Keywords are a thing of the past

Traditional search engine optimisation has become vastly more complex. Anyone who wants to sell books today needs to understand how language models "think". But how do you get AI to not only give writing tips, but also make specific purchase recommendations? Sebastian Zahn, Performance Marketing Manager at Droemer Knaur, talks about the change in SEO for books and why we need to think in categories instead of titles.

Sebastian's methodology boils down to getting AI to trigger purchase recommendations. He shows that "explanatory mode" is the biggest enemy of sales – and how prompt triggering can revolutionise the book industry's metadata strategy.

We've known the game for 20 years: enter a keyword, hope for number 1. You say that era is over. How radical is the change really?

The change is fundamental because user behaviour has completely shifted.

People used to type static keywords like "crime bestseller" into a search box. Today, they converse – just like we are doing right now.

They ask Gemini (AI Mode, AI Overview) or ChatGPT questions like they would an experienced bookseller.

The currency is no longer the ranking on a list, but the exclusive recommendation in a response. The crucial question today is no longer "Who has the best SEO ranking?" but "Who does the AI recommend as the answer?" If we ignore this paradigm shift, we will become invisible.

You often talk about "explanation mode" as a problem. What do you mean by that?

That is the biggest misunderstanding when dealing with generative AI. The systems are trained primarily to impart knowledge.

When a user asks, "How do I write a crime novel?", the AI explains the writing process.

That's nice, but it doesn't sell a single book for us as a publisher. Our strategic task is to force the AI to switch modes – out of explanation mode and into transaction mode, i.e. sales consultant mode. To do this, I have developed a mechanical formula called "prompt triggering".

How does this system work technically?

It is based on a three-element system. In order for the AI to ultimately say "Buy this book", the user's prompt must contain three specific triggers, which we must match using our data. If one is missing, the result remains vague.

Firstly: the intent trigger, i.e. the user's clear decision to buy or search.

Secondly, the category trigger – this is our playing field, for example psychological thrillers or New Adult.

And thirdly – this is our most important lever – the criteria trigger. This involves specific attributes such as "unpredictable ending", "dark atmosphere" or "Spiegel bestseller level". Only when these three come together does the AI generate a specific product as a response.

Let's make that concrete. How do you apply this to Droemer Knaur's programme?

Let's take a title by Vincent Kliesch. Today's readers are often looking for a feeling or an intellectual challenge.

Let's imagine the user tells the AI: "I need new reading material (intention), specifically a psychological thriller (category) that challenges me intellectually and has plot twists I don't see coming (criteria)."

At that moment, the algorithm searches for books that are mathematically most strongly associated with the attributes "mentally challenging" and "surprising".

If we have optimised our metadata, blurbs and press work so that our title aligns perfectly with these attributes, then the AI must name this title.

Not by chance, but by mathematical probability.

What does this mean for metadata maintenance in publishing houses? Do we need to re-tag our books?

Absolutely. We need to move away from thinking purely in terms of titles and start thinking in terms of criteria. We need to ask ourselves: What atmosphere are we simulating? Are our romance novels classified as "heart-warming" or "emotionally stirring"? Are we perfectly occupying the "cosy crime" niche for holidays? How strong is the brand authority of a non-fiction author?

It's not magic. It's understanding how machines simulate human purchasing decisions. If we understand which trigger elements our readers use, we can influence the AI's response. Our goal is for publishers and authors to be the answer.