AI chatbots keep failing every accuracy test thrown at them
PLUS: Why Substack’s new subscriber milestone is so significant
Welcome! I'm Simon Owens and this is my media industry newsletter. If you've received it, then you either subscribed or someone forwarded it to you.
If you fit into the latter camp and want to subscribe, then you can click on this handy little button:
Let’s jump into it…
AI chatbots keep failing every accuracy test thrown at them
The Columbia Journalism Review performed a pretty rigorous test assessing the accuracy of all the major AI chatbots, and I don’t think it’s an exaggeration to say that they failed miserably:
Most of the tools we tested presented inaccurate answers with alarming confidence, rarely using qualifying phrases such as “it appears,” “it’s possible,” “might,” etc., or acknowledging knowledge gaps with statements like “I couldn’t locate the exact article.” ChatGPT, for instance, incorrectly identified 134 articles, but signaled a lack of confidence just fifteen times out of its two hundred responses, and never declined to provide an answer. With the exception of Copilot—which declined more questions than it answered—all of the tools were consistently more likely to provide an incorrect answer than to acknowledge limitations.
Nearly every major tech platform is pushing AI chatbots onto their customers and claiming these tools will vastly improve their work product while also saving them time, but every time these chatbots are subjected to any rigorous testing they're shown to be wildly inaccurate a sizable percentage of the time.
This is why every single publisher that leaned into AI-generated content has ended up with egg on its face. Without assiduous fact checking, these bots are going to insert errors into copy. And if you have to hire someone to fact check every statement, then what efficiency is the chatbot driving? You could just as easily pay that person to write the story.
Why small-donor members are an important component to any nonprofit newsroom
The Houston Landing is one of the most ambitious nonprofit news outlets to debut in recent years, but its continued operation is already under threat due to a major donor pulling out:
In January 2022, Arnold Ventures committed $4 million in initial funding, to be paid out over three years. From the beginning, the organization referred to the money as seed funding, a catalyst to get the Landing off the ground. But staff at the foundation were enthusiastic enough during the first year that at least some of the Landing’s staff felt optimistic that the Arnolds would eventually re-up their commitment. Indeed, in December 2024, as the first three years wrapped up, leaders at the Landing believed they had secured $850,000 in additional funding from Arnold Ventures. But no contract was ever finalized and, when the Landing came asking about the money in December, Arnold Ventures announced it would not provide any more funding.
A lot of local news startups over the past few years have been buoyed by large donations from wealthy individuals who are concerned with the decline of local journalism, but an obvious flaw in this strategy is that it only takes one or two pullouts to seismically impact the business.
This is why these startups need to focus on building their small-donor membership programs — not necessarily to replace the large donors, but to protect the outlet from large swings in revenue that can trigger layoffs. Many of the media leaders I've spoken to aim to generate at least half of their revenue from low-priced memberships.
The challenges of expanding into new media niches
Jessica Lessin, the founder of The Information, recently talked with Dealbook about her decision to make seed investments in multiple niche media companies:
I’m absolutely all about revenue and controlling your destiny and direct subscription revenue, and that being the true north.
I’ve also always been about that founder that has the real expertise. And I think big media companies dismiss the niches. They think they’re too small. Across all of these investments, the criteria I’m looking for is there’s got to be real revenue and a revenue model that is direct and user-driven where the brands can control their own destiny. But also a very passionate founder.
I really like Lessin's approach to media expansion; rather than stretching herself thin and launching new verticals, she simply makes seed investments in companies founded by people who are extremely passionate about those verticals. That way, she can stay focused on her core competency — tech reporting — while leveraging her contacts and media expertise to identify promising niche players.
A lot of media entrepreneurs take the opposite approach by trying to copy-and-paste their media format across more and more niches. And sometimes that works! It certainly did for companies like Industry Dive and Aging Media. But that requires a special kind of media operator who isn't too invested in any one niche, whereas the success of The Information is rooted in Lessin's deep love and obsession with tech reporting.
Please don’t take my newsletter for granted
I rely on paid subscriptions for the vast majority of my revenue. Without enough paid subscribers, I can’t continue justifying spending 40+ hours a week on my newsletter and podcast, and I’ll need to shut them down so I can seek out other work.
Let me put this another way: if you’d be disappointed that I suddenly announced I’m shutting down my newsletter — a very real possibility — then you should probably subscribe.
Seriously, it’s only $50 for a full year, and if you’re using insights from my content to improve your own business, then that $50 pays for itself. And if you use the link below, you get 20% off for the first year:
Disney adopts the freemium model
Last month, Apple TV + released the first full season of Severance on free streaming platforms to promote its second season, and now Disney is doing something similar with the new season of Andor:
Disney+ and Lucasfilm are widely making the debut season of its acclaimed Star Wars drama more widely — and freely — available ahead of the highly anticipated release of season two. Starting today, the entire first season of the series — which consists of 12 episodes — is available on Hulu. In addition, the first three episodes are now available for free on the Disney+ YouTube channel.
I've been advocating for this type of strategy for years. Basically, the idea is that you can promote an upcoming season of a show by releasing the previous season on free, ad-supported platforms. By that point you've wrung out all the subscription revenue you could from the previous season, and by releasing a free version you're not only driving additional revenue through ads, but also generating more demand for the new season, which sits behind the paywall. There's really no downside.
In fact, I'd love to see an even more radical version of this strategy: Imagine Netflix launching a free tier of its service that just offers the first season of all its TV shows. It would vastly expand its ad inventory while creating an even better subscription funnel for its paid tiers.
Some good longform journalism
What do you do when you become so famous that your mere presence creates a dangerous situation? The streamer iShowSpeed realized he needed to radically change his approach to public appearances when he found himself trapped in a store, unable to escape the throngs of fans that had clogged the streets outside. [golfcart]
Why did several left-leaning journalists — people like Matt Taibbi and Glenn Greenwald — suddenly become loud Trump apologists? Most likely because the rightwing media ecosystem recognized these journalists' potential to pull wavering liberals over to the other side of the aisle and so showered them with attention and money. [The Nation]
ICYMI: How an engineering student accidentally started a thriving science news site
My other newsletter: The best longform journalism we consumed this week
Are you following me on social?
You can follow me on Substack Notes, Threads, my private Facebook group, LinkedIn, Bluesky, and Twitter.
Behind the paywall
Here’s what I have on deck for paid subscribers:
Why Substack’s new subscriber milestone is so significant
The rise of YouTube apprenticeships
Why the Washington Post is reorganizing its newsroom
Let’s jump into it…