Context Window 81
This week there is a tight cluster of stories around how AI companies access content, what that is doing to publishers, and the economic model for compensating creators. But it’s not just a theoretical set of discussions, with at least one practical step publishers need to take in the next couple of weeks.
As a great philosopher once said, life moves pretty fast. Last week I wrote about the narrow window to test Anthropic’s Fable model before pricing kicked in, but the day I sent the newsletter, Anthropic pulled access altogether in response to a US government directive.
On the heels of the Kindle AI product releases last week, Google wrote to Google Play Books publishers to let them know that its AI Book Insights feature will be rolled out to most English-language ebooks starting 6 July. Titles will be opted in to the feature automatically, unless publishers have opted out before then. Check your email and act accordingly.
Audio platform Storytel is launching an AI assistant, Genie, which provides recaps and recommendations leveraging LLMs from OpenAI, Google and Anthropic. The press release mentions a proprietary metadata layer to ensure results are grounded in source material, which I’d be asking questions about if I worked with them. There’s also no mention of publisher opt-ins or controls.
There’s been lots of coverage of John Wiley’s latest results, which included nearly $50 million of AI revenue for FY 2026. But the bit that interested me most was the below-the-line number: $23 million of cost savings driven by AI, $9 million of which was in the last quarter, suggesting a quarter-on-quarter acceleration over the fiscal year. As far as I can recall, or my agents can verify, this is the first time I’ve seen a publishing company put a dollar value to AI benefits.
In an example of media companies taking the initiative against AI companies, a group of online publishers have added contractual language to their website terms and conditions stating that any use of the site by LLMs will result in an invoice being sent—with the option for simple legal enforcement if it is not paid.
I haven’t seen legal advice on the enforceability of this, particularly in other jurisdictions, and I’d encourage readers to do their own diligence, but there is a how-to document here if you’re interested.
What sorts of things might AI be looking for on the web? Muck Rack published an analysis of over a million AI citations: key insights included journalism driving 25% of citations, press releases accounting for 1% (though with a 5x growth in the last year) and half of all citations pointing to material published in the last year. It’s a reminder that your publicity department is helping you talk to LLMs as well as journalists and readers.
On the subject of authors and publishers getting paid, there’s a new opinion piece in HBR suggesting a model for this through collective management organisations. I don’t disagree with the basic premise of this, but the piece goes on to argue that Hollywood-style profit shares are the right way to compensate creators.
The viral post of the week, at least on my channels, was bestselling author Tim Ferriss talking about the collapse of some non-fiction categories with reference to his own sales figures—he’s currently nearly 60% down year-on-year. Beyond those headlines, it’s a really interesting thought-starter on what AI answers can do better than books or videos, and where human stories continue to have value. I recommend it.