AI writing detection

Your readers already know what AI prose looks like

"Delve," "robust," "comprehensive." These appear 5 to 20x more in machine text than human writing.

Tier 1 words: replace on sight

AI wordReplace with
delve / delve intoexplore, dig into, look at
landscape (metaphor)field, space, industry
robuststrong, reliable, solid
comprehensivethorough, complete, full
leverage (verb)use
seamless / seamlesslysmooth, easy, without friction
cutting-edgelatest, newest, advanced
pivotalimportant, key, critical
utilize / utiliseuse
holisticcomplete, full, whole
actionablepractical, useful, concrete
impactfuleffective, significant
empowerlet, help, enable
streamlinesimplify, speed up
harnessuse
fosterbuild, encourage, support
in order toto
due to the fact thatbecause
serve asis
learningslessons, findings, takeaways
best practiceswhat works, proven methods
synergy(describe the combined effect)
thought leadershipexpertise
deep divelook at, examine, explore
game-changer(describe what changed)

One is fine. Two in the same paragraph signals a pattern.

Structural patterns

PatternWhat to do
Em dash overuseHard cap: zero. Use a comma or period.
Bold overuseMax one bolded phrase per section.
Rule of ThreeAI always lists three items. Sometimes two or seven is right.
Uniform paragraph lengthVary deliberately. Mix one-sentence and multi-sentence paragraphs.
Synonym cycling"Developers... engineers... practitioners... builders" in one paragraph. Pick one word.
"It's not X, it's Y"Once is fine. AI repeats it every other paragraph.
Copula avoidanceAI writes "serves as," "features," "boasts" instead of "is" and "has." Default to "is."

Chatbot artifacts: P0 credibility killers

ArtifactExample
Sycophantic openers"Great question!" / "Excellent point!"
Helpful-assistant closers"I hope this helps!" / "Feel free to reach out!"
"Let's + verb" transitions"Let's dive in!" Start with the point.
Chain-of-thought leakage"Breaking this down..." State the conclusion.
Cutoff disclaimers"As of my last update..." Never in published copy.

The severity ladder

SeverityWhat it catches
P0: credibility killersChatbot artifacts, vague attributions, significance inflation
P1: obvious AI smellTier 1 words, "Let's" transitions, bold overuse
P2: polishUniform paragraph length, copula avoidance, synonym cycling