Remember when Amazon SEO was like throwing spaghetti at a wall to see what stuck? Those days are over. The new AI systems are sophisticated enough to spot—and actively punish—the tricks that used to work.

I learned this the hard way when one of my books seemingly vanished overnight. No algorithm update announcement, no email from Amazon—just suddenly invisible. After digging into what happened, I realized I’d violated some of the new “AI etiquette” rules without even knowing they existed.

Here’s how to avoid getting ghosted by Amazon’s AI.

The Reality Check

If your travel memoir is stagnating while books with fewer reviews keep getting recommended, it’s likely because COSMO decided those books were a better fit for what readers were actually searching for. The AI isn’t trying to hurt you—it’s trying to connect readers to the right books. But that means being specific about who your book is for, not just what it covers.

The Keyword Stuffing Death Kiss

This is probably killing more books than any other single factor right now. If your title looks like this: “Romance Billionaire CEO Enemies Lovers Small Town Second Chance Standalone Series,” you’re actively confusing the AI about what your book actually is.

Why this backfires: The AI can’t determine your “semantic relevance” if your metadata is a confused mess of every possible keyword. It needs clear, consistent signals to understand what reader problem your book solves.

The fix: Choose 3-4 core concepts that genuinely describe your book’s main appeal. Be specific rather than broad.

Instead of: “Romance Contemporary New Adult College Billionaire CEO Boss”

Try: “A workplace enemies-to-lovers romance perfect for fans of emotional contemporary fiction”

The AI rewards clarity over comprehensiveness.

Ignoring Negative AI Feedback (The Thumbs Down Problem)

Here’s something most authors don’t know exists: When Alexa for Shopping gives information about your book to a customer, there’s a little thumbs-down button they can click if the AI got something wrong.

If readers consistently indicate the AI is providing incorrect information about your book, you get deprioritized. The system learns that your metadata isn’t reliable.

What to watch for:

• Discrepancies between your book description and what the AI tells customers
• Genre mismatches (your cozy mystery being described as a thriller)
• Incorrect series information or reading order
• Wrong content ratings or trigger warning information

The fix: Regularly search for your books using the AI assistant and see what it says about them. If you spot errors, submit a ticket to Amazon’s help center immediately.

The Metadata Mismatch Trap

The AI is constantly testing whether your keywords match reader behavior. If you tag a gentle cozy mystery with “psychological thriller” keywords, the AI will notice that thriller readers don’t convert on your book and stop showing it to them entirely.

This creates a death spiral: Wrong keywords attract wrong readers → poor conversion rates → AI learns your book doesn’t satisfy the intent → reduced visibility across all searches.

The solution: Audit your keywords against your actual reader reviews. What language do satisfied readers use to describe your book? Those should be your keywords.

The Out-of-Stock Invisibility Cloak

This one’s simple but deadly: If your book isn’t immediately available for purchase, the AI removes it from conversational recommendations entirely.

This includes:

• Paperbacks temporarily out of stock
• Kindle books caught in publishing delays
• Audiobooks in production limbo
• Any “coming soon” status

The AI won’t frustrate customers by recommending unavailable books, so availability becomes a “hard filter” for AI discoverability.

The Review Quality Spiral

Bad reviews don’t just hurt your feelings—they actively train the AI that your book doesn’t satisfy reader intent. But it’s not just about star ratings anymore.

The AI analyzes review content for specific signals:

• Did the book deliver what the description promised?
• Are readers surprised by content, pacing, or genre elements?
• Do reviews mention the book being “not what I expected”?

Reviews that hurt AI discoverability:

• “This wasn’t really a romance” (genre mismatch)
• “Much shorter than expected” (expectation mismatch) 
• “Too much explicit content” (content warning failure)
• “Couldn’t get into it” (reader targeting failure)

The Cross-Platform Consistency Requirement

The AI builds knowledge about your books across Amazon’s entire ecosystem—Kindle, Audible, physical books, even Prime Video if you have adaptations.

Inconsistent information across platforms confuses the AI’s knowledge graph. Make sure your book description, genre tags, and series information are identical across:

• Kindle store
• Audible (if you have audiobooks)
• Physical book listings
• Author Central profile
• Any A+ Content

The Sponsored Content Backfire

Here’s a sneaky one: If you’re running Amazon ads that drive traffic to books with poor conversion rates, you’re actually teaching the AI that your book isn’t a good match for those search terms.

The AI learns from ad performance just like organic search.** Poorly targeted ads can hurt your organic discoverability.

How to Audit Your Books Right Now

1. Search for your books using Alexa for Shopping. What does the AI say about them? Is it accurate?
2. Check your keyword-to-review alignment. Do your keywords match the language in positive reviews?
3. Review your cross-platform consistency. Are your book details identical everywhere?
4. Analyze your conversion rates. Are you attracting the right readers?
5. Monitor your availability status. Any gaps in availability hurt AI recommendations.

The Bottom Line

The new Amazon ecosystem rewards authors who clearly communicate what they’re offering and deliver on those promises. It punishes confusion, inconsistency, and attempts to game the system.

Think like a helpful librarian, not a keyword optimizer. The AI is trying to match readers with books they’ll love. Help it do that job by being crystal clear about what reading experience you provide.

Have you noticed any of your books suddenly becoming less visible? What changes did you make that helped or hurt your discoverability?

I’ve got some news that might sting a little: Only 22% of books showing up on page one of traditional Amazon search results actually match what the AI assistant recommends when customers ask questions.

Let that sink in. If you’re celebrating because your book shows up on page one for “romantic suspense,” there’s a 78% chance the AI is actively steering readers *away* from your book when they ask for personalized recommendations.

As a tech writer, this makes perfect sense to me. We’ve always known that what users say they want and what they actually need are often different things. Amazon’s AI has just gotten really, really good at bridging that gap.

The New Reality: Ranking vs. Recommending

Here’s what’s happening: Amazon now has two separate discovery systems running in parallel.

The old system shows you search results based on keyword matching and traditional SEO signals. This is where most authors are still focusing their optimization efforts.

The new system uses conversational AI to understand what readers actually want and actively recommends specific books. This is where 250+ million customers are increasingly going for book discovery.

The problem? Most authors are optimized for the old system while readers are migrating to the new one.

They’re Reading Your Metadata Like Tea Leaves

Unlike the old keyword-matching system, this AI pair analyzes something closer to what you could call your book’s intent signature. They study what people buy after searching terms related to your book, which books get purchased together, how readers describe your book in reviews, and whether your metadata actually matches what you deliver.

I learned this slowly with “Valley of Flowers.” I’d categorized it broadly under Travel and Adventure—too vague, apparently, for the AI to do much with. Once I repositioned it specifically for “hiking enthusiasts seeking friendship stories,” visibility picked up. The AI had enough to go on.

Strategy #1: Optimize for Use-Cases, Not Keywords

Stop thinking like an Amazon algorithm. Start thinking like a helpful bookstore employee.

When someone walks into a bookstore and says “I need something to read on the plane,” a good bookseller doesn’t just hand them any book with “travel” in the title. They ask follow-up questions and make contextual recommendations.

Amazon’s AI is doing the same thing now. It needs to understand the *context* of when and why someone would choose your book.

Instead of this: “A thrilling romance with unexpected twists”

Try this: “Perfect for readers who love romantic suspense but want to sleep tonight—thrilling without being disturbing”

Instead of this: “A cozy mystery series”

Try this: “Ideal for mystery fans who prefer puzzles over police procedurals, perfect for weekend binge-reading”

You’re giving the AI “information anchors”—specific contexts it can grab onto when making recommendations

Strategy #2: Treat A+ Content as Technical Documentation

This was my biggest aha moment as a tech writer. The AI isn’t just looking at your cover and reading your blurb. It’s parsing every piece of text in your A+ Content, including text within your graphics.

What most authors are doing: Using A+ Content for pretty visuals and marketing copy

What works now: Using A+ Content as a detailed product specification sheet

Include modules that clearly explain:

• What type of reader experience you deliver
• Specific mood and pacing expectations
• Content warnings or comfort levels
• Comparison points to similar authors/series
• Clear reading order for series

The AI reads this information and uses it to generate better recommendations and “sponsored prompts”—those suggested questions that lead readers to your book.

Strategy #3: The 4.0 Star Minimum

Here’s something concrete you can act on immediately: The AI typically won’t recommend books with ratings below 4.0 stars.

This isn’t about gaming the system—it’s about quality control. Amazon’s AI acts as a filter, and it’s not going to recommend books that other readers didn’t enjoy.

If you’re below 4.0 stars:

• Focus on reader satisfaction over discoverability 
• Consider whether your book description is attracting the right readers
• Look for patterns in negative reviews that you can address in future releases

If you’re above 4.0 stars but have low review counts:

• Gentle review requests become more important
• Consider strategies to increase engagement with current readers

Strategy #4: Write for Conversation

The AI uses your listing content to generate “sponsored prompts”—automated questions and recommendations it shows to customers. The clearer and more conversational your listing, the better those prompts will be.

Frame your bullet points as answers to reader questions:

• “Is this book appropriate for teens?” → Clear content guidance in your description
• “How steamy is this romance?” → Specific heat level information
• “Do I need to read the series in order?” → Clear series reading guidance
• “Will this keep me up all night?” → Specific pacing and intensity information

The Measurement Shift

Stop obsessing over keyword rankings. Start tracking:

• Conversion rates from views to purchases
• Read-through rates for series
• Review quality and sentiment
• Cross-selling to other books in your catalog

The AI is tracking all of this to determine whether you’re a good recommendation. High conversion rates signal that the AI is matching your book with the right readers.

What’s Working Right Now

Authors who are succeeding with the new system are those who:

1. Clearly define their ideal reader and reading context
2. Use specific, conversational language in their metadata
3. Maintain high reader satisfaction metrics
4. Provide detailed, scannable information in A+ Content

Coming up next

The mistakes that can get your books completely hidden by the AI—and how to avoid them.

What’s your current star rating, and have you noticed changes in your discoverability? I’d love to hear what patterns you’re seeing.

When I published “The Camino Ingles” back in 2017, Amazon book marketing was pretty simple. You’d load your title with keywords—”Camino Santiago Pilgrimage Walking Spain Guide”—upload your manuscript, and wait for readers to find you through the search bar.

That approach is mostly dead now.

If you’ve checked your dashboard lately and something feels off about how your books are surfacing, you’re not imagining it. Amazon has deployed two AI systems—Rufus (recently rebranded as Alexa for Shopping in the US) and COSMO—that have changed how travel books, memoirs, and adventure guides get discovered.

The New Sheriffs in Town

These aren’t search upgrades. Rufus is the conversational layer—the AI assistant that over 250 million customers now use to ask things like “What’s a good walking book for beginners?” or “I want to read about spiritual journeys.” COSMO is the reasoning engine underneath, using contextual logic to figure out what readers actually mean by vague searches.

When I dug into my Camino book’s performance over the past year, I found that most new readers weren’t arriving through traditional keyword searches anymore. They were asking Rufus things like “books about walking adventures in Spain” or “pilgrimage stories for inspiration.” The discovery path had quietly shifted.

So What Exactly Are These Things?

Think of them this way: Rufus is the friendly face, COSMO is the brain.

Rufus is the conversational AI assistant that over 250 million customers are now chatting with. Instead of typing “romance books,” they’re asking full questions like “What’s a good enemies-to-lovers romance that won’t make me cry?” Much more specific, much more telling about what they actually want.

COSMO is the sophisticated algorithm working behind the scenes. It’s using what Amazon calls “common sense knowledge” to figure out what readers actually mean when they search. When someone types “books for small apartments,” COSMO knows they probably want organization guides, not novels set in tiny homes.

They’re Not Attack Dogs—They’re Picky Curators

Here’s the thing that blew my mind as a tech writer: these aren’t simple keyword-matching bots anymore. They’re analyzing patterns—what people buy after specific searches, what books get purchased together, how readers behave after they click on your listing.

They’re building what’s called a “Knowledge Graph” around your book. Essentially, they’re creating a detailed profile of who your ideal reader is, when they’d want your book, and why they’d choose it over others.

This explains so much. Ever wonder why a book with fewer obvious keywords is outranking you? It’s because COSMO decided that book was a better “semantic fit” for what readers actually needed.

The Metadata Sniffing Operation

Unlike the old search system that just looked for keyword matches, this duo is reading between the lines of your entire listing. They’re analyzing:

• Your cover design and what genre it signals
• The emotional tone of your description 
• How your keywords relate to actual reader behavior
• Whether people who view your book actually buy it
• What other books your readers typically purchase

The crazy part? They’re doing this in real-time, constantly updating their understanding of your book based on new data.

What This Means for Your Books Right Now

Remember how we used to stuff titles with every possible keyword, hoping something would stick? That strategy is not just ineffective now—it’s actively hurting you.

The AI can’t figure out what your book is actually about if your metadata is a confused mess of every popular keyword. It needs clear, consistent signals about who your book serves and what problem it solves for readers.

Example: Instead of “Romance Enemies Lovers Billionaire CEO Small Town Second Chance”

Try: “A second-chance romance perfect for fans of emotional small-town stories”

See the difference? One is keyword soup, the other tells the AI exactly what kind of reader experience you’re delivering.

The Big Shift: Intent Over Keywords

This is the fundamental change every author needs to understand: Amazon’s AI doesn’t care about matching words anymore. It cares about matching intent.

When someone searches for “fast-paced thriller,” the AI isn’t looking for books that mention “fast-paced” the most times. It’s looking for books that readers actually describe as fast-paced, books that keep people turning pages, books with certain plot structures and pacing patterns.

Your job as an author has shifted from keyword optimization to intent optimization. You need to clearly communicate not just what your book is, but what experience it delivers and what kind of reader will love it.

What’s Coming Next

In my next post, I’ll break down exactly how to optimize for this new system—because there are specific strategies that work with AI curation vs. traditional keyword matching.

But here’s your homework for now: Look at your current book descriptions. Are they talking about your book, or are they talking about your reader’s experience? The AI can tell the difference, and it’s making decisions about your visibility based on that distinction.

Have you noticed changes in how your books are being discovered lately? Drop a comment and let me know what you’re seeing in your dashboard.