Become an expert at spotting AI writing and know when to use it
As AI develops quickly, so does its language. Here is how to spot where AI has been used, and a publisher’s ‘all you need to know’ guide on how to choose wisely when and where to embrace it.
Publishers are facing a genuinely new pressure around AI authorship. While AI written content is becoming more similar to human written text, a valuable skill is to be able to tell the difference between AI and human writing.
It is empowering to be able to read, and identify AI writing. Understanding the world where information seems endless and at times dense, this article strives to make that world more legible.
The Economist recently published a guide to navigating these uncertain waters - how to detect AI written articles. Building on the growing body of literature on recognising AI generated content, I want to pivot this discussion to the context of publishing. I’ll first outline a few methodical ways to identify AI writing, before moving on to how publishing benefits from a more sophisticated type of collaboration with AI.
Only a few months ago, we thought em dashes were an obvious marker of AI-written content, and now this is no longer really true since AI models are using them less. But don't worry, there are other tell-tale signs we can turn to for this:
AI’s grouping of buzzwords
Buzzwords grouped together like ‘streamlining your workforce, pushing boundaries and overall significantly impacting market sales’ bundled together in threes that are often followed by a semicolon.
Negative parallelisms: ‘Not just X, it’s Y’
This clause structure is commonplace. AI does this to make what is sometimes a plain, simple statement seem more profound and more poetic than it is.
Language and lexis choice
AI models prefer scientific, academic wording to sound more credible. Also, its style tends to favour using -ing constructions to describe future impact without drawing on any specific evidence or example. This may look something like this: “Publishers embracing AI stand to unlock transformative growth, streamlining operations while driving unprecedented engagement across the reader journey”. A human written sentence would look more like this: “It would help if publishers used AI-generated analysis to focus on where their sales are dropping, and to eventually improve readership”.
Structure of body
AI generated text also tends to posit an artificially balanced argument, with consistently sized paragraphs throughout. Human writing shows far more variety in paragraph size, fluid transitions and willingness to take a side rather than spotlighting a neutral, overly diplomatic tone with consistent paragraph breaks to visually punctuate ideas.
Rhetorical questions are also one of AI’s favourite ways to portend an engaging, lively tone of voice.
So now you hopefully feel more familiar with AI written styles, and more equipped to detect it if need be. This is useful in helping your writing feel personable and distinctive amidst a sea of increasingly AI-shaped texts. Most importantly though, Inkara exists to help publishers use AI to their own advantage. We consider our technology a collaboration with AI, where our clients learn how to use it efficiently. With knowledge comes power, and pairing your expertise with the Inkara technology will help make navigating this sometimes intimidating world of Artificial Intelligence a lot less daunting.
What matters most is the type of writing that AI is delivering, and, more importantly, where to best use it. Knowing when to use AI in literary spaces is down to identifying and trusting where it is useful and similarly where you should privilege human authorship.
The current environment of written communication and expression is swimming with AI shaped text as well as human writing. This cocktail of AI and human writing makes for a confusing information environment, and so developing your confidence with understanding where to use AI writing, and being able to recognise it for when it appears is paramount. Readers are becoming suspicious, and it is crucial to rely on human authorship where it matters.
Current research does not confidently reveal any reliable statistics showing how LLMs are used in terms of writing. Future studies should look to reveal the percentage of LLM prompts used to write administrative text, like email correspondence and summarising meeting notes versus drafting creative prompts for original thought such as speeches and novels.
As a recent article written by The Economist explores, it might be ‘churlish’ to snub AI’s competency for writing administrative work. For practical, tedious tasks such as writing emails, proofreading documents, sourcing articles and summarising meeting notes, AI equips us with a valuable fast track to success. This is where we should let AI take the lead. This creates more time and space for original human thought to take the stage.
The most important thing that we, and The Economist, recommend is AI disclosure. Inkara has taken a particularly committed approach to transparency with our AI usage. With a guide detailing our specific AI usage and what Inkara’s use of AI actually does, and each article specifying where AI has been used, and where it hasn’t, no user nor audience member will feel confused as to who has wielded the pen. This is key, as meaning is derived from when we know a human has written something. As put by The Economist, ‘making sense of a poem relies on the promise a human wrote it, and thus that it meant something to someone’. Meaning is exclusively human. These steps help us to derive and safeguard meaning in your work and your writing.
Let’s not underestimate the power of expressing honesty. When it has been useful to use AI for research, proofreading, all enriching what you want to say and to express - your original thoughts and advice are best labelled as your own. Declarations as such are powerful, and I think this the most effective way to collaborate with AI.
References
