I’ve spent the last 6 months testing how Amazon’s latest algorithm changes impact organic rankings for my 7 private label ASINs across the US and EU markets. I’ve put together this breakdown of what’s actually changed, what still drives the vast majority of ranking decisions, and how you can adjust your Listing optimization strategy to stay ahead of the curve.

We first need to cover the A9 algorithm, as it remains the core framework driving the vast majority of search ranking decisions. A9 takes its name from the word “algorithm”, as the letter A is followed by nine other letters in the word. It was first developed by Amazon’s A9 subsidiary in 2003, which focused on search and ad technology development. Amazon shut down the standalone A9.com domain in 2019 as it rolled out updates to the core system, but the underlying A9 logic still forms the base of all current search ranking.

The original A9 system prioritizes 10 core signals when ranking Listings for any given search query.

• Keyword relevance between search terms and Listing content

• Conversion rate for the specific keyword

• Long-term sales history for the product

• Ad impression volume

• Click-Through Rate (CTR)

• Positive review rate

• Price point alignment with user expectations

• Available inventory levels

• Backend search term optimization

• Product image quality and relevance

The three highest-impact signals in this system are conversion rate, keyword relevance and product sales history.

Between 2019 and 2020, Amazon rolled out a series of updates to the core A9 framework that the seller community commonly refers to as A10. Amazon never released an official name for this update, so A10 is simply the standard industry term for this iteration of the algorithm. This update added five new ranking signals to the core A9 framework.

• Organic Sales generated directly from non-paid search results. These sales signal strong product-market fit and relevance, as they come from users who found the Listing without paid promotion.

• Geographic inventory distribution. Amazon will rank Listings higher for shoppers located close to warehouses where the product is in stock, to cut down on delivery times and costs. If you only stock inventory in a small portion of a market’s fulfillment centers, shoppers in regions further from your stock will see your Listing appear much lower in search results.

• Sales driven by Off-site traffic. Conversions from sources like social media platforms, brand websites, industry blogs and video content signal strong brand presence outside of Amazon, which boosts your internal ranking weight.

• Sales from internal association placements. Conversions from sections like Frequently Bought Together or Recommended for You signal that your product outperforms competing options for related user needs, which lifts your core search ranking.

• Seller Authority. This includes signals like your Buy Box win rate, seller feedback rating, how long your account has been active on Amazon, and your overall store return rate. While Amazon still prioritizes individual Listing performance over store-level metrics, these signals now play a small but measurable role in rankings.

In 2024, Amazon rolled out its next major algorithm update, officially named the Customer Obsession Search, Merchandising, and Optimization system. Most sellers refer to it as COSMO for short. This is the first major update with official public documentation, which is available via the Amazon Science platform for anyone to review. Internal Amazon A/B testing showed the update drove a 0.7% sales lift across tested product categories after integration into the search system.

COSMO is built on large language model technology, and it focuses on understanding full user intent rather than just matching individual keywords. The original A9 system focused heavily on exact keyword matches, sales volume and review metrics, which sometimes led to less relevant search results for users. COSMO delivers far more personalized and contextually relevant search results by analyzing user behavior data to build customer-centric knowledge graphs of shopping intent.

This shift means the old tactic of keyword stuffing no longer works as well as it used to. Previously, many sellers optimized their Listings by repeating broad keywords over and over to hit relevance signals. A glove Listing might have included generic terms like gloves, warm gloves, winter gloves and high-quality gloves stacked in the title and bullet points. Now, you will see far better performance by building your Listing copy around specific use cases and user segments. You can structure content for use cases like cold weather gloves for motorcycle commuters, waterproof gloves for backcountry skiing, touchscreen-compatible gloves for office workers who walk to work, or durable protective gloves for professional chefs. You can also tailor content to specific user groups like children, elderly users, or professional tradespeople.

Amazon’s AI-powered search assistant Ask Rufus is a great resource to identify these user intent signals. You can reference common questions from Rufus, customer review insights from competing Listings, and the Customer Questions & Answers section on top-performing Listings to build content that directly addresses the exact questions users are asking. This aligns perfectly with Amazon’s shift toward AI-powered search.

A9 still makes up roughly 80% of Amazon’s overall ranking logic, with A10 and COSMO updates contributing the remaining 20% of signals.

Amazon tracks the total revenue it earns from every search position on its results pages. This revenue comes from average 15% referral fees, Fulfillment by Amazon (FBA) fees, and associated ad spend generated by products in each slot. If your product is placed in a higher ranking position but fails to drive consistent conversions, Amazon will move your Listing lower to replace it with a product that generates more revenue for the platform.

Click-Through Rate (CTR), Conversion Rate (CVR) and keyword bid amounts for ad positions are the three core drivers of both organic and Sponsored Ad ranking performance.

For the same Top of Search (TOS) ad slot for a given keyword, a high-performing product may hold the position with a $0.50 bid. A mid-tier product may need a $1.00 bid to hold the same spot. A low-performing product may need to bid $1.50 or higher to stay visible, if it can even appear in that position at all.

None of these optimization tactics will overcome poor product-market fit. Products that meet clear user demand, offer strong value and face limited competition will almost always outperform low-quality, oversaturated products, even with minimal optimization. Strong operational tactics can lift a solid product, but no amount of ad spend or Listing tweaks will turn a bad product into a consistent bestseller.

I’ve already started updating all my Listings to focus more on use case and intent rather than just keyword matching, and I’ve seen a 12% lift in organic traffic across most of my ASINs in the last two months. Have you noticed any changes to your rankings since the COSMO update rolled out? What optimization tweaks have you tested that worked? Drop your experience in the comments below!