One of the most common questions I get from potential customers at Ecomtent is "your Rufus optimization looks great, but are Amazon Shoppers actually using it?"

We now consumers are using AI chat bots for product research - A16Z research indicated that 60% of US Adults have used an AI Bot (e.g. ChatGPT or Rufus) for product research in the last 30 days. It's clear that Amazon believes this is the future of shopping, as they are ramping out the rollout by integrating Rufus into the main search bar.

Amazon has not released any data on Rufus usage publicly, except from stating that Rufus had serviced "hundreds of millions of purchases" in beta testing when announcing the full roll out of Rufus onstage at Amazon Accelerate in September '24.

However, whilst they have not given us raw numbers, in October '24, in an AWS blog they did tell us how much compute they had to allocate for Rufus queries during Prime Day: 80,000 AWS Inferentia and AWS Training chips, collectively processing 3 million tokens per minute.

In true Amazonian style, we can therefore work backwards to calculate how many queries Rufus got during this period. Assuming the average query length on Rufus is similar to that on Perplexity, 10-11 words, and each word comprising approximately 1.5 tokens (common in natural language processing), this results in an average of 15.75 tokens per query.

Given Rufus's infrastructure was built to process 3 million tokens per minute, Rufus' handled approximately 190,476 queries per minute. Over a 24-hour period, this translates to a capacity of 274.3 million queries per day.

Amazon receives approximately 2 billion searches per day, thus Rufus accounts for ~13.7% of these in October '24. This percentage is growing rapidly, and we expect could reach 25-35% by the end of 2025.

To stay ahead, Brands clearly need to be thinking about optimizing for AI powered search. This is our mission at Ecomtent, and our case studies include multiple examples of how we have helped sellers slash their Best Seller Rank by half and drive sales through optimizing for RUFUS. You can see a demo below.

‍

‍