Ally? Or creating bland content: Generative AI in content marketing and creation.

Linkedin-Merchant (definition #16)

Dear Network,

As promised, we are back with the somewhat new flavour of Officially Unemployed. This week, we jump into the extremely relevant topic of generative AI and how content creators and marketers are embracing this technology to optimise their work. We will also explore the (numerous) pitfalls of using these tools in isolation and how, despite the claims of many, human influence remains absolutely necessary and not all of our jobs are completely at risk as creatives!

The Rise of the Robots

Since the 2022 release of ChatGPT, and particularly whilst I was at university, it was more or less impossible to have a conversation about academic matters without mentioning this new cure-all for academic stress. Sensationalist articles from high quality news outlets like The Sun would flood the headlines, talking of students' abilities to create stellar essays simply by writing a prompt into this new-fangled computer programme with the intention of finding the 46th reason for calling our generation lazy, entitled and over-sensitive. This fearmongering spilled over, with jobs as linguists and creatives now seemingly under threat from ignorant people, being told that we'd all be replaced by robots and our lives were pointless. Either way, this technology was and still is everywhere, with Google and AWS among others coming up with their response to ChatGPT pretty rapidly too. This was what got me interested in its applications in content creation and marketing.

Despite the standard overreaction that comes with any new creation like this though, the hype is not unfounded. The manual and repetitive elements of marketing work such as optimising ads, keyword research and discovery, creation of SEO criteria and landing page optimisation can be done by Gen AI, saving huge amounts of time and increasing ROI. Given that these tasks make up the bulk of my daily activities during my marketing internship, it is unsurprising that this technology received the response that it did. The Sun was not entirely wrong either, these programs can generate longform content like essays, blog posts, even stories and song lyrics. However, it is with the generation of this kind of content that the pitfalls of generative AI start to become apparent.

AI Pitfalls

The most obvious of these is the tendency for generative AI to create bland and uninspiring content if prompts are not constructed well. To illustrate this, I have actually posted another blog at the same time as this one that is created solely by ChatGPT with the following prompt.

Whilst I was able to do this in a matter of seconds, there are some obvious differences between my usual content and the post generated by the AI which you will be able to identify, the main one being the tone and (hopefully) the engagingness of what I am writing. For some types of content, the robotic nature of these mass-produced and likely SEO friendly articles are not unattractive and it is no secret that many firms and news outlets will publish articles that are completely written by generative AI. That being said, effective brand activation and successful communication of a particular tone of voice must either be created by humans, whether that means being written from scratch, or time being taken to teach these programmes to understand your desired tone of voice. Going back to the post I generated, whilst there is nothing inherently wrong with it, it does not resemble any of the content I have written previously. You can therefore see how content may fall short if it is not proofread, edited and adjusted by a sentient being. Going back to the university essay, it's worth noting that the person who generated the essay said:

"You definitely can’t cheat your way to a first class degree, but you can cheat your way to a 2:2"

However, the combination of these widely capable tools and a competent human will create content that effectively communicates the desired brand and message, leading to a better ROI and more successful campaigns.

Another aspect of AI content creation that remains blurry today is the concept of ownership. Who owns the content generated by AI? To be completely honest, this article by Reuters explains this a lot better than I can and I encourage you to take a look: Who owns AI created content? The surprising answer and what to do about it | Reuters. Essentially though, there are a plethora of legal issues that can come with content created by AI and the reality is, you may own less of the content than you think if you used AI to help you create it.

Finally, AI gets things wrong. That is a simple fact. First off, we can see that the free to use ChatGPT 3.5 has not been updated since January 2022, meaning that some of the information is outdated and some is not present at all. A human element is required to be able to verify this information and be responsible for preventing misinformation, or the spreading of false information that will ultimately bring you or your brand into disrepute.

How to overcome these pitfalls: a conversation around prompts

So, how do we tackle these issues with Generative AI? A lot of it comes down to how we perceive it. I took David Birss's Linkedin Learning course about this to help better my understanding.

Birss says that the general misconception around GenAI, and the one that was shared by me until recently, is that people felt AI could effectively replace human thought. For the reasons we have discussed already, this is simply not true. It is important to think of AI as part of your creative team and arsenal rather than a replacement for it, a collaborator. Your interactions with ChatGPT should be a conversation, not a simple answer to a query as if it were a simple search engine. Moreover, you have to challenge the output given to you and continue the process in order to refine it.

Starting with the issue of bland and uninspiring content, the best way to kickstart the process is with an effective prompt. This should be a paragraph rather than a simple sentence with specific information to help the AI generate the content you need. Looking at this week's definition and the prompt I used to create the other blog post, it is completely unreasonable for the AI to create the perfect responses. Birss proposes the CREATE mnemonic to help us write effective prompts.

C - Character

Imagine the AI is an actor and you are assigning it a role to play. For example, "You are an experienced copywriter with 20 years experience creating engaging email campaigns surrounding the health and wellness industry".

R - Request

Give context and useful information, the more the better. This could be the USPs of the product you're marketing so it knows what to include or indeed not include.

E - Examples (optional)

One thing that AI is really good at is learning from previous works to come up with a consistent tone and content type. Therefore, including examples of things like previous headlines can be useful.

A - Adjustments

These are refinements that you may put in after the first output is generated such as asking it to (not) use bullet points, add some jokes or use tables, graphs and other formatting.

T - Type

The type of content you want it to produce must be very clear whether that's an article, blog post, email template etc.

E - Extras

Additional instructions such as "ignore everything before this prompt" or "ask me questions before you answer". Birss says that this helps to direct the AI more effectively.

From this framework alone, we can see how a prompt can unlock the huge amount of potential that AI has when used correctly as a collaborator. Being able to write effective prompts will skyrocket your ability to create high quality content fast. It is not a question of reducing your workload, but increasing the quality of it.

In terms of issues around data accuracy, this is where we also have to put some work in. These programmes have access to huge amounts of information of varying quality and, once again, the quality of our input dictates the quality of the output. You can instruct the AI to draw upon specific sources of information and widen or narrow search criteria. The critical thing with this though, is proofreading and verifying the information presented. As well as drawing upon incorrect information, AI will sometimes make things up, especially if you are asking it to be persuasive. You must stay in the loop. You can also achieve accuracy by inputting the data yourself and asking AI to synthesise, summarise or reformat the information presented into your desired outcome.

Lastly, in terms of legal and ownership issues, the most important thing is to be open about whether or not you are using Gen AI. It is a growing reality of pretty much all industries and it is a beyond useful tool that would be wasted if not utilised. That being said, it is important to level with readers, managers and clients that it is being used.

Takeaways

Ultimately, generative AI is a tool with wide-ranging capability that can be effectively harnessed as a valuable asset for creators and marketers alike. However, in order to unlock this potential, our perception of its role as a collaborator must be clear. We need to learn to speak its language and treat our interactions as ongoing conversations to constantly challenge the output we are presented with. We must also be aware of the legal and ownership side of AI usage and remain in the loop to edit and verify the information we present to consumers of our content.

I hope you enjoyed this edition of the new-look Officially Unemployed! As always I'd love to get any thoughts you may have on this matter and I encourage you to read the Reuters article and take David's course! I'll see you in the next one, merchants!

Your favourite prompt-writer,

AT