If you are still writing book descriptions using “Attention, Interest, Desire, Action,” you aren’t just outdated. You are invisible.
For the last decade, the advice was simple. Write a hook. Agitate the pain. Tease the solution. Use the AIDA model (Attention, Interest, Desire, Action) to grab the human reader by the lapels and force them to click “Buy Now.”
In November 2025, that advice is actively killing your book sales.
While you were busy optimizing your copy for a human emotional response, Amazon quietly replaced the engine running its entire marketplace. They moved from the keyword – matching A9 algorithm to a tripartite intelligence system that fundamentally changes how books are discovered.
There is a new “Silent Reader” on your product page. It reads faster than any human, it remembers everything, and it decides whether your book lives or dies before a human eye ever sees it.
I’ve spent the last six months deconstructing this new “A11” ecosystem – specifically the interplay between the A10 ranking framework, the COSMO knowledge graph, and the Rufus generative AI.
Here is what I found: The old way of writing blurbs is insufficient. We are in the era of the Dual – Audience Protocol.
The Ghost in the Machine
Most authors think Amazon is still a search engine. You type “Romantasy,” and it looks for the word “Romantasy.”
That is no longer how it works. Amazon has evolved into an Answer Engine.
The new system, powered by the COSMO Knowledge Graph, doesn’t care about your keywords. It cares about your Entities. It maps relationships between concepts. It knows that “dragons” are semantically linked to “political intrigue” and “fourth wing” in a way that goes beyond simple text matching.
Then there is Rufus.
Rufus is the generative AI assistant that now sits between your book and the reader. When a user asks Rufus, “Find me a sci–fi book that feels like The Expanse but with more ground combat,” Rufus doesn’t scan your emotional hook. It scans your structured data.
If your blurb is 100% flowery prose designed to make a human cry, Rufus sees nothing but noise. It cannot extract the “knowledge triplets” (Subject → Predicate → Object) it needs to verify that your book matches the user’s request.
The Dual – Audience Dilemma
This creates a paradox.
The Human Reader still wants “Subjective Product Needs” (SPN). They want “vibes,” “competence porn,” and “emotional payoff.”
The AI Agent wants “Subjective Truth.” It wants hard data, specific noun phrases, and verifiable attributes.
If you write purely for the AI, your blurb reads like a technical manual, and humans bounce.
If you write purely for the human (the old AIDA way), the AI cannot index your book, and you never appear in the recommendation feed.
The secret to ranking in late 2025 is mastering the Hybrid Blurb.
The Secret Sauce: NPO and Data Blocks
I have developed a proprietary framework for rewriting metadata that satisfies both masters. It involves two key techniques that most authors – and even most publishers – are completely ignoring.
1. Noun Phrase Optimization (NPO)
Adjectives are fluff. Nouns are anchors. The algorithm ignores “heart – pounding action” (a vague sentiment) but it devours “kinetic bombardment,” “orbital drop,” or “enemies–to–lovers trope.”
By restructuring your descriptive copy to replace weak adjectives with strong, specific Semantic Entities, we can “teach” the COSMO Knowledge Graph exactly where your book belongs.
The Science: According to Amazon’s own research paper, “COSMO: A large–scale e–commerce common sense knowledge generation system,” the search engine has moved beyond keyword matching to map user intent using “Knowledge Triplets” (Subject → Predicate → Object). It needs structured context to understand that ‘Starship Troopers’ is semantically linked to ‘Bug Hunt,’ not just the word ‘Space.’
2. The “Data Block” Strategy
This is the most powerful tool in my arsenal. I’ve started implementing a structured “Data Layer” at the bottom of my clients’ blurbs. This is a specific sequence of bullet points designed exclusively for the AI to ingest.
Why? Because Rufus (Amazon’s AI assistant) is built on Retrieval – Augmented Generation (RAG) architecture.
The Science: As detailed in the AWS Machine Learning Blog, RAG systems do not just “know” things; they must retrieve factual data from a trusted source text to generate an accurate answer. If that source text (your description) is vague, the AI hallucinates.
The Data Block acts as that source text. When Rufus scans the page, it hits this block and instantly indexes the book’s Tropes, Heat Level, Tone, and Comps.
The result? When a user asks a specific question – “Is this book spicy?” or “Is the science realistic?” – the AI answers with a confident “YES” because it found the answer in your Data Block, rather than guessing from a confused review.
References & Further Reading:
Amazon Science: “COSMO: A large–scale e–commerce common sense knowledge generation and serving system at Amazon.” Link to Paper
AWS Machine Learning Blog: “The technology behind Amazon’s GenAI – powered shopping assistant, Rufus.” Link to Article
Don’t Let the Algorithm Bury You
The “A11” era is not about tricking the system. It’s about speaking its language.
The authors who are winning right now aren’t just writers; they are architects of semantic data. They understand that a blurb is no longer just sales copy – it is code.
I have spent the last year mapping the “Knowledge Triplets” for genres ranging from Military Sci–Fi to Romantasy and Pre–Colonial History. I know exactly which levers to pull to make the Halo Effect work for you, driving external traffic that sticks because the metadata promise matches the reader experience.
Your book deserves to be found. But in 2025, being good isn’t enough. You have to be indexed.
Ready to future–proof your book launch?
If this resonated with you and your books have hit the “Sudden Stop,” the full 37‑page playbook is waiting inside my paid Substack.
Become a paid subscriber to get instant access to The Algorithmic Copywriter – the complete hybrid blurb framework with step‑by‑step prompts and genre‑specific examples.
You can also buy the Algotithmic Copywriter eBook without subscribing for $19.99.
I would like to know if anyone else has tried Novel Report and received questionable results. Several of the recommendations just really don't make sense for my book in both keywords and categories.