Regardless of whether you are a believer that AI is going to revolutionize every industry for the better or a skeptic who sees AI as a threat to original content creation and a host of other ‘human’ roles, I think everyone agrees that AI is here to stay. In the email marketing arena, AI has already shown it can deliver value in a variety of ways, from content creation to testing strategy and performance analysis. But it’s in the content creation arena where the debates are most polarizing.
The topic of AI and content generation has come up on more than a few occasions on weekly OI Members Only Zoom Discussions, and Doug Morneau wrote about it in a recent article for the OI blog (Is Honesty the Best Policy? The Truth About AI Content Disclosure.) These conversations have tended to raise as many questions as clear-cut answers, as the landscape continues to shift. So, I wanted to take a stab at creating a framework for considering not only AI’s place in various types of content creation, but also where disclosure of that involvement is either more likely or less likely to be necessary. If nothing else, it may simply move the discussion forward, as definitive answers remain elusive.
Why Does Disclosure Matter?
This seems like a good place to start, just to set the foundation. Why are we all concerned about disclosure when it comes to AI in content creation?
Legal Requirements - Laws around AI are still very much in their infancy. Any that have been passed likely haven’t been challenged in court yet, to determine whether the requirements they define will hold up to legal scrutiny. Today, there may not be many established legal requirements around disclosure related to AI’s involvement in published or otherwise shared content generation. Still, it’s probably only a matter of time. So, while this may not be a necessity today, it will likely come into play in the years ahead. The open question is where the laws will draw a line after which disclosure is required.
Ethical Requirements - We don’t need legal regulation to know that there are ethical questions about using AI to create content and who gets to take credit for that content. You also don’t need to be an ethics expert to recognize that there are questions to consider if someone uses AI extensively in creating content and then publishes that work as their own authored content. But, even here, it’s not as straightforward as we might like. How much involvement does AI need to have in creating the content in order for disclosure to be ethically required? Spell-checking? Grammar review? Initial ideation? Creating a content outline? Revising content the author writes? Creating entire sections of content based on a prompt? These answers will vary depending on who you ask.
Good Business Practice and Transparency - How about just good business practices and general transparency? Is disclosing the involvement of AI in creating content something that should be looked at as a best practice by businesses? Again, the answer is probably ‘it depends’. It may depend on the type of content being created. Is it a standard marketing email or thought leadership content the company publishes as evidence of its knowledge of the industry? Those seem like two common but very different scenarios where disclosure might seem appropriate in one but perhaps unnecessary in the other.
A Few Different Kinds of Content and Possible Approaches to Disclosure
Authored/Byline Content - If an individual is putting their name on an article, white paper, story or book, the reader has a reasonable expectation that the named author actually wrote it. So, if AI was used beyond a minimal extent, and certainly if it is used extensively, this is a situation where disclosure seems to be in the public interest and lack of disclosure might even be considered deceptive. But again, where do we draw that line of involvement?
Contracted Content - How about if you are a marketing consultant or agency contracted by a client to write content for them. If it isn’t detailed in the agreement, it might be reasonable for the client to assume a professional copywriter is creating the content. But, if the agency is using an AI chatbot to write the content, that might need to be disclosed in the contract. In some instances, non-disclosure could represent a breach of contract. So, at the very least, in a situation where content creation is being contracted, the use or non-use of AI should likely be described in the agreement.
Generic Marketing Content - What about the marketing language used in your next sales email? If that was created by AI, would you need to disclose it to the recipients? In this case, people who receive your marketing email have no clear expectations about how the content was created. Does it even matter to the consumer in a case like this, unless your brand positioning somehow suggests or relies on the idea of original, authored content?
The Extremes and Where the Line Exists in Between
From conversations among OI members and with other people in the industry, the general consensus I’ve found is that the line for disclosure exists somewhere between two extremes.
Spelling and Grammar
On one hand, content creators have been using software tools that now fall under the auspices of AI for years and never felt the need to disclose that usage. Spell check and grammar review software are often encompassed in current AI content generation solutions. So, if you use one of these programs to check the spelling or basic grammar of the content you’ve written, do you need to disclose that?
I think the vast majority of us would agree that this doesn’t fall under the umbrella of AI usage that needs to be disclosed. Writers have long been using software to catch misspellings, typos, and grammatical errors without ever feeling the need to let their readers know that it wasn’t a human editor handling this function for them. Students have also been using these tools on their papers for high school or college without concern that the professor or the university will consider this to be cheating. I don’t think many readers feel the need to have this information either, as most likely assume this is the case anyway. Using spelling and grammar-checking software is something nearly everyone uses in their personal content creation (writing emails, etc.).
Complete Content Creation from a Prompt
On the other extreme, you could simply enter a prompt into ChatGPT or another AI tool, asking it to write an entire piece of content. If you then just take that content and publish it, should you disclose that it was written by AI as opposed to something you wrote yourself?
This situation feels fairly clear as well. It feels deceptive to pass off content that was 100% created by an AI tool as something you wrote yourself. I’m pretty sure this will get you an F in English class and it seems unfair to claim credit for content that you essentially had no part in creating, other than providing a basic prompt to an AI tool to write it for you.
The In-Between
If we can agree there are cases where disclosure is clearly unnecessary and also where it feels required (at least from an ethical perspective), then the challenge is figuring out where in the middle you reach the line. While it almost certainly depends on the type of content, when it comes to content with an author’s name on it, there is a point at which AI involvement becomes so significant that it should be disclosed. Consider these common uses for AI in author-credited content creation and see where you feel the line might exist.
- Brainstorming a topic to write about
- Suggesting potential titles or subject lines for the content
- Creating an outline for the content
- Providing a list of character names for a fictional story
- Creating a general plot or content flow from beginning to end
- Rewriting sections of content written by the author
- Rewriting the entire piece of content written by the author
- Creating sections of final content from prompts defined by the author
- Writing an entire piece of content based on prompts, which is then reviewed and revised by the author to put it into their voice before publishing
- Writing an entire piece of content based on prompts which is then published by the author, with their name on it
Did you find a point where it felt clear that disclosure seemed appropriate? What made you decide the line had been reached? Was it just a gut feeling or something more logical, such as feeling that AI had contributed a certain percentage to the final product? Even in this exercise, I’m fairly confident that if 100 readers gave their opinions on where the line was reached, there would be a range of answers. While we can likely agree on the extreme edge cases, once we get into the middle ground, it all becomes much more subjective.
Well, I certainly haven’t done a great job of defining where exactly the line exists with AI disclosure. But, for me the process of writing this article did at least help me find where the line exists in my mind, and I hope it’s done the same thing for my readers. That is a good starting point for what’s to come regarding AI content generation in the months and years to come.
Photo by Nahrizul Kadri on Unsplash