When AI-powered language models popped up in marketing, there was a rush by overworked marketing teams to drop an idea or keyphrase into a language AI program, slap in a tone and a target audience, and out popped words for blogs, ads, social, and pretty much everything words. The result is copied and pasted, and the PUBLISH NOW button hit hard.
WONDERFUL! You have shaved hours off your already limited time, and you have an excellent piece of content that says most of the things you need it to say. You don’t really need to do much editing; throw in an image/video, and you’re golden – let your customers rave about your awesomeness and see your rankings surge.
Right?
Not so fast, lightning fingers! Don’t fire your writers just yet.
Anything generated by an AI-powered language model is aggregated from millions of inputs from millions of users with their own agendas. You need to research every statement, sentence, theory, and theme and verify that each line of content is a) True, b) Relevant, and c) Unbiased. That’s a lot of research and time vs. a human doing the research from reputable sources and writing the words.
Unfortunately, AI-powered language models, whether images or text, have been influenced by the inputs of millions of people worldwide, tip-tapping their needs into that language engine. AI-powered language models cannot research and verify. Millions of inherent biases, false information, and personal preferences influence the words popping onto your screen. Do you want to be that marketer with egg on their face from publishing false information or blatantly untrue content?
The London Interdisciplinary School recently published fascinating research on generative AI bias. It highlights how awful (and prevalent) bias is in image-generating AI. It’s a must-watch.
How AI Image Generators Make Bias Worse
Here’s my brief experiment in Bing Image Creator with the task of creating a bank CEO; what do you think? Biased.. Or not?
No thought, understanding, reflection, or editing comes with AI language generation tools.
AI-powered language models also seem to have set formatting for each ask. A great example is LinkedIn social posts liberally sprinkled with seemingly random emojis and bullet points in the same place each post, or those weirdly repetitive tips blogs – and don’t get me started on my soapbox about the tone, repeated sentence structure and cadence of generated content.
Then… there’s grammar. Spelling is generally correct, but I have yet to find an AI-powered language model that can do ‘the grammar,’ unless you are talking about Grammarly – a handy tool! But even that is often wrong.
Oh, and tone? No matter how many prompts you throw in, AI will not sound like your original voice and style – in its current iterations.
So… what do you use??
My first option is to use my brain and the brains of the brilliant content people around me. But, since we are talking AI… I guess I will throw out some more opinions.
Having played around with AI language generation tools – and wished there was such a thing as eye bleach – I will use the free ones as examples.
Bard.ai is the most relevant for my topic research, story outlines, and ideas. You can ask it for the source of information, and it gives you the link so that you can judge for yourself whether it is relevant and take steps to do more research if needed. It also has up-to-date research papers, articles, etc. Here is an article on what to be aware of and how to use it. Just be aware that it is not the tool if you are a one-human marketing team wanting to generate publishable content (spoiler alert: none of them are!).
ChatGPT can help with headline ideas, meta description ideas, basic outlines, and rough draft social posts – but PLEASE change the formatting!! Don’t let those LinkedIn experts try and tell you that ‘If you just get your prompts right and follow my amazing guide to prescriptive repetitive prompts, you’ll produce amazing content,’ that’s rubbish. Its content lacks flair, innovation, and accuracy and has repetitive sentence structures that any seasoned editor can spot with accuracy at least 80% of the time.
But – it’s a writing buddy, not a writer replacement.
The truth is, there is no real shortcut to creating great content. AI-powered language models cannot replace real writers at this stage of their development. If you are a start-up with a small budget, AI-powered language models might be a good start before you hire someone full-time – but don’t make it your marketing team. Please.
AI may be able to replace us one day – but it’s not today.
How do you use AI-powered language models?
For more information about our AI-enabled marketing services, please contact us here.