As I often do, I like to check the submissions from our press releases and blog request inbox. After a quick drink of my early morning tea (a British essential), I noticed something strange.

Each blog has the following style of title “Something about noting: a guide to nothing” Notice the: in the title?

I then proceeded to read down, and bang, I would say 60% of the submitted guest blogs were AI-written; what makes this worse is thstorat the Public Relations agencies submitting are charging customers for each blog written. Think about that: they care that little about the brand they are paid to represent.

Were these blogs interesting? Not really, they were complex, offered not a single broken-down explanation, full of technical mumbo jumbo and at worst lazy.

Why does this matter?

My personal view is that each blog submitted is posted to inform and entertain our audience, it is a primary motive for not sun-setting the blog and allowing it to be a free-flowing exchange of thought leadership ideas.

I asked ChatGPT to give me a description of what it considers to be the writing uncanny valley.

The “uncanny valley” in AI writing refers to a phenomenon where AI-generated text is almost, but not quite, indistinguishable from human writing. This near-perfect imitation can evoke a sense of unease or discomfort in readers. Several factors contribute to this uncanny effect:

Subtle Errors: AI writing might contain small but noticeable mistakes in grammar, context, or style that a human writer would typically avoid.
Lack of Genuine Emotion: AI-generated text can feel mechanical or insincere, as it lacks the true emotional depth and understanding that humans bring to their writing.
Predictability: AI writing can sometimes follow overly predictable patterns or clichés, making it feel less authentic.
Inconsistencies in Tone and Style: AI might struggle to maintain a consistent tone or style throughout a piece, leading to abrupt changes that feel unnatural.
Contextual Misunderstandings: AI can misinterpret context or cultural nuances, resulting in awkward or inappropriate phrases that a human would recognise as incorrect.

These elements combined can create a sense of dissonance for the reader, who might recognise the text as almost human-like but still detect something off-putting or unnatural about it.

I did have to correct the US English, but I left the post intact. I am a heavy Grammarly user, which sometimes causes a written article to be flagged as AI.

Back to my morning tea, I began to realise that AI has some great uses but also some significant drawbacks. The main drawback is simply trust!

The mainstay of human connection Trust

When we adopt the uncanny valley mindset, we then erode the element of trust: trust that the author wrote the blog with meaning, trust that the information is fact-checked, trust that you are not breaching copyright or plagiarising a body of work.

Checks and balances for AI content

1. Always have a human edit, and never believe sources or citations until you have confirmed them.
2. Cite the AI insert above to show transparency to the reader.
3. If you are a Public Relations agency, respect the client’s brand, stop taking shortcuts, and be transparent about using AI.
4. Content spinning and paraphrasing tools have been around forever; just don’t use them; the content just feels “uncanny”
5. The accuracy of online detection tools is somewhat flawed. Get AI to do the work for you and prompt you to make you write better.
6. Be human.
7. Word count, don’t write to fill a word count, perhaps just let it flow and then come back without the need to use filler content.
8. Storytelling, immerse the reader in the story you create, give a piece of yourself away.

As we are now in the maturity stage of chatbots, especially with the excellent Claude 3.5, my advice is to be transparent, open, and cite AI content. If it reads well and adds value, then go for it; if it is jargon-led mumbo jumbo, rewrite it.

As we look forward to chatbots’ continual maturity, the cost of using AI to produce blogging and brand content will fall. With this fall, though, comes the further step of the shadow of the Uncanny Valley. I can spot a blog written with AI within 10 seconds of reading it, and I am sure the majority of people reading this post could also.

To the person who just walked into the boss’s office in a slightly elevated manner saying, “Look what I produced in 1 minute”, think about why it took a minute, and then ask yourself, does this really create an advantage for you and your customers?

I asked ChatGPT to give me a quote to end the blog.

“Time invested in learning and adapting is always worthwhile, especially in a world where change is the only constant.”

It’s attributed to no one.

The post The Uncanny Valley of AI Writing appeared first on Compare the Cloud.