Exciting times: AI is producing more coherent and human-style copy, saving everyone time and effort! In just three years, tools like ChatGPT, Google Gemini and specialised marketing platforms have put power in everyone’s hands to create media from simple texts to graphics, to multi-part copy concepts and campaigns, including video executions.
It’s a wonderfully democratic resource that’s certainly improving the spelling, grammar and general coherence of much that lands on social media, LinkedIn, in our inboxes and through other digital and traditional content channels. The downside is that AI is prone to feeding itself and thereby creating a homogenised best practice writing style, unless it’s guided thoughtfully and edited carefully.
Does this matter? In my view, it depends whether the AI style is off-putting. AI generated content can be off-brand, repetitive and bland. This can be subtle or jarring. It’s my belief that B2B audiences are noticing this, as they experience increasing volumes of very similar-sounding posts on core topics, designed to show thought leadership or competitive parity. It’s going to feel increasingly boring, and perhaps even insulting, to be assailed by high volumes of undifferentiated copy from organisations intent on grabbing a share of your attention. AI copy masquerading as a friendly human, but with a few subtle tells, can make you feel like a brand or organisation takes you for a fool.
If you’re using AI to generate copy or reviewing and signing off content created by someone else, it’s useful to be able to identify the hallmarks of an AI-generated text, so you can decide whether they’re likely to affect the impact of your content and give feedback to mask them if necessary.
Let’s take a look at an example of some AI copy on a general topic.
I asked ChatGPT 4.0 “Explain good writing, in roughly 100 words.” It responded:
Good writing is clear, purposeful, and emotionally resonant. It shows rather than tells, using vivid imagery and precise language. For example, instead of saying “She was sad,” good writing might say, “She stared at the cold coffee, unmoving, as the rain traced slow lines down the window.” Good writing also has rhythm—sentences flow naturally, varying in length and tone. It avoids clichés and chooses words with intention. In persuasive writing, strong arguments and a confident voice matter. In storytelling, authentic characters and subtle tension engage the reader. Ultimately, good writing leaves a lasting impression through clarity, connection, and craft.
- Repetition: The text repeats the same point several times, saying very little with many words.
- Lack of engagement: The text is accurate and informative but it doesn’t make any real connection. It sounds like a textbook. This may be fine in some situations, but creating emotional engagement is important when you want to inspire action, interest or loyalty.
- Structure: Although AI is generally accurate and clear in its style and grammar, it tends to over-use the long ‘em dash’. It also leans hard on the ‘rule of three’ in lists of nouns and adjectives. This is a well-established and effective device, but it’s easy to overwork it. In our example, nearly every sentence is either divided into three chunks or has three concepts.
- Starts and finishes: Pay particular attention to clichéd style in opening and closing sentences (or paragraphs in longer texts.) A humdinger in the example above is the use of “Ultimately” to signal the concluding sentence. It’s true that individual writers often have their favourite tics of this kind, but they’re better at mixing them up.
- ‘Uncanny Valley’: This term describes phenomena in life that are almost human, but not quite. People tend to react to them with fear or anxiety. In our excerpt, I’d say that the example of conveying sadness creatively doesn’t quite work. It feels more annoyed or oppressive than sad. That’s a subjective response, but the effect of these micro-misses will build for any individual, as instances they personally react to.
- Lack of personality: AI can’t draw from experience; it repurposes data and practice in a clinical way. It doesn’t add personality and original anecdotes to writing. Instead, it seeks to explain concepts neutrally. This means that it can’t effectively imprint a distinctive brand, voice or opinion.
Of course, by working on an initial draft provided by AI you can improve the style. You can feed it other examples and ask it to write in a similar way. You can tell it what you don’t like. You can refine your prompts and specify structures to avoid. You can ask it to use British rather than American English. There’s no doubt that AI research and copy drafting can help to save time, but it needs skilled and confident direction if you want to maintain the impact that good quality human-generated copy can have. Unfiltered AI outputs might fool some of the people, some of the time… but it may well disengage many others.
The best antidote to the homogenous AI voice is to be specific about the language you want it to use and to make edits to both the style and content. Unless you’re happy with a completely neutral tone, you’ll need some skilful human intervention to bring the text to life for your intended audience and impact.
Photo credit: Eric Krull via Unsplash