What You Should Know
Why Identifying AI Content Will Always Be Important
For communicators, like so many in other industries, AI quickly became part of the workflow. It provides an assist in brainstorming, pitching, research, and, of course, content generation. AI’s speed has contributed to a surge in new material. Research from Graphite suggests AI-generated articles now outnumber human-written ones online.
The flood of unedited, AI-generated writing is prompting pushback from audiences and reporters, who are already rejecting AI text that looks polished but lacks substance and humanity. To be clear: AI can absolutely be a helpful tool when used well, but it’s up to communicators to prompt the tools and edit the outputs well enough to not leave a sour taste in a reader’s mouth.
Publications like Entrepreneur, InformationWeek, and VentureBeat say they won’t accept contributed content created with AI at all. The AI detectors used to determine what is or isn’t created by humans are often faulty, but the perception of AI within the media is growing more consequential.
Some reporters have pushed back on quotes, commentary, and background materials they believe were generated by AI. One reporter took to LinkedIn to share an anecdote about “unquotable” responses that she believed were AI in part because they “don’t really say anything.”
Communications teams have always served as a final filter, but that role is more important than ever, as passing along unedited AI output from an executive or client — especially if it lacks substance and specificity — is a new way to burn your reputation with reporters.
Sometimes the AI element can be hard to spot. Platformer reporter Casey Newton just wrote about how he thought he had uncovered a major whistleblower story, only to learn the “source” was feeding him cleanly written but meaningless AI-generated responses. The reporting unraveled once the substance didn’t hold up.
Recognizing AI-generated content goes beyond spotting certain words or structures. There are certainly tells (feel free to build off our list), but making the writing feel natural and human requires a deeper level of analysis and discernment. AI outputs often repeat ideas, add filler that doesn’t move the story forward, and avoid taking a clear stance. They can sound polished without being specific.
Human writing reflects intent and judgment. It understands who it’s for and why it exists. Pure AI output rarely gets there without detailed guidance on voice, audience, and purpose, and even then, it probably needs heavy editing.
Identifying AI content matters because every piece of communication carries a name and a reputation with it. If what you share feels automated or soulless, it can undermine trust with editors, reporters, and audiences who expect a real person or brand behind the words.
Elsewhere …
- AI Begins Prescribing Medications in Utah
- At CES 2026, Everything Is AI. What Matters Is How You Use It
- Why AI Predictions Are so Hard
- Motorola Is Entering the Wearable AI Game
- 40 Million People Turn to ChatGPT for Health Care
- How AI Turned a Utah Police Officer into a Frog
Tips and Tricks
Tweak your personalization settings
What’s happening: Late last month, OpenAI added a new layer to personalization settings with four characteristics: warmth, enthusiasm, formatting, and emojis. It gives users more levers to pull than just the base style and tone (which has eight options) and custom instructions.
So what?: More options mean more ways to tinker with the style of responses to get what you want. For example, the “cynical” style, described as critical and sarcastic, defaults to forcing (often nonsensical) punchlines in nearly every response. There’s not so much substance, but almost always something that borders on insulting the prompt. Tweaking the new options with more warmth and less enthusiasm seems to remove the insults but maintain the independent evaluation and criticality of the information provided.
Try this: It will take some trial and error to get this where you want it to be. Start by removing personality-related directives from your custom instructions, keeping it to just style preferences like banning words or encouraging it to write in active voice. Next, experiment with the styles and characteristics. For instance, I find ChatGPT often drones on without saying much, so “efficient” style with more warmth, less enthusiasm, and no emojis (does anyone really want them in their responses?) provides better results..
Quote of the Week
“The ChatGPT moment for robotics is here. Breakthroughs in physical AI — models that understand the real world, reason and plan actions — are unlocking entirely new applications.”
— Jensen Huang, Founder and CEO of NVIDIA, in a press release about a new set of models unveiled before CES
How Was This Newsletter?
