How AI is driving more customers than social media for Fox & Fable bookshop

Fox & Fable is getting in early on a selling technology that is conversational and, potentially, well placed to take advantage of an agentic future even in a space crowded with established retailers. Its founders admit that while using a chatbot to find books is currently likely to send you to Amazon, much of this digital territory remains up for grabs.

Fang tells The Bookseller: “Our website is optimised for AEO (Answer Engine Optimisation), and we focus more research on what works than other retailers. Many retailers receive ONIX metadata [from publishers], but not all make that data available to AI search agents or optimise it properly for them.”

At a time when parts of the industry are already sounding the alarm about the death of BookTok, the narrowing of space for book reviews or other coverage in the media, and the sheer volume of books being published, Generative Engine Optimisation (GEO) is a huge priority for publishers. The idea of some retailers treating ONIX data in a way that is more or less useful to LLMs for specific shops, however, has been met with scepticism.

Graham Bell, executive director at EDItEUR, the trade standards body for ONIX metadata, says: “If [publishers] are updating metadata for GEO, then almost certainly they’re doing it for all the retailers. That metadata will be widely distributed, including through Nielsen. I’d be a bit disappointed if one bookstore was saying: ‘Update your metadata backlist and only send it to us.’ I’d be surprised too if publishers agreed to do this. Obviously their interest would be to get improved metadata out to everybody.” A further question Bell raises is: why wouldn’t publishers simply share their ONIX data directly with LLMs?

For publishers, the rise of bookselling via LLMs is an opportunity to ‘surface’ backlist and understand consumer behaviour. “It is a really high priority for us,” says Dibble. “We’re doing loads of work to make sure that our books and authors are visible in answer engines – that’s ChatGPT, AI mode on Gemini, Perplexity, all of them. They all have slightly different flavours of what they like.” Dibble confirms that while Pan Macmillan is working with Fox & Fable, they do not have a “special relationship” and are not sending “individualised metadata”.

She adds: “For us, metadata has never been a kind of ‘one and done’. I think you probably find that for most of the big publishers, and probably lots of the small ones too; it’s a constant evolving, re-energising of your metadata and the product data, so that it kind of works with all the algorithms to be visible. Now with LLMs, it’s a whole new ball game.”

Fox & Fable is a nascent business not old enough to have filed its first annual accounts, but it says its LLM-based search referrals are outstripping sales from social platforms. “We’ve seen a 400% rise month-on-month from ChatGPT-referred sessions of people going on ChatGPT, typing something in, and coming to our website because it’s been linked. This means it’s by far our fastest-growing channel – above paid social, above TikTok or Instagram shops. And, of course, these consumers are much more likely to convert because they’ve been recommended,” Fang says.

‘Consumers and booksellers will make their own choices as to what extent they engage with AI. What gives us confidence is the proven ability of bookshops to adapt’ – Booksellers Association MD Meryl Halls

Publishing might have been an unwitting early battleground in the development of AI when tech companies began either licensing or using pirated works to train LLMs, but bookselling is set to open up another front. Amazon recently killed its Rufus chatbot in favour of Alexa for Shopping, and it does not allow third-party LLMs to crawl it, meaning the information it contains will be limited off-platform as this space develops.

The wider implications for retail of the predicted shift in consumer behaviour linked to chatbots remain to be seen, but Meryl Halls, managing director of the Booksellers Association (BA), is confident that independent bookshops will endure. “The BA recognises that, as with any major technological shift, consumers, businesses and booksellers will ultimately make their own choices about how and to what extent they engage with AI. What gives us confidence is the proven ability of bricks-and-mortar bookshops to adapt and evolve,” she says.

Halls adds: “At a time when many people are seeking meaningful offline connection in an increasingly digital world, bookshops provide something unique, personal and irreplaceable. That experience cannot be replicated by Amazon, algorithms or AI. The continued growth of independent bookshops, with numbers expected to reach a 15-year high in 2026, demonstrates the enduring value readers place on these spaces.”

Fox & Fable’s founders are aware of negative attitudes to AI in their dealings with publishers who, they claim, “don’t really know what to do with us”. Forgan and Fang are clear that their approach does not require the ingestion of a copyrighted work, and that ONIX metadata is enough to feed the LLMs. Under US copyright law, the transformative ingestion of entire books might be viewed as “fair use”, but this is not the case under UK law. Either way, Fang and Forgan seem to understand the wariness around AI in the book trade. “I think AI use in publishing is dubious and morally a bit unsure, but I’m very pro AI use in bookselling and getting the books out there,” says Fang. “I think it’s great.”