Placeholder

(Originally posted on LinkedIn): Every December since 2020, I’ve posted about my books of the year: to mark the passing of the year, as suggestions for anyone looking for holiday gifts, and to start a conversation with my network about books at a time when many of us would like to read more, but struggle to (I wouldn’t have got through many of them without the option to listen as audiobooks on xigxag).

Publishing seems to have been slowing down for the holidays for the last couple of weeks, and while AI isn’t on quite the same trajectory, there are fewer notable developments this week. I hope that you have a good festive break if you’re taking one—I will be, so there will be no newsletter next week. Normal service resumes in January.

A slightly shorter newsletter than normal this week, as I have been balancing four days on the road for a conference and board meetings, and trying to close off various projects before the holidays. I’m planning to do one final newsletter for the year next Friday, 19 December, then take a week and resume publication in the New Year.

This was a week in which AI refused to behave according to anyone’s preferred narrative: strong environmental claims unravelled, legal cases tightened, policies were challenged, and AI models themselves showed new points of fragility. I was particularly interested in new research that challenges a common false dichotomy between AI use and creativity. ​ Talking to publishers, the environmental impact of AI is regularly cited as one of the major reasons not to use it—and it is one of the central criticisms in Karen Hao’s bestselling book Empire of AI. But researcher Andy Masley recently highlighted significant methodological errors in Hao’s calculations, which overstate water usage by several orders of magnitude. In response to Masley, Hao has acknowledged the error and will correct the book. The fine detail of this is quite abstruse, but for me the takeaways are simpler. The response to the controversy is a Rorschach blot for how people feel about AI. Many of the people commenting in support of Masley condemned Hao for being ideologically anti-AI. In turn, some of Hao’s supporters dismissed Masley out of hand for his links to the Effective Altruism movement. Reality is more nuanced than either characterisation: Hao may be directionally right even if some claims are discredited, while Masley’s analysis highlights an important epistemic risk, namely, how easy it is to take weak analysis at face value when it aligns with a preconceived worldview. ​ Second, the uncertainty over environmental impact exists in large part because the standard of sustainability reporting from AI companies is so poor—it was interesting to consider this in the context of Satya Nadella’s comments this week that tech companies need to earn social permission for their resource use, a standard the AI sector is currently failing to meet. ​ A fundraising document for the AI music platform Suno claims it is generating the equivalent of Spotify’s entire catalogue every two weeks. I’d love to know the equivalent number for generation of self-published books; Amazon/KDP is probably the only organisation with data robust enough to attempt that calculation, and they don’t publish it. It’s also worth noting that music’s subscription economics make this sort of volume easier to monetise than in books, which remain largely sold a la carte. So if this isn’t happening at this scale in publishing yet, it’s not because it would be technically difficult—it’s because the incentives haven’t lined up. ​ The Information (paywalled) reported that OpenAI is forecasting 220 million paid subscribers to ChatGPT by 2030, which it compares favourably to Spotify’s 280 million paid users. I’d suggest a better comparison is with Microsoft Office, which has just shy of 500 million paying users: to get to just under half of that in less than a decade would imply a remarkable level of ubiquity. Whether the forecast is achievable, or enough to justify OpenAI’s stratospheric valuation, is another matter. ​ However, not everything is going OpenAI’s way: a court ordered that internal communications related to the company’s use of pirated books for training, which OpenAI argued were legally privileged, should be revealed to authors suing the company. OpenAI’s in-house legal team will now be deposed. As I noted after earlier disclosure losses, there is a long road to resolution of this litigation. But Bartz v. Anthropic set an important precedent: a court suggested that training on lawfully acquired books may qualify as fair use, but obtaining books from pirated libraries may be copyright infringement. If OpenAI’s disclosures reveal extensive use of illicit material, pressure for a settlement would likely increase. ​ A major new analysis of 168 million UK job ads finds that demand for AI skills and creativity skills is rising together, especially since the release of ChatGPT, with the strongest overlap appearing in highly-skilled roles and the UK’s established creative clusters. For publishers, this reinforces a point many of us are already seeing in practice: AI skills are becoming a complement to good creative judgement, not a replacement. It also suggests that publishers who invest in both sets of skills, rather than treating them as competing priorities, will be better positioned as GenAI becomes embedded across publishing workflows. ​ Epic Games CEO Tim Sweeney started a debate in the games business by arguing that games marketplaces like Steam shouldn’t label games made using AI, because it will soon be part of how all games are made. Consumer signalling is an equally hot issue in books, though no-one is asking for less transparency, with at least four organisations offering an AI-free checkmark for book covers. ​ Mistral launched a new range of models. On a like-for-like basis, they are not really competitive with larger models from OpenAI, Google or Anthropic, but the interesting aspect is some of the smaller models, designed to be run locally on smartphones or laptops without an active Internet connection, which reportedly benchmark in line with competitor models four times their size. ​ Researchers from MIT, Northeastern and Meta have shown that some LLMs can over-rely on sentence structure rather than meaning, answering questions based on familiar grammatical “shapes” even when the words are nonsense. For book publishers, this helps explain why models sometimes hallucinate confidently in unfamiliar contexts, misinterpret bibliographic or rights information, or respond erratically to out-of-domain material. It also highlights a potential safety and quality risk when AI is used in workflows such as metadata creation, research, fact-checking or customer-facing discovery tools, where subtle shifts in phrasing can produce unreliable results.

Happy Black Friday. This Sunday marks three years since ChatGPT first appeared: a toddler in human years, already learning to run, leaving a mess in its wake, and showing signs of what it might grow into. It has even added a new word to our vocabularies—slop—and several of this week’s links explore whether AI-generated content has any value or is just that. ​ ChatGPT introduced a new shopping research feature this week, which matches products to a user query. I tried using it to find some gifts and it did a pretty good job of matching recently published books to recipients, though it pulled pricing and availability from a range of bookshops and publishers. It’s also unclear how frequently this data is refreshed or how well it handles backlist titles. These are questions anyone who cares about book discoverability should be thinking about too. ​ The new 2025 Edelman Trust Barometer on Trust and AI has some really interesting conclusions for anyone implementing AI in the workplace: about 60% of employees would accept AI aimed at productivity rather than cost saving, and the provision of high quality training increased employee willingness to use AI (unsurprisingly, I endorse that message). ​ Google Labs launched a new pedagogical experiment called Learn Your Way, which uses AI to transform linear textbooks into interactive learning materials. They claim an 11% improvement in retention scores for the AI texts over traditional ebooks. There’s a waitlist signup for users to upload their own PDFs: for publishers, that might be an experiment to consider, but it will also be interesting to see what copyright guardrails are built in to control use of third-party content. It also raises rights questions: at what point does an AI-modified textbook become a derivative work, and who owns those adaptive outputs? ​ On the subject of education, EdTechnical is running a forecasting competition on the impact of AI on education to the end of 2028—the same timeframe forward as from ChatGPT’s release to now. I’m sure plenty of subscribers have a view on this, and there are prizes for the best contributions. ​ Rahim Hirji’s newsletter has a fantastic list of nearly sixty AI habits that would get you sued, fired or embarrassed. Based on your score, you can determine whether you’re a cautious sceptic, a normal human or a walking liability according to Rahim’s scale. There’s a part of me that thinks that with such a fast-moving technology, if you’re not experimenting—and occasionally making mistakes—you’re not learning. But it’s certainly safer to learn from other people’s errors. ​ The AI platform Descript has published a useful guide to slop-free content creation, in particular repurposing an existing asset into different media formats. This is something content and marketing teams in publishers do all the time, and the guide provides some clear, practical advice. ​ New research suggests that AI-produced adverts can’t be dismissed as slop, as they perform considerably better than traditional commercial messages. I have questions about the research methodology, particularly the sample size. And the study measures perceived effectiveness rather than real-world conversion—still, it’s a sign that AI-generated creative may not be as disposable as many assume. ​ On the subject of AI slop, The Wrap returns to a subject that I’ve discussed before: the AI podcast studio Inception Point, now generating 3,000 podcast episodes a week, with a team of eight people. It’s easy to dismiss this as slop, and as I’ve previously argued, it hurts the signal to noise ratio for traditional podcast publishers. But it’s working, as Inception Point has over 400,000 subscribers. The piece goes into new detail about how the company operates. What’s striking is how tightly their production model is tied to algorithmic opportunity—filling keyword gaps with astonishing speed. It’s not hard to imagine the same volume x velocity playbook applied to ebooks. ​ A more traditional publisher, The Atlantic, has struggled with AI platforms crawling its site for data: one company tried to crawl over half a million times in a week. This piece looks at its strategy for managing access to its content. What’s particularly useful is the data-led approach that the Atlantic took, using its logs to determine which bots brought referral traffic and which should be blocked (less than a third brought any value). On that point, I’ve spoken to two publishers in the last six months who were proposing to make decisions about their websites without even reviewing their logs. Be more Atlantic. ​ Finally, I don’t think I’ll ever be comfortable listening to my own voice, but I did a very traditional podcast interview with the brilliant Alison Jones this week, talking about our respective careers in books, digital change, and the impact of AI on publishing.

This week has been full of stories that show how fast the conversation around AI, authorship, and creative integrity is moving. From new research on writers’ concerns to real-world disputes over AI-generated artwork, it’s clear that evidence and transparency matter more than ever. ​ A new research report from the Minderoo Centre for Technology & Democracy at the University of Cambridge received significant media attention this week for its headline claim that half of UK novelists believe AI will replace their work (there’s a lot more beyond the headline and I really encourage you to read it, especially if you’re a trade fiction publisher). It’s a strong piece of exploratory, qualitative research, though its focus on novelists alone excludes the equally important perspective of non-fiction writers (particularly since many authors work across both fiction and non-fiction). I have expressed some doubts about the methodology and whether one can reasonably draw population-level inferences from a convenience sample of 258 authors. But this is really to say that I hope it demonstrates the topic is important enough to do a larger scale study across types of authorship. ​ Coincidentally there’s a very useful piece by Marc Zao-Sanders in HBR on how to make sense of research on how people use AI, which makes important points about vested interests and inherent bias. For clarity, I’m not suggesting these are problems with the Cambridge research, only that the principles are worth keeping in mind with every data source on the subject (including what I write). I think I’m going to take this sentence from Marc’s piece as the mission statement for this newsletter: “The path to a sensible, defensible, and useful view of what’s going on lies in the synthesis of many different sources.” ​ Second only to the Cambridge research in press coverage this week was the news that two eminent authors were disqualified from a leading literary prize in New Zealand for the use of AI in their cover artwork, after new rules were instituted by organisers. This highlights a number of issues: the fact that the authors were unaware of the use of AI by their publisher, the concomitant need for transparency between authors and publishers, and the practicality of applying rules on this fairly and consistently. ​ On this subject, I came across an interesting practical AI feature this week: Google Gemini now has the ability to look at an uploaded photo and detect SynthID watermarks that are added by Google’s own image generator to determine whether it is likely to be real or generated. It takes one to know one? This could become a helpful tool for publishers trying to verify the provenance of submitted images. ​ However good automated detection gets, it’s never going to be completely foolproof in detecting AI, and there’s a real problem with false positives. How do you defend yourself if you’re accused in error of using AI? (This is far from a theoretical problem: I know book publishers this has happened to in the last year.) This is a really interesting case study from the brilliant Watershed in Bristol, which was accused of using AI in its marketing. Their response is really clear, and includes an explanation from their designer. It’s a model of clarity, and it raises an uncomfortable question: how many of us could offer an equally confident and well-documented rebuttal of our creative processes? ​ I suspect Google Scholar is one of the company’s lesser-known offerings, but it’s an essential tool for many researchers and academic publishers. This week Google released an updated, AI-powered search called Scholar Labs in limited preview. This is particularly helpful in providing a short summary of relevance to a search topic for each item returned, and could be especially valuable in the exploratory stages of research or literature review. ​ I’m always interested in AI case studies from other industries, and this post from broadcasting about developing a complex content workflow in under thirty minutes offers a compelling look at how AI can accelerate production. Of course, the prerequisite for doing this was having an MCP server that already interacted with key systems: the book or journal publishing equivalent would first require an existing integration with bibliographic databases, content management systems and other infrastructure. But the underlying principles hold true. I particularly liked this assessment of the developer’s role in the results: “To be honest, I don’t see a world where AI replaces engineers. It’s more about all engineers operating at a fundamentally different velocity, albeit constrained by purpose. The knowledge I’ve accumulated over three decades didn’t become irrelevant—it became leverage. I knew what to ask for. I could evaluate whether what [AI] produced was sensible… The AI handled the tedious translation of intent into implementation, yet the customer still owns the ‘purpose’ that drives the intent.”

There’s an interesting new report out from the Minderoo Centre for Technology and Democracy at Cambridge University on the impact of AI on the novel, authored by Dr Clementine Collett. There’s much that I completely agree with and that accords with my experience of working with book trade on AI. But I have some constructive thoughts about the framing/communication.