Tired of guessing how Amazon’s algorithm actually works? Sick of conflicting "guru" advice that never lines up with your actual sales data? For years, Amazon’s ranking and recommendation logic felt like a total black box for most sellers. We all picked up surface-level takeaways like "conversion rate matters" or references to the A10 algorithm, but very few of us had the time or technical background to dig into actual official documentation.

Amazon does publish all of its core algorithm research publicly on the Amazon Science website, including papers on everything from semantic search to the COSMO knowledge graph system. The problem? These are dense academic papers written for data scientists, not eCommerce sellers. Even with generic AI assistants, parsing one paper and translating it into actionable takeaways used to take me a full day or more. That all changed when I started combining Gemini and NotebookLM. Now I can get clear, cited answers to every algorithm question I have in 60 minutes flat.

Let’s start with two of the most common seller myths I was able to bust immediately using this workflow, with evidence pulled directly from Amazon’s own research:

Myth 1: Ad sales suppress organic ranking

After cross-referencing multiple official Amazon algorithm papers, there is zero evidence to back up this claim. Every search-to-purchase action, regardless of whether the buyer clicked an ad or an organic Listing, counts as a strong positive signal for your Listing’s ranking.

I tested this myself with my own portfolio of high-ticket products priced at $100+ with healthy profit margins. We intentionally hold Top of Search (TOS) ad positions for our core keywords long-term, and ad sales make up 60-70% of our total unit sales for these ASINs. Even with that high ad share, our organic search positions for those same keywords have stayed locked in the top 3 spots on page 1 for months. If ad sales hurt organic rank, we would have seen our positions drop off long ago.

Myth 2: All ad formats contribute equally to organic rank lift

This is another common misconception. When it comes to boosting organic search rankings for specific keywords, Sponsored Products (SP) ads deliver far more weight than Sponsored Brands (SB) or Sponsored Display (SD) ads. This is not an arbitrary bias—it’s tied directly to how search intent signals are structured in Amazon’s ranking system.

You can verify this yourself during your next product launch. Cross-reference your SP search term report with organic rank movements for the same keywords in the weeks after launch, and you will see a clear correlation between SP conversions for a keyword and organic rank gains for that term. This does not mean SB or SD campaigns have no value—they are extremely effective for brand awareness and retargeting—but SP should be your primary ad format if organic rank lift is your core goal.

I’ve run dozens of other questions through this workflow and gotten clear, cited answers for every single one, but I wanted to focus first on the step-by-step process so you can test it yourself:

  1. Source official algorithm papers with Gemini

Start by prompting Gemini to act as an eCommerce search and recommendation system expert, with a focus on Amazon’s published research from Amazon Science and top academic conferences like KDD, WWW, and SIGIR. Specify the exact areas you want to research—for example, organic ranking logic, the COSMO knowledge graph, or Rufus generative AI features—and ask it to return a curated list of official papers with direct download links. This step alone saves hours of manual searching through academic databases.

  1. Build your private knowledge base in NotebookLM

Open Google’s NotebookLM, create a new notebook, and upload all of the paper PDFs you sourced in the first step. The biggest benefit of NotebookLM is that it only pulls answers from the exact materials you upload, so you never have to worry about hallucinated information or generic, unproven advice. Every answer it gives will be cited directly to the official Amazon papers you provided.

  1. Ask your most pressing operational questions

You can ask any question you’ve debated with your team or seen floating around seller forums, from broad strategy questions to specific edge cases. You can even ask it to translate academic findings directly into actionable operational steps, create training flashcards for your team, or run quick quizzes to test your own knowledge of algorithm logic.

One of the most useful insights I pulled from this process relates to Amazon’s Common Sense Knowledge Generation and Serving System (COSMO), the technology powering Amazon’s shift from keyword matching to intent matching in search. If you want to align your Listing with how COSMO works, these are the highest-impact adjustments you can make:

  • Prioritize use cases and target audiences in your Listing content, not just product features. If you sell non-slip shoes, for example, explicitly mention that they work for pregnant users or elderly users in rainy or snowy conditions, rather than only listing material and size specs. COSMO connects user intent to these use case signals, not just exact keyword matches.

  • Address the "why should I buy this" question directly in your A+ Content and product videos. COSMO powers features like the "Help Me Decide" tool, so your content should clearly explain how your product solves a specific user problem. For a four-person tent, for example, highlight that its windproof design makes it suitable for all-season camping, so the algorithm can connect your product to users searching for winter camping gear.

  • Add layered, intent-focused keywords to your backend Search Terms. COSMO breaks down broad search queries into hierarchical navigation paths (e.g., camping gear → winter camping gear → cold-weather sleeping bags). Pull these long-tail, intent-specific terms from your search term report and add them to your backend Search Terms to appear in filtered search results.

  • Pair complementary products for promotions and cross-campaigns, not just similar products. COSMO analyzes co-purchase behavior to identify functional connections between products (e.g., camera cases and screen protectors both serve the purpose of protecting a camera). When creating promotions or Sponsored Brands campaigns, group products that share a common use case, not just products in the same category.

  • Amazon’s own research shows that intent-aware models improve semantic matching performance by over 20%, so optimizing for these intent signals will have a measurable impact on your visibility.

This workflow has already changed how I run my business. I no longer waste thousands of dollars testing unproven algorithm hacks I see on social media, because I can pull official answers directly from Amazon’s own research in an hour. For team leads, it cuts down on training time significantly, since you can build a shared knowledge base of official, cited takeaways for your entire team to reference. For new sellers, it eliminates the confusion of conflicting advice and lets you validate your operational ideas quickly, without wasting months testing strategies that don’t work.

I’ve been testing a ton of AI-powered workflows tailored to Amazon sellers over the past two months, and I’ll be sharing more of the highest-impact ones as I refine them.

Have you used NotebookLM to analyze Amazon’s official docs or research? What algorithm myth do you want debunked most? Drop a comment below—I may answer your top questions in my next post.