Ever have something you wrote flagged as being generated by Artificial Intelligence (AI) when it wasn’t? This is happening to lots of good writers lately. There are memes on the internet showing old writing — like the opening to Mary SheIly’s Frankenstein (The Modern Prometheus 1818) — being flagged as written by AI. So what’s going on here?

Let’s go to the source and ask an AI bot! The information here comes from Gemini — the AI by Google.

~~~~~~~~

Query: What are some things to be aware of when a writer’s writing is flagged as potentially being written by Artificial Intelligence when it was not? Apparently this can happen to people who have large vocabularies and are good at spelling and grammar, and it also might be related to using spellchecks.

Answer: This is a real and frustrating problem, well-documented among strong writers. Here's what to be aware of:

Why it happens to good writers?: AI detectors flag patterns common to both AI and skilled human writing: consistent grammar, varied sentence structure, low typo rate, formal vocabulary, and logical paragraph flow. Ironically, the better your writing, the higher the risk.

Specific patterns that trigger detectors:

• Perplexity — AI detectors measure how "predictable" your word choices are. Sophisticated writers who consistently choose precise, fitting words score similarly to AI.

• Burstiness — Humans naturally vary sentence length dramatically (short punchy sentences mixed with long complex ones). AI tends toward uniform rhythm. If your writing is consistently well-paced, it can paradoxically read as AI-like.

• Transition words — Heavy use of "furthermore," "however," "additionally," "moreover" is a strong AI signal. Use them sparingly.

• Hedging language — Phrases like "it is important to note" or "it is worth mentioning" are AI staples. Avoid them.

• Passive voice — AI overuses it. Active voice reads more human.

• Oxford comma consistency — Ironically, perfect punctuation consistency is a flag.

Practical adjustments for job assessments specifically:

• Leave a minor intentional imperfection — a casual contraction, a sentence fragment used for emphasis, a conversational aside.

• Vary your sentence rhythm deliberately — break up polished paragraphs with a short, blunt sentence.

• Use first-person perspective and personal framing where appropriate ("I think," "In my experience").

• Avoid overly symmetrical structure (e.g., three bullet points each exactly two sentences long).

• Don't over-edit. A first-draft feel in places actually helps.

The spell-check factor: You're right that this contributes. Zero spelling errors combined with perfect grammar is a statistical outlier among human writers and a flag. Leaving one or two natural, minor errors — or simply not correcting every informal construction — can help.

The core principle: controlled imperfection signals humanity. You don't need to write worse, just less uniformly polished.

~~~~~~~~

It’s amazing if you think about it. AI is basically cornering the market on good writing. In this response above, it is literally telling good writers that “controlled imperfection signals humanity.” Wow! So if you spent years perfecting your craft, learning to use the right words (and never using a word like ‘burstiness’), using bullet points consistently, polishing your grammar and spelling, you’ll now be flagged as AI and might even lose your job over it. So get stupid!

There’s another aspect to this that’s relevant for writers. Many employers use Applicant Tracking System (ATS) software driven by AI to scan resumes. These systems automatically collect, parse, rank, and filter applications based on keywords and formatting. Job sites recommend using AI to rewrite your resume to get through the ATS stage. There are sites online designed to merge your your resume with a job description so you’ll get past the ATS stage. Some jobs sites now do this automatically and apply to jobs for you with a resume you’ve never proofed. The problem is that resume-writing AIs have ‘hallucinations’ where they add false experience. They lie so you don’t have to! And artificial intelligence knows this is happening. The information below is from Claude AI by Anthropic. For writers specifically, it’s known as falling into the computer science (CS) trap.

~~~~~~~~

Query: Explain the computer science (CS) trap as it relates to resumes and writers.

Answer: The CS trap in practice: AI resume tools are trained on high-performing resumes, which for some online platforms can skew heavily toward technical/CS profiles. So when you ask an AI to rewrite your resume or “modernize" your terminology, it pattern-matches toward the job descriptions and the computer science field — adding acronyms (like RLHF, LLM) and data pipeline language that reads as CS to both automated screeners and human reviewers. Your actual background will get buried under borrowed credibility you didn't need and couldn't defend.

Going forward best practices:

• Rewrite your resume yourself, in your own voice

• Lean into documentation, content strategy, editorial work, style guides, structured authoring — the actual vocabulary of your field

• Terms like "technical communicator," "information architect," or "content developer" are current without being CS-coded

• RLHF and LLM should only appear if you're describing actual evaluation work you've done in those disciplines; they aren’t terms to replace reviewing human work or knowing language models

• Have an AI review your resume for clarity rather than rewrite it

The bigger lesson: AI is genuinely bad at resumes for people whose expertise is adjacent to AI work, because it keeps pulling you toward the technical center of gravity. Reduce reliance on AI for resume writing or it might lead to lots of awkward and unsuccessful interviews.

~~~~~~~~

This explains some of the foundations of the misinformation age we live in. There’s strange goings on in corporate and government spaces these days. People are claiming to be experts when they aren’t experts, and while that’s always been true to some extent, now it’s powered by artificial intelligence. We have AI bots literally telling smart people to be stupider, and telling stupid people they are smarter than everyone else. There are standards and practices in writing resumes that add keywords from job descriptions that are things the applicant never did. Meanwhile, hiring managers express frustration over not finding qualified candidates, while qualified candidates don’t get interviews and see idiots doing important jobs.

Perhaps the best advice from an AI on the use of AI is to “reduce reliance on AI”.