How to write manually and avoid AI flags - Essential tips for content writers

As content writers, we've all faced challenges in our careers, but the advent of AI content checkers has introduced a new layer of complexity that many of us did not anticipate. Recently, a fellow writer, Craig Brett , shared a frustrating experience where his meticulously crafted content was flagged as AI-generated by an AI checker. This incident highlights the growing issue that AI content checkers are becoming more of a hindrance than a help in our industry.

Craig's content was flagged as 43% AI-generated by Quillbot, despite being entirely human-written. In an effort to prove his authenticity, he even offered to rewrite the article while recording his screen, only to see the checker flag his content once again. This inconsistency raises serious concerns about the reliability of these tools and the impact on our livelihoods.

Why AI content checkers are misleading - educating my fellow content writers

Structured Sentences

AI checkers tend to flag well-structured sentences as AI-generated. For example, Craig's content had syntactic complexity and varied sentence constructions, which are hallmarks of good writing, yet these attributes were marked as AI-generated.

Tip: While maintaining clarity is essential, try to vary the sentence structures within your paragraphs to introduce a more natural flow.

Long Sentences

Sentences longer than 13 words are often flagged by these tools. In my own experience, I wrote detailed, research-driven content only to have it flagged as AI-generated. It's disheartening to see hours of hard work and personal insight reduced to an algorithm's false positive.

Tip: Break longer sentences into shorter ones where possible without losing the intended meaning or detail.

Minor Punctuation Changes

In Craig's case, adding two full stops in two sentences reduced the AI-generated percentage from 43% to 0%. This highlights how minor changes can drastically alter the results, questioning the reliability of these tools.

Tip: Introduce slight variations in tone and style, perhaps by including more conversational phrases or varying the formality of the language used.

Detailed Explanations

Providing detailed explanations and thorough elaboration on points is often flagged by AI checkers. AI-generated text tends to be overly detailed and methodical, so human writers need to balance detail with conciseness.

Tip: Balance detailed explanations with more concise statements and vary the way information is presented to avoid a repetitive pattern.

Repetitive Patterns

Repetitive sentence constructions and patterns are common flags for AI-generated content. AI models often follow specific templates, leading to repetition that can be mistaken for AI generation.

Tip: Use varied sentence starters and mix up the structure of your sentences to avoid repetitive patterns.

Lack of Errors

AI-generated text is often free from grammatical errors, typos, or casual mistakes, which might not always be the case in human writing, especially in first drafts or casual settings. We have a habit of proofreading, using Grammarly, Quillbot, and multiple other grammer checkers to ensure we have our work proofread other than our human eyes. The absence of these small, natural errors can lead AI checkers to flag content as machine-generated.

Overly Comprehensive Coverage

The tendency to cover all possible aspects in a very comprehensive manner is another trait flagged by AI checkers. Human writing might miss out on some details or not cover every single aspect as systematically.

For example, when I write a piece of content, I ask gpt to critically analyze the content from a specific perspective and tell me where does it lack. It gives me a list that this is what could have been better and this is what is better. To be clear, if I am writing a cold email, I tell all the demographics, that this email is meant to be sent to decision makers and that what would a decision maker of a specific company would like to see in this email that would intrigue him to buy our services.

Based on the analysis, I fix the issues, which, at the end results in being flagged AI because its structured in a comprehensive manner.

My personal encounter - Misunderstanding due to AI checkers

In a similar vein, I had a client who doubted the authenticity of my work based on an AI checker’s report. Despite my repeated assurances that the content was original, they refused to believe me and eventually terminated our contract. This experience was not only humiliating but also financially damaging, as I was not compensated for a month's worth of work. The frustration and helplessness I felt were overwhelming, knowing that my dedication and hard work were being questioned by an imperfect tool.

Adding to this, another client flagged my content as 51% AI-generated.

The irony was that the content was purely derived from voice notes in Urdu, which I had translated into English. Despite this manual process, the AI checker failed to recognize the human effort and flagged it inaccurately. This example underscores how AI checkers can misinterpret genuine human content.

Why the below content was flagged AI?

The highlighted text in the image above shows content that was flagged as AI-generated, even though it was translated from spoken Urdu. Let's break down why AI checkers might have made this mistake.

• AI-generated text often maintains a consistent tone throughout. In this document, the tone is formal and highly consistent, which might differ slightly from the natural variation found in human writing.
• The text lacks a personal or subjective touch, which is common in human writing, especially when recounting events or personal experiences.

• The highlighted text is very logically organized, with clear enumeration and categorization (e.g., listing points 1 to 14). This type of structured format is typical of AI-generated content, which tends to follow a systematic approach.
• Each point is elaborated with detailed explanations, often repeating the same structure (e.g., stating a problem, explaining its importance, and suggesting a solution). This methodical and exhaustive detail is characteristic of AI-generated text. However this is funny because it was exactly the way the speaker spoke in the voice notes.

• The language is highly formal and technical, using precise terminology without any colloquial expressions or idiomatic phrases. Human writing, especially from voice notes, might include more casual language or regional expressions.
• There is a repetitive pattern in sentence construction. For example, many sentences start with “The stage manager did not…” or “The risk assessment was…”. This repetition can signal AI generation as AI often follows specific templates for sentence construction.
• The text is free from grammatical errors, typos, or casual mistakes. Human-transcribed notes, especially from voice recordings, often contain some minor errors or idiosyncrasies.
• The text covers all possible aspects in a very comprehensive manner. Human writing might miss out on some details or not cover every single aspect as systematically.

Conclusion

AI checkers are not infallible and often misinterpret high-quality, human-written content as AI-generated. This can lead to misunderstandings with clients and damage to our professional reputations. It is crucial for content writers to understand the limitations of these tools and to educate our clients about the nuances of AI detection. By standing together and sharing our experiences, we can push back against the over-reliance on these flawed tools and ensure that our hard work is recognized for what it truly is—authentic, human creativity.