How Different Platforms Decide Which Books Get Seen

We’ve all been there before: the joke that usually lands suddenly falls flat; watching everyone rave about a film you couldn’t stand; the song you’re convinced should be a hit never catches on.

And this is perhaps the best way to depict exactly what happens to books once they arrive on retail platforms.

Two titles can appear remarkably similar on paper: comparable metadata, positioning and marketing, yet algorithms may decide one deserves visibility while the other quietly disappears into the background.

So understanding the ways in which discoverability functions through different platforms is essential, if not, a make-or-break for books.

Perhaps unfortunate, yet a way of life is that publishers now aren’t only marketing to readers, but also to the systems that decide what they see. And if you want to play the game, you need to understand the rules.

Amazon

Shaping up to be a lot like human behaviour, Amazon’s algorithms pay attention to what’s hot – and ultimately what’s trending. The platform is powered through behavioural signals, picking up on what people like and rewarding books that appear to already be generating momentum.

P.S. Behavioural signals: think of them as the measurable actions users take on a platform that help the algorithms decide whether a book is valuable, relevant and most importantly, likely to convert.

While metadata still matters, particularly for categorisation and search relevance, Amazon increasingly behaves like a predictive commerce engine. The platform constantly measures these behavioural signals to essentially determine whether a book deserves wider visibility.

What the algorithm likes:

• Click-through rate
• Conversion rate
• Sales velocity
• Review activity
• Pricing behaviour
• Kindle engagement
• Ad performance
• Customer similarity modelling

Clearly then what Amazon algorithms prioritise is attention, meaning if a book is ‘having its moment’, it will almost instantly feed it into recommendation loops such as “Customers Also Bought”, place it onto home pages and plant it in email recommendations.

Google

On the other side of the spectrum lies Google, a favourer of trust rather than precipitate demands. The platform is more concerned with how much it can trust the book to best answer the query. So rather than responding to bursts of demand, Google looks for whether it genuinely understands a book.

And the way in which Google determines this is by: what the book is about, who the author is connected to, how authoritative the content appears, and whether readers searching for that topic are likely to find the result genuinely useful.

What the algorithm likes:

• Contextual relevance
• Structured metadata
• Schema markup
• Backlinks
• Author authority
• Semantic relationships
• Search intent alignment
• Topical consistency
• User engagement and dwell time

So unlike Amazon, where visibility often follows momentum, Google pays attention to confidence. If readers click on a result, stay with it, continue exploring and appear satisfied with what they found, the platform interprets this less as popularity and more as trustworthiness.

TikTok

Perhaps the most complex of all, TikTok’s algorithm is the forefront of riding the wave, in which the wave appears rather randomly and at a demanding speed.

The platform is not primarily driven by search, categorisation or even traditional discoverability logic. Instead, TikTok is powered by emotional velocity - how quickly a piece of content can provoke curiosity, reaction or obsession before the user scrolls away.

So in the name of books, this means how quickly you can entice people to like it, so in many ways, TikTok treats books less like products and more like social objects.

What the algorithm likes:

• Watch time
• Rewatches
• Comments
• Saves
• Shares
• Creator engagement
• Emotional reactions
• Rapid interaction velocity
• Trend alignment
• Visual identity

If enough people suddenly stop scrolling, react, comment and participate, the algorithm understands this as cultural relevance and pushes the content outward at immense speed. Backlist titles can become bestsellers overnight and single creators can create more discoverability than months of carefully planned campaigns.

It's notoriously the most frustrating of all platforms and yet, despite its unpredictability, TikTok remains one of the most powerful discoverability engines in publishing today. Just the other week we reported on our guide to cracking the platform’s discoverability logic, which you can read here.

Chat GPT

A newer contender to the party (and certainly not one that arrived quietly) is ChatGPT. Unlike traditional retail platforms, the way discoverability functions within AI systems is still fairly elusive, both to publishers and to the wider industry. There’s no public bestseller chart, no visible recommendation shelf and very little transparency around exactly why certain books surface while others do not.

And discoverability on the platform functions almost like a culmination of all the systems before it. Like Google, ChatGPT prioritises contextual understanding and authority. Like Amazon, it responds to signals of relevance and usefulness. And much like TikTok, it increasingly rewards the content, topics and entities that are already being discussed, referenced and reinforced across the wider web.

But unlike those platforms, AI systems do not simply rank results - they synthesise them, creating a kind of double-edged situation in which publishers have the opportunity to become embedded directly within the answer itself, while simultaneously risking complete invisibility if their books fail to be contextually understood by the system.

What the algorithm likes:

• Contextual clarity
• Authoritative sources
• Semantic relevance
• Trusted entities
• Structured information
• Strong topical connections
• Conversational usefulness
• Consistent web presence
• Interpretable metadata
• Widely referenced content

Essentially, the more contextually visible a title becomes: through reviews, citations, metadata, conversations, articles, recommendations and semantic connections, the easier it becomes for AI systems to confidently interpret and surface it.

The Key Takeaway

Every platform now behaves differently, rewarding different signals and interpreting value through entirely different forms of audience behaviour.

Amazon rewards momentum.

Google rewards trust.

TikTok rewards reaction.

ChatGPT rewards contextual understanding.

And while metadata remains foundational across all of them, metadata alone is no longer enough. And perhaps ironically, for systems built entirely on algorithms, discoverability is becoming increasingly human: rewarding attention, trustworthiness and cultural resonance.

The challenge, then, is not simply to understand metadata, but to understand behaviour: how platforms interpret it, amplify it and use it to shape discoverability altogether.

‍