AI and A9: The Future of Amazon’s A9 Algorithm and Its Impact on Search Rankings

Note: This blog takes inspiration from a LinkedIn post from Kevin King, an industry expert and influential voice in the ecommerce community. 

2023’s run toward AI adoption has been quicker than Wile E Coyote chasing down Usain Bolt and only getting faster. With the rapid acceptance of ChatGPT– 100 million users in 2 months– almost every big company has followed suit. Google created Bard, Snapchat added AI chat, Meta introduced Code Llama, and new AI tools appear daily.

Among these tech giants, Amazon, the ecommerce behemoth, is adopting AI in every part of its product development process. In this post, we’ll look at how they are using AI for their A9 search algorithm– the search engine that powers Amazon search results. 

Related: AI in Product Development: Leveraging The New Age of Innovation

Amazon Experts Weigh In on the A9 Algorithm

In a recent LinkedIn post from Kevin King, an 8-figure ecommerce seller and AM/PM Podcast host, he talks about how Amazon is at work on their A9 algorithm, using a unique ecommerce LLM (Large Language Model). 

“It’s the A9. It evolves over time like you did during puberty. But its name doesn’t change.” - Kevin King.

Amazon unveiled these insights at the 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining in Osaka, Japan. Kevin explains the current limitations of traditional search methods that rely only on exact word matches in product data and descriptions. This approach can overlook products when someone searches with spelling errors or synonyms. 

Using LLMs, Amazon addresses this with “semantic matching,” where the LLM can understand the meaning behind words and their context and not rely solely on the keywords. Kevin uses the example, “A search for ‘sneakers’ would also show ‘running shoes.” 

The Current State of Amazon’s A9 Search Algorithm

As it stands, here are some of the main factors influencing the A9 algorithm:

• Relevance: First, the A9 algorithm checks the relevance of your keyword compared to the title, description, and backend keywords of the listings. 
• Price: Competitive pricing affects ranks, i.e., if you are priced too high compared to similar products, your listing may rank lower. 
• Customer Satisfaction & Retention: Amazon looks at product reviews, ratings, and overall customer feedback. A high return rate can also negatively affect ranks. 
• Conversion Rate & Sales Velocity: The more frequently customers buy your product after viewing it, the higher it ranks. This is why getting high-quality images, descriptions, and positive reviews is essential. 
• Stock Availability: If your product is out of stock, it will rank lower. Make sure your product has consistent stock levels and avoids stockouts. 

New AI Technology Coming to Amazon’s A9 Algorithm

In his post, Kevin King mentions, "Amazon is working on both physical GPU-based technology and its own unique ecommerce LLM. It can process massive AI-related search analyses to serve up results in the blink of an eye." 

He adds, "This could take away the need to be indexed for everything, eliminate keyword stuffing, and change how we try to optimize listings to rank today."

As Kevin describes, the entire landscape of search engines is changing. The focus in the future will be less on keyword stuffing and more on crafting a sophisticated listing focused on conversion and quality. 

RELATED: AI in Business: How to Find the Right Tools to Make Selling on Amazon Easier

How Amazon Trains Its Ecommerce LLM in Four Steps

Now, you may be wondering how Amazon trains this Large Language Model. Let’s break down the process into four easy-to-understand steps. 

1. Domain-specific pre-training: Talking shop
2. Query-product interaction pre-fine tuning: Matching words
3. Fine tuning for matching: Finding stuff
4. Knowledge distillation to a smaller model: Make it simple  

If you’re scratching your head deciphering the technical jargon, here[s what that means in layperson's terms: 

1. Talking shop: First, Amazon trains the LLM on over 1 billion products and words relevant to online buying.
2. Matching words: Next, the system guesses missing words and context and learns to connect what you ask for with the right thing to buy.
3. Finding stuff: Then, it uses a unique method called ‘fine tuning’ to get good at matching what you ask for with things you can buy.
4. Make it simple: Last, the complex system is transferred to a smaller model suitable for real-time product searches, making it quick and easy to find what you want to buy.

New A9 Algorithm Impact on Sellers and Tools

One of the biggest takeaways from Kevin’s post is the potential impact this change has on sellers' keyword strategies: "A single sentence in a product listing could soon match hundreds of search phrases without keywords explicitly mentioning them all."

This A9 algorithm change is potentially paradigm-shifting in how sellers approach product listing optimization. Tools heavily focused on keyword research may need to adapt to this new world of LLM-assisted search algorithms. 

Related: How to Use Listing Quality Score (LQS) by Seller.Tools to Optimize Your Amazon Listing