Each week we'll gather headlines and tips to keep you current with how generative AI affects PR and the world at large. If you have ideas on how to improve the newsletter, let us know!
What You Should Know
The New Signs Your Writing Feels AI-Generated
A recent report found developers are leaning on AI to write code faster and stepping in to shape and refine what it produces. The output gets them started, but their judgment is still a huge part of the equation.
The same shift is happening in writing.
AI can generate clean, structured content in seconds. That speed is a driving force behind the adoption boom. Roughly 87% of content marketers say they’re using it, according to Ahrefs, and more than half of what’s published online is now machine-generated. The draft is no longer the challenge. But is that draft good enough? Does it sound like you (or whoever you’re writing for)? Does it even sound human?
Most people have learned to avoid the obvious tells. The phrases that gave AI away a year ago, like “ever-evolving landscape,” don’t show up as often. What’s replaced it is harder to spot.
Most patterns now are structural. Sentences that are technically fine but follow the same rhythm and length. Transitions that sound conversational but don’t carry meaning. Punctuation used incorrectly. Language meant to stand out without actually saying anything specific. None of it is necessarily wrong on its own, but over the course of a piece, it starts to stack up and look like all the other AI content on the internet.
When so much content is produced the same way, uniformity is a risk. Readers may not be able to call out a specific phrase or sentence, but they can tell when something feels generic. You want writing to evoke emotion, but disenchantment isn’t the one you’re going for.
We’ve developed a couple tools to help. There are plenty of AI tells lists out there, and many of them are quite long. Ours is kept to one page to distill only the most pressing phrases and structures that make writing look robotic. We’ve also built an AI tell detector that flags those items. Unlike most AI detectors, the point here is to identify what specifically looks like AI, not to have a witch hunt that (often incorrectly) accuses someone of using AI.
You want writing to be accurate, structured, authentic, and unique. AI can get you most of the way there, but making the story sound like it came from you is still on you.
Elsewhere …
- LISTEN: Deepfakes Used To Be Funny. Now They Threaten Every Business
- Anthropic Holds Mythos Model Due to Hacking Risks
- Google CEO Sundar Pichai Says ‘AI shift’ Opens Opportunities to Invest in Startups
- OpenAI, Anthropic, Google Unite to Combat Model Copying in China
Tips and Tricks
🗣️ Talk it out
What’s happening: Before you start creating content, you need to think through your concept. It’s an important step between two common AI uses, brainstorming and content creation, and is yet another area that AI can help with. Instead of prompting it for answers, prompt it to ask you questions that make you consider new approaches and angles.
How: If you’re a more conversational person, use voice mode on your AI tool of choice and literally talk it out. You can also perform the same exercise in writing, which might be easier to keep track of all the details you want to iron out, like defining the story for the right audience, making sure you’re using and appropriately citing data to support your point of view, and making sure the story flows. It’s a way to keep you honest and make sure you’re offering a well-thought-out version of an idea.
Quote of the Week
“Over 99% of the vulnerabilities we’ve found have not yet been patched, so it would be irresponsible for us to disclose details about them (per our coordinated vulnerability disclosure process). Yet even the 1% of bugs we are able to discuss give a clear picture of a substantial leap in what we believe to be the next generation of models’ cybersecurity capabilities—one that warrants substantial coordinated defensive action across the industry. We conclude our post with advice for cyber defenders today, and a call for the industry to begin taking urgent action in response.”
— Anthropic’s red team report on the launch of a new model, Claude Mythos Preview, and its cybersecurity capabilities
How Was This Newsletter?
