning marketer who has worked for Microsoft and was featured on the cover of The St. Louis Business Journal for his unique approaches to digital marketing.