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
Why Thoughtful AI Use Is More Important Than Ever
Multiple regional winners of this year’s Commonwealth Short Story Prize face allegations that their work may have been generated with AI. Readers pointed to the increasingly common “not X, not Y, but Z” structure, nonsensical metaphors, and AI detector results. The Commonwealth Foundation said it does not use AI checkers, in part because the tools are not reliable enough to settle the issue.
Most text-based AI detectors produce far too many false positives, but AI detection is improving for some types of content. OpenAI announced this week that it is adding Google’s SynthID watermarking to images generated by ChatGPT, Codex, and the OpenAI API, while also expanding its use of C2PA Content Credentials and previewing a public verification tool. Those kinds of provenance systems could help when someone needs to know whether an image, video, or audio clip was generated or altered.
For text, it’s not that easy because copy can more easily be edited, altered, and combined. There’s no clear forensic trail to follow, just a judgment about how the writing sounds compared to typical AI patterns. But in communications, perception is everything.
Reporters are already sensitive to vague pitches and hollow commentary. If a quote sounds like it was generated from a prompt, it may not matter whether a person technically wrote it. If an apology feels generic, the audience may read it as evasive, even without knowing what role AI played.
AI is a great tool to help communicators produce a draft, whether that’s through brainstorming, editing, or even the creation of the content. But the content still needs to be good. A unique point of view and a specific story are what make it stand out and not feel like generic AI slop.
But even then, if writing contains AI tells, readers will start making judgments. We created a free AI Tell Detector to flag the phrases, structures, punctuation habits, and vague language that raise suspicions of AI generation. It is not meant to prove whether something was written by AI, but to help writers catch the patterns that make otherwise solid work feel automated or unfinished.
The Commonwealth Short Story Prize controversy shows where this is heading. You may not be able to prove whether a piece of writing involved AI, but you can make sure it does not read like someone handed over the job to a prompt and called it done.
Elsewhere …
- LISTEN: What 5 HumanX Builders Know About AI that Most Companies Don’t
- Google Announces Slew of AI Advances, Including a Personal AI Assistant
- Pope Creates AI Study Group as Vatican Prepares to Release His First Encyclical
- Google Reinvents Search Before AI Rivals Replace It
Tips and Tricks
💭 How agentic memory works
What’s happening: ChatGPT and Claude retain memory passively — the platforms handle it mostly in the background. Agents are built differently and let you be a bit more explicit about how you want it to remember and work with that memory.
Long-term memory: This is where you want day-to-day context to complete your work — information about you, your company, your clients, your work preferences, and other information that informs its outputs. Skills can also be a form of long-term memory, giving the agent set instructions to act on for a task or process.
Short-term memory: This is what the agent has available while it’s working — context you’ve given it, documents in a shared folder, internet searches, and other information specific to the task at hand.
Scheduling: The big advantage of agents is that they can act on your behalf without prompting. Recurring work tasks might draw on long- and/or short-term memory. Some will be on a slower cadence, like daily summaries or weekly reports, while others might require more frequent checks.
Quote of the Week
“Agents in search is the next step. One of the cool things we get to do here at Google is build technologies that get immediately deployed into multibillion-dollar products.”
— Demis Hassabis, CEO and Co-Founder of Google DeepMind, as quoted by Axios about how Google is changing search
How Was This Newsletter?
