UpTrajectory Review

Russell Brandom at TechCrunch digs into the linguistic fingerprints that betray AI-generated copy, focusing on a single word: 'dependable.' According to the research he cites, Opus 5.5 uses this adjective 23 times more frequently than human writers do. That is not a subtle statistical blip. It is a glaring, almost comical marker that anyone who reads enough AI-assisted content will start to notice everywhere. The piece arrives at a moment when businesses are flooding the internet with machine-drafted marketing copy, product descriptions, and blog posts, often without disclosure. For small-business owners who have experimented with AI writing tools, this is a wake-up call: your customers may already know.

If you run a small business and have leaned on AI to generate website copy, email newsletters, or social posts, this matters more than you might think. The word 'dependable' is not the only tell. AI models gravitate toward certain safe, middle-of-the-road adjectives and sentence structures because they are trained to sound confident and inoffensive. The result is copy that feels strangely generic, even when it is grammatically flawless. Customers pick up on this, even if they cannot articulate why. If your brand voice suddenly shifts from conversational to oddly formal and stuffed with words like 'dependable,' 'seamless,' or 'robust,' you risk eroding the trust you spent years building. Authenticity is one of the few advantages small businesses have over larger competitors, and AI copy can quietly undermine it.

What is genuinely new here is the specificity. People have speculated about AI writing tells for years, but pinning a 23x frequency spike on a single word is the kind of concrete, testable claim that moves the conversation from vibes to evidence. We agree with Brandom's implicit point: these tells are not going away. As models improve, the obvious markers will fade, but new ones will emerge. What we are skeptical of is the cat-and-mouse framing. Some businesses will treat this as a challenge to edit AI copy just enough to dodge detection, rather than asking the harder question of whether the copy actually serves their customers. The real risk is not getting caught. It is publishing content that sounds like everyone else's.

The second-order effects cut in both directions. On one hand, human writers may see a resurgence in demand as businesses realize that generic AI copy blends into the noise. On the other hand, the proliferation of AI-generated content could make readers more skeptical of everything they read online, which hurts honest businesses that genuinely write their own material. There is also a cost to consider: if you have already invested in AI tools and workflows, retrofitting your content strategy to sound more human is not free. It takes time, editorial judgment, and often a return to paying skilled writers. Businesses that ignore this shift may find that their 'dependable' copy is quietly costing them conversions.

What should you do next? Audit your recent content for AI tells, starting with the word 'dependable' and branching out to other overused adjectives and formulaic sentence structures. Read your copy aloud. If it sounds like it could belong to any company in any industry, rewrite it. Consider adding a disclosure if you use AI assistance, not because regulations require it everywhere, but because transparency builds trust. And watch for follow-up research on AI writing tells. As detection methods improve, the businesses that thrive will be the ones that use AI as a tool for ideation and drafting, not as a replacement for a genuine voice.

“Opus 5.5’s biggest tell is the word “dependable,” which pops up 23 times more often than in human samples.” — TechCrunch

Takeaway: Audit your AI-assisted copy for overused words like 'dependable' and invest in a genuine brand voice before customers tune you out.