Hi Wayne.

Very interesting that the 'Most-Streamed' broadcast was 13.38 minutes in a 60 minute broadcast.   13.8 minutes in an hour is about 22% of the duration.

It's a valid mathemetical calculation but many people take that as a rating, assuming it is large and worthwhile rating of advertising.   A genuine rating reports the average audience over the broadcsat time, which is a better estimated audience for any minute in the broadcast, rather than cumualting minute-by-minute.

This has been going on for years and beneficial to all- additional revenue for the sites and additional easy spend by the agencies who can tell their clients they bought x dollars on Hulu or whoever without mentioning 20% of spend were spent on the same content or in the same pod.
if not then this would have been solved years ago 

Maarten if a time buyer gets twice as many impressions at half the price at one  seller this doesn't necessarily mean that many or all of those horrid, repetitive, exposures weren't effective--providing they were reasonably well dispersed. More likely, the problem is caused by relying too much on computers to do work that humans should play a role in--re how the commercials are scheduled.  It's a simple problem to deal with if handled correctly.

The walled gardens make it hard across channels/olatforms. But I saw the same ad, usually an insurance with a bird in its commercial (not a duck) or a weight loss drug, multiple times in one show just like you. That's not a walled garden issue. That is either an advertiser issue who is only interested in cheap CPM's and not the consumer experience. Or an disingenuous agency that thinks it's ok to schedule for profit, not effectiveness. 

Spot on Cory.

In AU it is happening as well ... three blokes on the sand is one and it is nauseating.

We also have an ad of a garden hose that 'extends further than the standard hose.'  The original ad was about 4 minutes .... it seems they have bought the entire 30 minutes and repeated showing the ad.    Vomating.

Wayne, the kids cable channels are almost  surely doomed so why not try to sell them--if a sucker buyer can be found? Same goes for most of the other CBS cable channels--but not so for WBD's CNN. So one has to be selective.

Not all cable channels are doomed to speedy extibction---the three news channels, for example, ESPN, Bravo and others may well survive and be profitable for some time to come.

What's happened  is that cord cutting has diminished in intensity which means that eventually cable may wind up with a hold on about 15-20% of all viewing. That will be enough to sustain anywhere from 15-30 chanels once the rest of the herd is culled. 

Gee, you combine moronic level intelligence with the impluse control of a 2 year old--what could go wrong? 

 Quite right, Scott. Not only is the amount of streaming GRPs barely 25% of the total for both platforms--I'd peg it a tad lower, by the way--- the upfront is not only about prime time. Add in all of the non-prime national time buys on the broadcast TV networks, cable and national syndication and the linear take swells to around $30 billion. Streaming will do well to garner half as much. 

I still see alot of the same commercials on streaming. Frequency remains an issue. Steamers have not solved that equation yet. Ad Supported Streaming of any "decent" audience is only about 25% of total TV Viewing time. Why do you think 55% of the Upfront commitment will go there? Clients feel they are more forward thinking if they are all over Streaming and Agencies make more money off of it. Nobody wants to deeply question the trend.

Steve, I want to give you real credit for something.
Your decision to explicitly label AI-generated passages is a pioneering step toward Truth in the AI age. Many writers use AI quietly. Very few make the synthetic layer visible to the reader. In that sense, you are not only writing about the problem — you are beginning to model a public standard for authorship, disclosure, and trust.
But this important step also reveals the next problem.
Disclosure is not verification.
“Written by ChatGPT” tells us where a passage came from. It does not tell us whether the claims are true, whether the citations exist, whether the sources say what they are claimed to say, or whether synthetic material has entered the scientific, medical, or journalistic record.
Your publication about AI hallucinations in medical research led me, as a reader, to look more closely at the work being done in this area. I should emphasize that I am not writing from the business side of this issue. I am simply an AI enthusiast and a reader whose thoughts were shaped by your publications.
That is how I came across the work of Max Topaz and Citadel. From the outside, Citadel seems to be not an anti-AI project, but an attempt to build a verification layer before the accident happens: checking citations, sources, claims, provenance, and synthetic insertions before they become part of the public knowledge stream.
With all due respect, it seems to me that it could be valuable for you to connect with Max Topaz directly, because your public concern with Truth and his practical work on verification appear to meet at precisely the same point.
So perhaps the next frontier is this: not only to disclose AI, but to verify AI.
Use AI. Label AI. But above all, verify AI.

The question is: why do the platforms mix synthetic (fake) answers with citable fact-based sources? Is it a bug or a design choice. I asked ChatGPT. It's answer: 

Humans are notoriously bad at distinguishing confidence from accuracy. If a statement is fluent, specific, and presented in a familiar format, we tend to treat it as authoritative. Plausibility is the superpower of modern AI. (Written by ChatGPT).

We're building systems that make fiction indistinguishable from fact, then deploying them at planetary scale. That's the danger (Written by ChatGPT).

To use, or not to use — that is no longer the question. Use AI. Trust, but verify. Steve, judging by your previous publication, this problem is already familiar to you. So I was hoping this new piece would go one step further: not only Wikipedia vs. Google, but verification. Have you already found practical ways to verify what AI inserts into the knowledge stream? Who should do it, how should it be done, and with what tools? I mean tools that can check claims, citations, sources, provenance, and whether a reference actually exists and says what it is claimed to say. Because if the answer is simply “trust Wikipedia and distrust Google,” that is not a strategy. It is a retreat. AI is already here. The question is no longer whether to use it. The question is: who verifies?

Fantastic way to give Ambient some more attention.
A highly underrated genre.

Excellent and pragmatic recommendations - thanks for sharing.

The practical follow-up to the Max Topaz story is that this is no longer only a warning about AI hallucinations.
Professor Topaz and his team have turned the problem into a working editorial verification tool: a system that can scan manuscripts or published papers, detect non-existent references, separate likely ordinary author errors from more serious patterns, and help editors decide what should be corrected, investigated, or prevented before publication.
AI is already inside publishing workflows. The answer is not denial or panic, but an immune system for the scientific record. Tools emerging from Topaz’s work, including Citadel, point exactly in that direction.

Nice to see two of my all favorite media people having an interesting discussion.
Good luck Donnie, and hi Joe1

Without the NFL lead-in in the winter, and the NCAA in the spring, 60 Minutes would be just be another news program like 20/20 and Dateline and 48 Hours in terms of ratings.

Remember 60 Minutes II or 60 Minutes Wednesday?

Neither do I

Impossible.

Nielsen Ad Intel says Colbert national TV revenue from June 1, 2024 to May 31, 2025 was $83.2 million US Dollars.

That fell to $63.5 million from June 1, 2025 to May 21, 2026

When the majority of your on air talent is between 60 and 84 years of age, it's hard to argue that changes aren't needed moving forward. 

The median age of the 60 Minutes of viewer is 65.  

Hardly the age group you want to build a future franchise on. A bit too early to see how the long game works out and the people who've been sitting in their anchor seats for 40+ years really shouldn't be complaining that it's tile
to start moving in a new direction. 

I watch Squatters on A&E pretty good show and yes, I hope laws do change because of these shows sad that we need these types of shows. The host of the show had a squatter living in his dead mothers house opening of the show in each EP they always have warning before the show starts.

Not sure which it could be ...

* “Squatters: Get the Fish Out of My House!” ?
* “Squatters: Get the Flea Out of My House!” ?
* “Squatters: Get the Fuzz Out of My House!” ?
* “Squatters: Get the Foil Out of My House!” ?
* “Squatters: Get the Flux Out of My House!” ?
* “Squatters: Get the Frog Out of My House!” ?
* “Squatters: Get the Fido Out of My House!” ?

Could be a series ...

Wayne, what's the source for that huge increase in the TV networks' late night ad revenues over the previous year that you referred to?

Mr. Rosenbaum has raised an extremely important problem: AI hallucinations are not merely technical errors, but breaches in the trust systems on which media, science, business, and public life depend.

The first emotional reaction to AI hallucinations is often rejection: perhaps the only safe response is not to use AI at all. That reaction is understandable. When AI fabricates a quote, a citation, a source, or a chain of evidence, it does not merely make a mistake. It damages trust.

But refusal cannot be the long-term answer. AI is already entering journalism, publishing, marketing, science, medicine, education, law, and corporate communication. The practical question is no longer whether AI will be used. It is how to prevent AI-generated errors from contaminating the systems that depend on credibility.

The better metaphor may be immunity.

A healthy body does not survive by denying the existence of viruses. It survives through detection, response, memory, and defense. Information systems now need something similar: an immune layer that can identify fabricated references, unverifiable claims, false quotations, and weak evidence before they are published, cited, or amplified.

That is why the next stage should not be anti-AI. It should be pro-verification.

One important example is the work of Professor Maxim Topaz of Columbia University, whose recent research helped expose the scale of fabricated citations in medical literature. His team has developed Citadel, an AI-focused verification tool that can be understood as a kind of immune filter for texts, claims, citations, and evidence.

This kind of work matters far beyond medicine. Media, publishing, marketing, public communication, education, and business will all need trustworthy verification layers as AI becomes part of everyday knowledge production.

The opportunity here is not only academic or technical. It is also institutional and commercial. People with experience in media, communication, reputation, markets, and scale — including voices like Mr. Rosenbaum, who are already bringing public attention to the issue — could play a decisive role in promoting such verification systems and turning them into practical infrastructure.

If this conversation reaches people who are interested in building that kind of AI-era immune system — publishers, media leaders, investors, platforms, agencies, or technology partners — this may be the moment to engage.

The future is probably not “AI or no AI.” It is AI plus responsibility, AI plus verification, AI plus trust architecture.