Why AI isn't the end (or answer) to everything

"One of the problems of deep learning is its lack of explainability. Deep learning models are big and configured automatically, making it hard to know how they come up with their outputs." -Emmanuel Maggiori, AI Consultant, from his book, Smart Until It's Dumb

Whether its use cases are heartfelt or hype, artificial intelligence is here to stay, although not as a technological replacement for writers. The emergence of generative AI and the introduction of some clients into their workflows instigated my investigation.

• How accurate are these AI tools (and AI detectors )?
• How ethical are they in terms of data sourcing and energy consumption?
• What makes human content....human?
• And, of course, how do AI-written tools affect search rankings?

So, I conducted a brief study of about 300 keywords and their top 10 results. I looked at a cross-section of them before and after the March 5th update, too, to evaluate what had changed.

AI content doesn't guarantee ranking

According to my findings, 0.46% of the collected links ranking in the top 10 results for their keyword were 100% likely to be AI-written. While 90% of writers and marketers I surveyed used AI (mostly for editing and content repurposing), using this technology alone isn't an indicator of content success.

More AI-like links also moved down in their rankings than moved up after the March 2024 update.

Much of this may stem from how AI processes information versus the human brain.

To be honest, I got pretty sidetracked in the research on their very topic. I already knew that modern AI uses large language models (LLM) to pull information from. I discovered that both AI and the human brain use probability to determine relevancy - at least to some extend.

But plausibility and wider world context are strictly human. Understanding context is what enables humans to still conduct better research and fact-check assignments, among other writing-related tasks. Human writers also win out on more complex and in-depth reporting(1).

Knowing this, it's easier to cut through the hype regarding AI writing tools, allowing marketers and writers to spend their technology budget on the most effective tools.

For example, while humans excel at research and fact-checking, AI does not. However, there are some research tools that provide significant exploratory power and can reduce the need to fact-check. I've started using Scite Assistant to scour journals and articles for full citations, which I use in addition to industry associations and client interviews, to rapidly access data.

SEO AI and automation tools, in particular, can be extremely helpful as well. Editors, such as Grammarly, aren't always correct but offer writers some initial feedback on potential problem areas.

It's better to leave out AI when it comes to:

• Finalizing strategy
• Drafting
• Sourcing/citing data

Since AI cannot explain itself, not to mention its lack of specificity, often poor sourcing, and potential plagiarism risks, it's better to leave the complex decisions to people who can explain them - the marketers and writers behind the desk.

If you'd like to read the longer report, you can get it (no email required) on my website .

And, of course, I'd love to hear your thoughts on how you are using AI (or not)!

Citations

1. Sundar, S. and Liao, M. (2023). Calling bs on chatgpt: reflections on ai as a communication source. Journalism & Communication Monographs, 25(2), 165-180. https://doi.org/10.1177/15226379231167135