7 Practical Solutions to Make AI Sound More Human: A Writer’s Guide

Who doesn't love the idea of whipping up entire blogs, academic papers, and a month’s worth of LinkedIn posts in a matter of seconds? When you’ve got deadlines looming large, it sounds like a dream, doesn’t it?

But here’s the problem: most AI writing is suspiciously flat and even, like a steamrolled airfield. There are rarely any bends, curves, or irregularities. 

Add to this equation AI content detectors —  the polygraph tests of the content world — and companies and universities' decreasing trust in human writers. Suddenly, good writing isn’t just about skill, it’s about proving that you’re not a robot.

So, what’s the game plan? Do we ditch AI tools altogether? 

Nope. We pivot. We adapt.

We learn how human creativity can be buttressed by AI’s speed to create emotionally resonant content. (So it doesn’t feel like you’re reading a manual on watching grass grow.)

Welcome to your no-BS guide on humanizing content with AI. 

Let’s get into how you can keep your writing fresh and personable, with a little help from our AI friends.

What are AI content detectors?

AI content detectors are algorithms that scan texts to determine whether something has been written by AI. They’re programmed to search for AI authorship through reverse engineering and predictive patterns — this includes everything from repetitions to unusual word choices to a jagged flow in writing. 

If you’re wondering, ‘But isn’t that sometimes just creative writing?’ you’re not wrong.

Are AI content detectors reliable?

The sales arms of AI detection companies will have you believe that yes, they’re reliable, with one company claiming a rate of only 0.2% false positives.

As a writer and editor of seven years, I can assure you, that in their current capability, they’re compasses with broken needles at best.

AI detectors are getting smarter as more and more information is fed into them, but they’re hardly of the same competence as a human editor. (More on this later.)

In the International Journal for Educational Integrity, a study evaluated the performance of five AI text content detectors: OpenAI, Writer, Copyleaks, GPTZero, and CrossPlag. The findings reveal: 

“The results of this study indicate considerable variability in the tools' ability to correctly identify and categorize text as either AI-generated or human-written…”

Turnitin’s lead AI scientist, David Adamson, has issued a statement clearly warning that:

“You’ll have to take the output from any AI-powered feature from any company with a big grain of salt.”

There are two operative phrases here — considerable variability and a big grain of salt. 

Unfortunately, for a lot of universities, AI content detectors are the new airport security checks: the final line of defense between a final draft and an A grade. 

While AI is in everyone’s digital satchel, false positives are a grave and growing problem.

Now more than ever, trust is the foremost currency in professor-student and company-contractor relationships. This isn’t inherently bad, however, the reliance on nascent tools’ judgment that overrides human testimony is slowly redefining how we relate to each other — accusations, terminations, and a failure to reassess evaluation methods. 

Why was my content flagged as AI?

AI detectors assess content based on two metrics: perplexity and burstiness. 

• Perplexity is the measure of how likely a word is bound to follow another. That is, if your writing moves in a relatively predictable manner.
• Burstiness focuses on sentence variation. AI writes in uniform sentences that are roughly of the same length. Humans, however, write with greater ‘burstiness.’ We like to vary our sentences in speech and writing. Some crisp, some meandering.

Since AI models are trained on statistical averages, they're likely to keep the writing flat, without teeth. 

But as humans, we’re contextually aware sentient beings with the great ability to synthesize emotional information and move our audience whether to action or tears through expressive writing. 

And expressive writing is far from perfect. The more I read and write myself, the more I move away from the traditional rules of grammar and style in my own work. You learn the rules to break them. If none of us ever experimented with punctuation or flow or narrative, some of the greatest novels ever written wouldn’t exist. On Earth We’re Briefly Gorgeous by Ocean Vuong is a particularly fitting example.

These musings, however, often take a backseat when you’re dealing with companies and universities and need your invoices or assignments to clear. 

Here’s why your content might be getting flagged as AI:

1. In my experience as a Brand Manager to C-suite executives, I’ve found that a lot of B2B marketing copy, especially on LinkedIn, relies on technical and tight-collared corporate language — driven by rationale and reason. B2B writing may well be the left, analytical side of the brain, with B2C as its right, imaginative counterpart. 

There are certain words and turns of phrases that AI is partial to in its own generations, and it might be flagging the same for you:

• In the ever-evolving landscape of business…
• The world of content is rapidly evolving…
• Delve into these strategies
• Elevate your brand’s voice
• Synergy
• Woven into the tapestry
• A myriad of problems
• Unlock the potential
• Unleash
• In the realm of 
• A crucial aspect/skill/component
• Resonate
• AI isn’t just another buzzword
• The digital landscape of content creation
• It is important to note…
• Writing is an intricate dance between…
• Embark on a journey
• The dynamic interplay between…

These are perfectly valid words in their respective contexts but they’ve lost their potency with wanton overuse. 

To loosely quote Bilbo Baggins from The Lord of The Rings: “...like butter scraped over too much bread.”

2. There’s also the case of being punished for perfection. No typos, missed capitalizations, or faults in flow? You’re either an accomplished writer or you’ve cheated with AI. There are scores of defeatist posts and comments on Reddit by professionals and students who’ve resorted to making their writing ‘less perfect’ so that they can avoid AI detection. 

Think, intentionally misspelling common words, or even swapping the letter ‘e’ for a Unicode letter. Why? Because AI wouldn't make these ‘foolish’ mistakes, but a tired human can. 

The question is — how do we as writers stand firm in the authenticity of our content while also ‘passing’ newfound metrics of AI content detection?

Humanize AI with AI

There are a few unsavory ways to bypass AI detection from my research, like adding random punctuation marks in the document and whiting them out or alternating between British and American spellings purposely.

For a long-term partnership with your clients and professors built on trust, I wouldn’t recommend them. 

Instead, here’s what you can do to humanize your content. Paradoxically, AI can save you time and double your productivity with a few clever hacks. 

Let’s add some teeth to your content, shall we?

1. Experiment with adversarial co-creation

I’d been reading about adversarial training in AI and realized that adversarial co-creation is excellent for “inter-model collaboration.”When I’m brainstorming a new topic or having a dialogue with myself after I’ve collected my research bits, I like to have an “inter-model dialogue” between ChatGPT, Claude, and Wordtune. ‍

• Let’s say I start with the same prompt on ChatGPT and Claude: “How does an AI writing assistant help writers?” 

• Then, with their respective answers, I feed them as ensuing prompts for more fleshed-out answers in the other model. 

• As a final step, I turn to Wordtune to make sure that my notes aren’t too wordy and have clear instructions for me when I come back to them.

Think of this thought experiment as dialoguing with someone who always challenges your opinions. This is an excellent way to beat monotony of thought, weed out unoriginal ideas, and therefore, sieve unoriginal content. 

‍

2. Get help from an AI text humanizer

This is how you beat the system. You could’ve written the words yourself, but if your professor or client isn’t convinced, it’s time for the battle of the detectors.

I run my text through a maximum of three AI content detectors, but I find that two are mostly sufficient. 

Most detectors will highlight the phrases that sound AI-generated. 

The next step is to have Wordtune rephrase the highlighted sections of the text according to your style and tone of voice. For a more personalized touch, I’ve configured the ‘Brand tone’ to suit my writing style — confident, conversational, and irreverent.

Based on your style, you’ll also vary the sentence lengths to convey a combination of emotion and logic in your writing. This ensures that your content isn’t flagged under the ‘burstiness’ AI score. 

3. Choose analogies over metaphors

If you’ve been using ChatGPT for a while, you’ll notice that it loves dance metaphors for some reason. Everything is an intricate and delicate dance between x and y. 

On the other hand, analogies are more explanatory by nature and can break the robotic hypnosis of flat writing. Metaphors bend toward the poetics, something that most LLMs (large language models) are terrible at. 

Let’s look at an example between ChatGPT vs. Wordtune. 

The prompt is: “AI is a new tool that’s important for a writer’s productivity.” 

ChatGPT: “AI is the wind beneath the wings of a writer’s creativity….” (An immediate no for me.)

Wordtune: “(AI is) like having an extra set of hands…” (Much better.)

The first generation is bound to be flagged by an AI detector because it’s non-specific, and doesn’t solve for a problem by clarifying further — it complicates the subject matter with philosophical overtones that are neither original nor creative.

‍

4. Use adverbs sparingly

Hemingway and King are famous grouches when it comes to adverbs, and to a certain extent, I’d nod my head along. The purpose here isn’t to burn off adverbs altogether from our writing vocabulary, rather to make intentional and judicious use of them. 

And to humanize your content, it’ll do you well to jettison superfluous adverbs. 

‍A lot of them end in -ly so they’re easy to spot.

‍If an adverb doesn’t add specific or significant information to your sentence, delete it and replace it with a stronger verb or adjective.Take a look at these two sentences:

• They ran very quickly towards the car.
• They sprinted towards the car.

The second one reads crisper, doesn’t it?

‍

5. Employ contractions

A lot of technical and corporate language forgoes the natural cadence of everyday language in favor of inaccessible acronyms and sales-first jargon as a pretext for professionalism. This is another reason why your writing might be flagged as AI. 

A simple fix is using more contractions in your content:

“We are going to improve our workflow.”

“We’re going to improve our workflow.”

‍

6. Integrate natural collocations

A collocation is an easily recognizable pair or string of words that is familiar to native speakers as part of their everyday language.

This is a classic write-like-you-speak hack and a tried-and-true solution to humanize AI content. When you’re writing or editing your content, pay special attention to how the words sound when you read them out loud. Are they natural or forced? 

Here are some examples of collocations:

• Fast food
• Make the bed vs. do the bed
• Do the dishes vs. make the dishes
• Blurry vision
• Listen closely
• A quick shower

The comprehension of collocations relies heavily on our contextual intelligence: our ability to determine the ‘who,’ ‘where,’ ‘why,’ and ‘when’ of situations that guide our decision-making. As smart as AI is right now, the nuance of human learning, and therefore, communication, is beyond compare. 

Let’s fix a mismatched one:

He’ll revert back to me on Friday ➞ He’ll get back to me on Friday 

7. Become an AI editor

There’s a new wave of AI-optimized writer roles, and here’s where you get comfortable with the duo of AI detection software and AI writing assistants to prevent false positives. (With Wordtune, you get both.) You’re going to want to swap words to suit your voice, rewrite sentences, and add anecdotes and examples.To humanize AI content further, follow these steps:

• Revisit your brand guidelines or past work and look for words you commonly use. Then, swap them in your current content. Also, swap words you wouldn't normally use. For example, I don't like using “super” as a descriptive adjective.

• A lot of LLMs, like ChatGPT, typically use the same verbs as starting points. Like this: “Elevate your brand’s tone of voice with Wordtune.” 

Rewrite these kinds of sentences to reflect your verbal identity:

• Wherever you can, include personal pronouns and active voice. For example, “I’m writing a piece based on my personal experiences on the fallacy of AI content detectors in evaluating human writing.”  vs. “This piece delves into the fallacy of AI content detectors based on a writer’s experience in evaluating human writing.”

An editor’s role is far from janitorial — you’re not just making the draft look ‘clean,’ you’re making sure that every word in your content corresponds to a larger WHY. 

This is why it’s important to vary sentence length to bypass the metric of burstiness, and include industry-specific knowledge to throw off AI’s perplexity meter.

‍Also, it makes for good writing overall. Win-win.

A note on AI-powered human writing

I’ve come to see AI as a sidekick in my plot line. In marrying our humanness with AI’s capabilities, the conversation goes much further than slipping past the new digital sniffers.

As you take on the role of an AI editor, you'll realize that your knowledge, experience, and emotion can't be xeroxed by an algorithm.

At the end of the day, what people crave is genuine human connection. And as writers, it’s our job to defibrillate our words to give them a pulse. 

A smart toy like Wordtune is the perfect companion to humanize AI content and convey the voice you want to connect to your readers.