This past Saturday morning, cup of coffee in hand, I did something I almost never do these days: I opened my LinkedIn account and decided to scroll my feed for a few minutes.
At the start of 2026 I made a decision to leave almost all social media. I keep LinkedIn because of the work I do. I thought I would go into some kind of withdrawal; I was surprised to find I felt a huge relief. I find the relentless feed-scroll dizzying. I come away feeling overstimulated, distracted, and disappointed at the lost time.
This Saturday, I intentionally went into my LinkedIn feed with a different purpose. It was a casual social experiment: observing the themes and topics that are getting the most visibility and traction as we continue to navigate the AI revolution.
Between two AI-driven product launch posts was a post offering reassurance and encouragement. “You are not behind,” it read. Clearly well-intentioned, the post was also revelatory: universal messages of comfort resonate broadly when they feel necessary; when the zeitgeist suggests the opposite is true.
My brief scrolling experiment produced the following: outside of advertisements, 38 posts appeared from people and organizations in my own network. Of those, 26 were about artificial intelligence - a little more than two thirds. Of the 13 ads I saw, 5 were selling AI.
The largest group was announcements: a new model from one of the frontier AI labs appeared twice. A new AI device also appeared twice, once in an interview with Sam Altman, and once in a user’s post, reporting how the device had solved one of his most challenging problems that morning. Marissa Mayer announced her new AI startup. Another person declared that ChatGPT and Claude had just replaced the note-taking app he had used for years.
The second largest group of posts was about AI at work. There was an event on AI and creativity, an essay on how a major agency was adopting and integrating AI into their workflow. There was a job opening post, looking to hire a director of “agentic” products. There was an announcement of a new award in the “digital health & AI” field. Then there were a handful of posts discussing AI public policy and risk, including the recent White House accord on “super intelligence.”
By far the smallest group of posts focused on the issue that matters most to me. Out of 38 posts, just 5 (13%) were about trust: can we trust the information these systems give us? Who owns the data they were trained on? What exactly is the data they were trained on? How do we verify where a piece of content came from? What happens to the conversations we are having with these machines? One of the 5 posts announced an upcoming talk entitled “Whose Intelligence Is It?”
Taken as a whole, my Saturday morning scroll through LinkedIn read more like a release calendar. Progress and success measured in launches, with launches arriving daily. It’s no wonder someone felt compelled to tell us all that we’re not behind.
But there was one other phenomenon that struck me: not only is the “feed” talking about AI. Increasingly, AI is also doing the talking.
This summer, LinkedIn added a new option to every post: members can now flag content that “seems like AI slop.” As reported by Fortune, more than a million members used it in the first two weeks. LinkedIn’s chief product officer, Hari Srinivasan, laid out the challenge: “we know we have more to do to ensure LinkedIn remains a place where you can find real people & real perspectives.”
So far, readers seem to know the difference. LigoSocial, which sells an AI tool for writing LinkedIn posts, compared more than 100,000 posts against each author’s other work and found that the ones that read as AI-generated drew 21% less engagement. Humans are still looking for other humans, for authenticity and connection. Will we continue to find those in social media forums like LinkedIn, or others?
So how much of LinkedIn’s current content is written by machines? I searched for an answer and came up with some interesting datapoints. The numbers range, depending on samples and methods. In July 2026, Originality.ai, a company selling AI detection, sampled 5,000 public posts on LinkedIn of 100 words or more and classified 81% as likely AI-generated. Pangram, another AI detection company, analyzed over a million posts that appeared in the feeds of people who use its browser extension between late April and early July of this year. Over 40% of long-form posts on LinkedIn - posts of more than 250 words - were flagged as fully AI-generated; a count that leaves out posts that were only AI-assisted.
When a professional “commons” departs from a mutually agreed-upon reality, the risk is that credibility will be increasingly influenced by the quantity rather than the quality of information. The rule seems to be that those who post most often, most confidently, will dominate the airwaves as well as the tenor, direction, and substance of public discourse. Perhaps this has always been true: the public square was historically a cacophony of conflicting views, and the loudest prevailed. What is new is the scale of the departure from a shared understanding of what is real, aided by the fact that a growing number of the voices aren’t people at all.
I don’t think the answer is less AI. This seems unrealistic, if not entirely implausible at this point. The tools are remarkable, and many of the people and companies who are posting about them are doing meaningful, serious work at a level and velocity unimaginable before now. What’s still missing, at the root, is the validation layer that - like mycelium connecting all plants and living things - keeps us connected and grounded in what’s real.
That question has been the center of my working life for some time now. In May 2025, our company filed the first patent application for our data taxonomy technology, ATLAS™. It grew out of our 50-year legacy of publishing field guides and other reference guides - books whose entire value rests on being accurate: the bird really is that color and that size; that mushroom really is that dangerous. ATLAS™ exists so that AI can be grounded in expert-validated, rights-cleared reference material of any domain, and so that an answer can be traced back to its source.
The five posts in my Saturday morning feed that asked about provenance and ownership were, to me, the most important ones on the page. Our company’s tagline is three words long, and after my weekend scroll-stroll it feels less like a slogan and more like table stakes for an authentic professional life:
Know the source.
A note on this newsletter. When I started writing here, I had promised an essay every Tuesday. While I continue to write every day, the voice that is emerging through this process is better suited for longer form thinking and exploration, rather than weekly push-outs. My self-imposed weekly deadline felt like it was turning my essays into what this one describes: a work product on a tight schedule, at volume. The issues I care about most need time to research and room to think. So, starting today, I’ll publish one essay on the first Tuesday of each month, built around a single theme. Fewer posts, each one considered: my small answer and antidote to the tyranny of the feed.
I am also working on two new manuscripts: two books in gestation. As the ideas take form and direction, I’ll offer a few tidbits here.
Thank you for reading, and for being here. I’ll see you on the first Tuesday of November. Should be an interesting day.
- S
Sources
Fortune (1M flags)
Entrepreneur (Srinivasan quote)
LigoSocial, State of LinkedIn 2026
Originality.ai
Pangram
Al Jazeera (White House Accord on Super Intelligence)