Which AI Writing 'Tics' Actually Hurt Engagement — and What Marketers Should Do About It

Which AI Writing ‘Tics’ Actually Hurt Engagement — and What Marketers Should Do About It

Search Engine Land’s analysis of more than 1,000 content marketing pages offers a clear-eyed look at which AI-produced writing habits are actually driving readers away — and which are being unfairly demonized. As the piece notes, “The problem with these debates is that they often confuse taste with performance.” (Adam Gnuse, Search Engine Land).

The study finds that not every perceived “AI tell” harms user engagement. Some common patterns—like overused em dashes—are surprisingly neutral or even slightly positive for engagement, while others—most notably repetitive constructions such as “not only… but also” and generic “Conclusion” headers—correlate with higher bounce rates. That means editors should be surgical, not reflexive, when polishing AI-generated drafts.

Key findings marketers should know

Tracking stylistic patterns per 1,000 words across a diverse dataset, the study highlighted three practical takeaways:

• Repetitive concessive phrases (e.g., “not only… but also”) show a measurable negative correlation with engagement; excessive repetition can feel formulaic and push readers away.
• Articles that use explicit “Conclusion” headers often perform worse—these headers had the strongest negative correlation in the dataset—likely because they signal formulaic wrap-ups or lower-quality content.
• Em dashes—often blamed as an “AI signature”—were the most common tic but showed a slight positive correlation with engagement, suggesting punctuation alone is not a reliable proxy for quality.

Why some tics matter and others don’t

Two forces explain the results. First, many AI “tics” reflect common human writing patterns because AI models are trained on human text. The presence of a pattern doesn’t automatically mean the content is low quality. Second, readers react to usefulness and clarity. When a repeated construction undermines clarity or reads as formulaic, it damages engagement; when a stylistic choice (like an em dash) helps nuance or flow, it can help.

As NPR observed in its coverage of the em dash debate, writers are split between abandoning the punctuation and reclaiming its purposeful use. Susan Lovett told NPR, “It’s like it’s the only piece of punctuation they’ve learned other than a period.” That blunt observation highlights why editorial judgment matters: punctuation can add personality when used well, but it can also make machine-produced text feel mechanical when misapplied.

Actionable steps for content teams

Turn the study’s insights into a short checklist editors can apply to AI-assisted drafts:

1. Scan for repetitive constructs and vary them. Use simple find-and-replace or pattern detection to flag repeated concessive phrases and rephrase where they add little value.
2. Treat “Conclusion” headers as a red flag. Rather than a boxed-off conclusion, fold final insights into analysis or add a fresh takeaway that deepens the reader’s understanding.
3. Use punctuation intentionally. Em dashes aren’t inherently bad—use them to improve rhythm and clarity, but remove them when they create run-on sentences or confuse meaning.
4. Prioritize usefulness and clarity over policing “AI style.” If a phrase or punctuation improves comprehension or adds explanation, keep it. If it’s formulaic or repetitive, edit it.
5. Implement a lightweight QA pass for AI drafts. A short editorial workflow—scan for repetitions, check section headers, and ensure sentence variety—delivers outsized improvements in engagement.

Implications for SEO and user metrics

Engagement signals—time on page, bounce/engaged session rate, scroll depth—affect how content performs in search and how users convert. Small editorial improvements that make content feel more human can increase time on page and reduce early exits, which supports both ranking and conversion goals. Rather than obsessing over whether content is produced by AI, teams should measure and optimize for the behaviors that matter.

Search Engine Land’s study reframes the conversation: stop treating style police anecdotes as editorial doctrine, and start measuring the actual reader response. As the study concludes, don’t rewrite content just because someone declared a phrase “AI writing.” Write for reader usefulness and clarity above all (Adam Gnuse, Search Engine Land).

Attribution: This article is based on Adam Gnuse, “The AI writing tics that hurt engagement: A study,” Search Engine Land, Feb 25, 2026. Additional reporting: NPR, “Inside the unofficial movement to save the em dash — from A.I.,” Nov 10, 2025.

Original Search Engine Land article: https://searchengineland.com/ai-writing-tics-engagement-study-470051