Book Description
If you published an ebook in 2025 and felt like your usual playbook suddenly stopped working — solid keywords, a decent launch week, maybe a few ads — you weren't imagining it, and you're not alone. Something changed underneath Amazon's search bar, and most authors still don't fully understand what. Amazon's search and recommendation system went through one of the biggest structural changes in its history, and most of what worked under the old rules quietly stopped mattering as much.
The shift is real, it's documented, and it has specific, practical implications for how you write your ebook's metadata, where your traffic comes from, and even how you think about pricing. This post breaks down what's actually confirmed, what's still informal industry shorthand, and what it means for your next ebook listing.
The Search Engine Behind Your Book Just Changed
For most of Amazon's history, its search system worked the way most people assume search works: it matched the words in your listing against the words in a shopper's query. Write "psychological thriller" in your title, and Amazon could match you to someone searching "psychological thriller." Miss the exact phrase, and you were often invisible no matter how well your book actually fit what the reader wanted.
That changed with the integration of COSMO, a semantic knowledge system Amazon built and published research on in 2024 together with researchers at HKUST. Instead of just indexing words, COSMO organizes shopping behavior into relationships — who a product is for, what problem it solves, what situation it's used in, and what else people buy or search for alongside it. For books, that means the system increasingly understands things like "quick, funny read for someone recovering from a rough week" as a coherent request, even if none of those exact words appear in your title or description.
In practice, this rewards listings that clearly communicate who the book is for and why — not just what genre box it fits into. A description that reads as generic marketing copy gives COSMO very little to work with. A description that's specific about mood, audience, and context gives it a lot.
Meet Rufus: Amazon's AI Assistant Is Reading Your Reviews, Not Just Your Blurb
The customer-facing part of this shift is Rufus, Amazon's generative AI shopping assistant. Since launching in early 2024, Rufus has grown to hundreds of millions of users, and Amazon has confirmed on earnings calls that AI-assisted shopping now drives billions of dollars in incremental sales. This is not a minor experimental feature — it's a mainstream part of how a growing share of readers now search.
Rufus works on a retrieval-based framework: when someone asks it a question — "what's a good beach read that's actually funny, not just fluffy" — it pulls from your listing text, your reviews, your Q&A section, and broader context, then generates a natural-language answer rather than a list of keyword matches.
Here's the part that matters most for ebook authors: Rufus weighs what your reviewers actually say as heavily as what your blurb promises. If your description says "unputdownable, fast-paced thriller" but your reviews consistently describe it as a slow-burn character study, that mismatch doesn't just risk a disappointed reader — it actively confuses the system trying to match you to the right audience. Honest, accurate positioning now has a direct, measurable payoff.
How Amazon Discovers Books in 2026 and Beyond
Title • Subtitle • Description • Categories • Keywords
Semantic understanding of audience, themes, mood, context and reader intent
Combines metadata, reviews, Q&A and product information to answer natural-language questions
Reviews • Conversion Rate • External Traffic • Engagement
Amazon identifies which readers are most likely to enjoy your book
Better recommendations • More qualified readers • More sales
Is There Really an "A10" Algorithm?
You'll see a lot of self-publishing blogs and SEO consultants refer to "the A10 algorithm" as the name for this whole shift. It's worth being precise here, because it affects how much weight you should put on any specific claim you read about it: Amazon has never officially confirmed a system called "A10." What is actually documented, with published research and confirmed company statements, is COSMO and Rufus.
"A10" functions as informal shorthand — a convenient label the author community uses to describe the cumulative, observable effect of these changes. Whatever you call it, the practical takeaway is the same: semantic relevance and genuine reader engagement now carry more weight than simple keyword density ever did. We'd rather tell you that plainly than repeat a term as if it were official Amazon terminology when it isn't.
How to Tell If Your Listing Is Already Losing Ground
Most authors don't get a warning when their listing starts underperforming under the new system — sales just quietly soften, and it's easy to blame the market instead of the metadata. A few signs worth checking right now:
-
Your description reads like every other book in your genre. If you could swap your blurb with a competitor's and nothing would feel off, there's nothing distinctive for a semantic system to latch onto.
-
Your backend keywords are single, generic words. "Romance." "Thriller." "Fantasy." These are the exact phrases every competing title is also using — and the ones least likely to match a specific, natural-language request.
-
Your Author Central profile is blank, or close to it. No bio, no photo, no linked website. That's a gap in the kind of authority signal discussed above.
-
Almost all of your traffic comes from Amazon's own ads. If external, organic interest is close to zero, you're relying entirely on the one traffic source that carries the least weight in this new system.
-
Your last metadata update was more than a year ago. Categories, keywords, and reader expectations all shift — a listing that hasn't been touched since launch is optimized for a version of Amazon that no longer fully exists.
If two or more of these sound familiar, the good news is that all of them are fixable in an afternoon — and none of them require rewriting your book.
What This Means for Your Ebook's Metadata
A few practical shifts follow directly from the above:
-
Write for the request, not the genre label. Instead of cramming "romance" or "thriller" into every field, use the specific, natural phrases a reader would actually type or say out loud — the mood, the trope, the comparable feeling. "Enemies-to-lovers small-town romance" gives the system far more to work with than "romance novel."
-
Don't waste space repeating your title. Amazon already indexes the words in your title and subtitle — using your seven backend keyword fields to repeat them is wasted space. Use that space for synonyms, tropes, and reader-facing language your title doesn't already cover.
-
Backend fields are measured technically, not just by character count. The field lengths you see in KDP's interface don't map cleanly onto how the indexing system measures them — special characters and accented letters can count differently than plain ones. If in doubt, keep phrasing simple and avoid packing a field to its absolute visual limit.
-
Ebooks are your fastest lever. Unlike a print edition, you can update an ebook's metadata, description, and categories instantly, with zero reprinting cost or delay. If you're testing how your listing performs under this new system, your ebook edition is where to run those tests first.
For a deeper look at how Amazon's underlying ranking mechanics actually work — conversion signals, category strategy, and the "30-day window" effect — see Inside the Amazon Algorithm: How Book Visibility Really Works.
Amazon Search: Before vs. After COSMO and Rufus
| Before COSMO & Rufus | After COSMO & Rufus |
|---|---|
| Exact keyword matching | Semantic understanding of reader intent |
| Genre labels carried most of the weight | Audience, mood, themes and context matter more |
| Keyword density | Natural language relevance |
| Book description alone | Description, reviews and metadata work together |
| Amazon Ads were often enough | Qualified external traffic has greater value |
| Individual book signals | Author authority and trusted web presence |
External Traffic Is Doing More Work Than Ever
One consistent thread across how sellers and consultants describe this new system: traffic arriving from outside Amazon — a newsletter, a blog, a promotion site, social media — increasingly carries more weight than traffic generated entirely by Amazon's own internal ads. The logic makes sense from Amazon's side: external traffic that converts is evidence of real demand the platform didn't have to manufacture itself.
There's an important catch, though. Not all external traffic helps — some of it actively hurts. If you drive a flood of untargeted clicks to your book page and few of those visitors buy, your conversion rate drops, and a falling conversion rate is a negative signal, not a neutral one. Quality and relevance of the traffic matter as much as the volume.
This is exactly why a targeted, genre-matched promotion works differently than a scattershot ad campaign or a random social media push. When you list your book through a Goodkindles promotion package, the readers seeing your book are already there because they're actively looking for books in your category — not cold traffic pulled in by an unrelated ad. That's the kind of external traffic that's more likely to convert, and conversion is what the system actually rewards.
For a real example of what happens when targeted external traffic and a well-optimized listing come together, see our case study: How One Indie Author Made $10k in a Month on Kindle.
The Print Royalty Change Most Ebook Authors Missed
Here's a confirmed, factual change that's had less attention than it deserves: as of June 2025, Amazon reduced the royalty rate on paperback and hardcover books priced below $9.99 in the US marketplace, from 60% down to 50%. This was the first change of its kind to print royalties in the program's history.
Ebooks were not affected. The familiar 70%/35% royalty structure for ebooks priced between $2.99 and $9.99 remains exactly as it was.
The practical effect is subtle but real: for many indie authors, particularly those pricing accessibly for a wide readership, this quietly widens the profitability gap between a paperback and its ebook edition. That's not a reason to abandon print — plenty of readers still want a physical copy, and print builds legitimacy — but it is a reason to make sure your ebook edition gets the lion's share of your promotional attention and budget going forward, since it's both the format most readers will actually buy and the one where your margins are strongest.
If you want to go deeper on how to price your ebook for maximum royalty and visibility under the current rules, see The Science of Pricing Your Kindle Book for Maximum Sales.
Seller Authority: Why Your Presence Beyond Amazon Feeds Back Into It
A concept increasingly discussed by self-publishing consultants — again, not an officially confirmed Amazon term, but a widely observed pattern — is that Amazon's system appears to weigh accumulated, account-level trust alongside individual book signals. Things like the age and consistency of your KDP account, how complete your Author Central profile is, and whether there's a credible presence for you as an author outside of Amazon itself.
An empty Author Central page with no links anywhere else on the web gives the system very little to anchor "who is this author" to. A complete profile, connected to an active author website, does the opposite — it gives both Amazon and AI systems like Rufus a clearer, more confident picture of who you are and what you write.
This is one of the more overlooked reasons a listing on Goodkindles is worth more than a single day of exposure. A permanent listing on a site that's been indexed and trusted for over 15 years is itself a durable external signal — and the Professional package takes this a step further by including a spot on our Authors' Website List, giving your own author site a link from an established, high-authority domain. It's a small addition with outsized long-term value if you're building a durable author brand rather than promoting one book in isolation.
Where Goodkindles Fits Into This Shift
We didn't design our submission process around this shift by accident. Every book listed on Goodkindles now goes through a five-question intake — covering ideal reader and moment, core themes and atmosphere, honest comparisons to well-known titles, the real story behind the book, and its central question or struggle. Structurally, that's almost exactly the kind of information COSMO's knowledge graph organizes around: who something is for, what context it fits, and what it's genuinely like — not just what category it sits in.
Every listing also carries structured data behind the scenes that explicitly tells search engines and AI systems what the book is, who wrote it, and what it's about — the same kind of clear, machine-readable signal that Rufus and similar AI shopping assistants rely on when synthesizing an answer for a reader. You can read the full story of how and why we rebuilt every listing this way in Why Every Goodkindles Listing Just Became AI-Readable.
As always, we want to be direct about what this does and doesn't promise: no platform, including ours, can guarantee a specific ranking outcome inside a system it doesn't operate. What we can do is make sure your book is represented with the kind of clear, structured, genuinely specific information that both human readers and today's AI-driven discovery systems are increasingly built to reward.
Frequently Asked Questions
Do I need to remove all my old keywords and start over?
Not necessarily. Review your existing backend keywords and description for generic, single-word phrases and keyword-stuffed lists, and replace them with specific, natural phrases a reader would actually type or say. If your listing already speaks clearly to a specific reader and mood, you're likely in good shape.
Does the royalty change affect Kindle Unlimited page reads?
No — the June 2025 change applies specifically to paperback and hardcover royalty rates. KDP Select and Kindle Unlimited page-read payouts are calculated separately and were not part of this change.
Is "A10" just a myth, then?
Not a myth exactly — more a useful shorthand without an official name behind it. The underlying changes it describes (COSMO's semantic matching, Rufus's growing role, and a general shift toward reader-behavior signals) are real and well documented, even though Amazon hasn't confirmed a system called "A10" itself.
Will listing my book on Goodkindles guarantee better Amazon rankings?
No, and we wouldn't want to tell you otherwise. What a targeted, genre-matched promotion can do is bring real, interested readers to your book page — the kind of external traffic that's more likely to convert, which is one of several signals this new system is built to notice.
Bringing It Together
Three things are reshaping ebook visibility on Amazon right now: semantic relevance replacing simple keyword matching, external traffic that actually converts carrying more weight than internal ads alone, and account-level author authority increasingly factoring into how confidently the system ranks your books. None of these are reasons to panic — they're reasons to be more specific, more honest, and more deliberate about where your promotional effort goes.
If you're ready to put a well-optimized ebook in front of readers who are already looking for it, take a look at our current packages:
-
Standard ($39) — A permanent, AI-optimized listing on our homepage and at least one genre category, plus promotion to our social media and newsletter audience of 100,000+ readers.
-
Standard Plus ($59) — Everything in Standard, with placement in at least two genre categories and a double promotional cycle for extended reach.
-
Professional ($69) — Everything in Standard Plus, with placement in at least three genre categories and a spot on our Authors' Website List — a durable, external authority signal for your author brand.
Whichever package fits your book, the goal is the same: a listing built for the way readers — and the AI systems increasingly helping them — actually search today. Explore Goodkindles promotion packages →
---
This post was prepared with the assistance of AI tools and reviewed by our editorial team.