How to write well with AI without losing your voice
Use AI Like an Editor, Not a Ghostwriter
I previously wrote a post titled: “When AI Writes the Article, I Unfollow: Three reasons copy-pasted AI writing erodes trust”. In that article, I shared my personal reaction to posts that are clearly generated entirely by AI. Beyond the issue of authenticity, unverified AI writing can also introduce inaccuracies, vague statements, and misleading information.
In this article, I want to take a more constructive approach. I will highlight some of the common signs that a text is heavily AI-generated and share practical ways to avoid them. More importantly, I will show how AI can actually improve your writing when used thoughtfully.
I have taken several excellent online courses from Jules White at Vanderbilt University on the platform Coursera. Some of his lectures are available on YouTube, but his full courses dive much deeper into how large language models (LLMs) work and how to use them effectively.
In this post, I’ll outline common AI writing pitfalls to avoid and share practical strategies for using AI to brainstorm ideas, fill knowledge gaps, verify facts, improve transitions, and polish your writing. I’ll also cover how to craft prompts that actually produce helpful results.
The Biggest Mistake: Letting AI Write the Whole Article for You
As described in my previous post (“When AI Writes the Article, I Unfollow: Three reasons copy-pasted AI writing erodes trust”), AI writing is recognizable, vague, and repetitive. As we become more familiar with using it, they quickly notice when a text is written primarily by AI rather than by the author. Even individual phrases can give it away.
Readers may not always consciously identify why something feels off, but they often sense that the writing is generic or detached from real experience. This can reduce trust in the author and engagement with future posts. Ironically, the more polished the text sounds, the more suspicious it can sometimes feel.
Even if you designed the content and asked AI to draft it for you, the result often lacks the perspective, nuance, and specificity that comes from real human thinking. The goal, therefore, is not to let AI write for you, but to use it as an intelligent editor and assistant.
Why AI Text Feels Generic (The Cognitive Reason)
The simple reason why AI writing often sounds generic is that LLMs predict the most statistically likely next word based on patterns in massive amounts of existing text it has access to. LLMs do not generate ideas the way us humans do. Instead, they produce the most probable continuation of a sentence. That means AI tends toward the average of everything it has seen.
Average writing is safe and grammatically correct, but it rarely contains original insight. It avoids awkward or unusual phrasing, unconventional ideas, or strong personal perspectives, precisely the things that make writing interesting. Based on this probable writing strategy, aspects of original thinking become flattened. This is why the most effective approach is to let the human generate the ideas and let AI refine the way it is presented.
How to Know if Text Was Written by AI
Once you know what to look for, AI writing becomes surprisingly easy to identify. This also highlights the necessity to make your own writing distinguishable, because others notice the signs too. Here are some common signs:
· Generic introductions
AI often begins with broad, predictable statements such as: “Throughout history…” or
“In today’s fast-paced world…”
AI example: “In today’s fast-paced world, many people are looking for ways to improve their productivity and overall well-being.”
This sentence sounds fine, but it could introduce thousands of different articles. It doesn’t tell the reader anything specific. These openings rarely contain a specific observation or personal insight.
A human example: “Last week, I opened Substack and realized something odd: I could predict which articles were written by AI within the first two paragraphs.”
The human version starts with a specific observation, which immediately creates interest.
· Repetitive phrasing
AI frequently repeats similar sentence structures or key phrases within the same article. The repetition makes the writing sound mechanical.
AI example:
“AI can help writers generate ideas.
AI can help writers improve grammar.
AI can help writers organize their thoughts.”
Human revision: “AI is particularly helpful during early drafting. It can suggest ideas, clean up grammar, and help organize scattered thoughts.”
The idea is the same, but the human version is written more naturally.
· Overly structured lists
Articles may become extremely formulaic: numbered sections, identical paragraph lengths, and predictable transitions.
· Vague statements
AI tends to produce statements that sound meaningful but contain little concrete information.
AI example: “This powerful approach can significantly enhance outcomes and create meaningful improvements in many areas of life.”
Big words, but enhance what outcomes? In what situations? For whom?
Human version: “Using AI to critique your draft can reveal weak arguments you might not notice on your own.”
Now the claim is specific and understandable.
· Confident tone about incorrect facts
One of the most important limitations of LLMs is that they can present incorrect information very confidently. This is why verification is essential.
AI example: “Numerous studies have proven that writing with AI dramatically increases creativity.”
This sounds authoritative but may be completely unsupported.
A careful human writer would instead say: “Some early studies suggest that AI tools can support idea generation, but the research is still developing.”
· Lack of personal experience
AI cannot describe what you personally observed, learned, or struggled with. When writing lacks concrete examples or descriptive subjective experience, it often feels detached.
· “Polished but empty” writing
Sometimes AI text is grammatically perfect but intellectually thin. It reads smoothly but leaves the reader with no new insight.
AI example: “By adopting innovative strategies and embracing new technologies, individuals can unlock their full potential and achieve remarkable success.”
This sentence sounds motivational but contains no concrete insight.
Human version: “If you let AI write your entire article, it will probably sound polished—but it will also sound like everyone else.”
Specific observations tend to be more engaging than generic inspirational big words.
What AI Is Good At and What You Should Leverage
While AI should not write your entire article, it can be extremely helpful for specific tasks.
Some of the most useful applications include:
· Brainstorming angles
If you already have a topic, AI can help generate alternative perspectives or related questions you might explore. You can also find ways to fill gaps or point out angles you have not previously considered.
Example prompt: “I’m writing an article about habit tracking. Suggest five unusual angles that would make the topic more interesting.”
Example output might include ideas like:
• Tracking emotional patterns to help identify triggers
• Using habit data for useful self-experiments
• How habit tracking can become obsessive
These suggestions can spark ideas you might not have considered.
· Generating outlines
AI can quickly suggest potential structures for an article, which can be helpful when organizing ideas. This can even contain additional suggested topics that help you build the story line.
Example prompt: “Here are my notes for an article about writing with AI.
Can you suggest a clear structure?”
AI might propose sections like:
Why AI writing is recognizable
The psychology behind generic text
Practical ways to use AI effectively
You can then adapt or improve the outline. For this post, I referred to the post I wrote last week and listed the topics I already knew I wanted to cover for a successive post.
· Improving transitions
Moving smoothly from one section to the next is often difficult. AI is particularly good at suggesting transitional sentences. This might identify gaps in your story line and allow you to choose from different options to add useful and relevant content.
Example draft: “AI writing is easy to recognize. AI can still be useful. The jump feels abrupt.”
Example prompt: “Suggest a smoother transition between these two sections.”
AI might suggest: “Although AI writing can be easy to recognize, the technology can still be extremely useful when used thoughtfully.”
· Rewriting awkward sentences
If a sentence feels clumsy or unclear, AI can offer alternative phrasing. I find it doing a fantastic job at turning my wordy sections in a much more concise and readable paragraph.
Original: “Using AI can be useful for writers in several ways that help them improve the quality of their writing.”
Prompt: “Rewrite this sentence to be clearer and shorter.”
Possible result: “AI can help writers improve clarity, structure, and grammar.”
· Proofreading
AI can catch grammatical errors, redundancies, and overly complex sentences. It is also helpful to check for logical gaps.
Example prompt: “Please proofread this paragraph and highlight grammatical errors without changing my tone.”
It can identify:
• Repeated words
• Punctuation and grammar issues
• Unnecessarily long or awkward sentences
· Identifying missing arguments
You can ask AI to analyze your article and suggest perspectives you may have overlooked.
Example prompt: “What arguments against my position have I not addressed?
This is extremely useful when writing opinion pieces or essays.”
· Summarizing
AI can help condense long texts into key points, which is useful when reviewing background material. This might be helpful in combination with brainstorming about a topic.
Example prompt: “Summarize the key findings of this article in three bullet points.”
This is helpful when reviewing greater amounts of background literature for your writing. However, if you end up using a specific text, you should read it completely and understand its content.
· Explaining unfamiliar topics
When you encounter a concept, you don’t fully understand, AI can often provide a quick overview to help you orient yourself. This can help you take a deeper dive in a more focused manner.
Example prompt: “Explain how large language models generate text in simple terms.”
You may then research further using reliable sources.
How to Use AI to Fill Knowledge Gaps
One of the most powerful uses of AI is identifying weaknesses in your own writing.
Design your workflow like this:
1. Write the first draft yourself: Start with your own ideas and perspective. This ensures the article reflects your voice and thinking.
2. Ask AI what you might be missing: For example: “What important aspects of this topic have I not addressed?”
3. Ask for opposing viewpoints: This helps strengthen your argument and prevents one-sided writing. An example prompt could be: “What would a critic of this argument say?”
4. Ask AI to identify weak arguments: You can ask the model to highlight sections that lack evidence or clarity.
5. Fact-check claims: Use AI to flag statements that may require verification (i.e. almost any fact statement). However, always verify important claims using reliable sources like scientific publications (never other people’s personal writing or social media science channels – I see many viral accounts post AI-generated text with no evidence that they checked the facts).
6. Prompting Tips That Actually Work
a. Bad prompt: “Write an article about healthy habits.”
b. Better prompt: “I wrote this article about habit tracking. Identify three places where the argument is unclear and suggest improvements without changing my writing style.”
i. This works better because it gives context, asks to complete a specific task, and preserve’s the author’s voice
My Personal AI Workflow
1. I Write a draft myself. I begin with my own ideas, structure, and examples.
2. Ask AI to give me feedback on the structure and flow, e.g. “Does the argument flow logically?”
3. I Ask for missing perspectives. This helps identify gaps I may not have considered.
4. I ask to Improve transitions and phrasing. AI is particularly helpful for polishing sentences and improving readability.
5. The final step is always human judgment. I remove anything that sounds generic or does not reflect my voice. I can rephrase, expand, skip, start an additional section based on the current story line, and look over the whole text to decide if I find it complete. It often helps to walk away for a while, many ideas happen as you give your brain a break from the focus.
Conclusion
AI is an extraordinary writing assistant that can help you be faster, more accurate, and more efficient in getting your writing done. But like any assistant, it works best when guided by a human who has something meaningful to say.
The most compelling articles are written from personal experience or perspective. AI can amplify the existing qualities of your writing and ideas, but it cannot replace them.
I am curious about your experience:
Does it matter to you whether a writer used AI, or only how the final article reads?
If you discovered that an article you liked was mostly AI-generated, would it change how you feel about the author?
Where do you personally draw the line between “AI-assisted writing” and “AI-written content”?
What is one task where AI has genuinely improved your writing process?