Blog

An update on Amazon’s Rufus and what it means for brands

This blog post is a follow-up to our original analysis of the beta launch of Rufus - providing further observations, recommendations for improvement, and strategies for brands learning to live in the world of AI and search.
Written by
Gabe Fishbein
August 19, 2024

Amazon has officially rolled out its AI-powered shopping assistant, Rufus, and it was the talk of the town during Prime Day 2024. Shortly after beta launch in February, we put our initial thoughts here.

Now that we’ve all had a few more months to test and learn with Rufus, we’re here to share how brands should be thinking about the tool in a constantly evolving AI-driven search landscape.

What we know about Rufus

Here’s what Rufus looks like in action:

Here are some of our key observations:

  • Rufus does not show any ads
  • Rufus results are not in the same order as organic rankings
  • Rufus will cite external websites, but won't link to them
  • Rufus looks around the web and will respond to prompts like “What does Reddit think about X?” or “What features are on [brand’s] website that are not listed on the PDP?”
  • Rufus can also recommend a number of products to help solve a particular need; i.e. “What decorations do I need for a race car-themed birthday party?”
  • Product recommendations from Rufus are not served with an add-to-cart option yet, meaning you have to navigate to PDP to add each item to the cart
  • Rufus seems to loosely propose questions based on purchase/search history, but also injects randomness into the suggested prompts that have nothing to do with activity
  • Rufus looks at ratings and reviews when summarizing pros & cons of a product, and its summary very closely matches the existing "AI" summary widget
  • Rufus uses the current PDP as context in order to tailor its answer to a generic prompt that doesn’t mention a product, eg. “Does this product have soy?” but it does not only use the PDP - it pulls information from brand websites as well

Similar to what we saw back in February, we know that Rufus is sourcing information from product descriptions, customer Q&As, and customer reviews. It just depends on the task - whether it’s more broad or more product-specific - to determine exactly where info is being pulled from. 

Progress has been made on Rufus’ capabilities, but there’s still room for improvement. Interestingly, Rufus is able to provide answers to existential questions (and then link to product resources) such as “What is the meaning of life?”. LLMs are still in the experimental phase, and the brightest minds have still not figured out explicit guardrails.

What does this mean for brands?

Every consumer now has the ability to discover and research your products via Rufus, and winning in AI-generated search may turn out to be quite different from traditional search. With everything we know, here our are recommendations for brands looking to stay ahead of the curve:

Optimizing your catalog and product detail pages is an important first step, and you can do this by keeping an eye on search term reports, and regularly optimizing your content based on trends. However, having up-to-date content on Amazon alone isn’t enough. 

Your content strategy should extend to your brand website, as we have found that clarifying questions to Rufus sometimes cite content directly from the manufacturer. GenAI search is more conversational, and the Rufus UX encourages customers to ask more questions, dig deeper, and compare products. Rufus’s results are contextual, meaning that they don’t follow the same order as search results.

Ratings and reviews are also a heavy influence on what Rufus knows. Feedback from customers, both positive and negative, are factors in Rufus’s responses. By integrating this feedback into your product listings, you can showcase key selling features or address questions and concerns upfront to drive more conversions. You can even use Rufus to drive new product innovation by learning about gaps in features and competition.

Given that Large Language Models are expensive to run and even more expensive to train, their knowledge base doesn’t evolve quickly. This means that brands need to “play the long game” with both content and assortment, knowing that it’s likely that changes won't immediately reflect within these tools. Obsessing over customer knowledge gaps and pain points remains the most important focus for brands.

How customers FIND products will always evolve, but content’s role in how customers LEARN about products remains just as important.

FAQs
No items found.
Gabe Fishbein
VP of Product Strategy, Flywheel
Gabe Fishbein is VP of Product Strategy at Flywheel. Since joining Flywheel in 2016, he has been responsible for the creation of innovative, first-to-market solutions to help brands disproportionately win on eCommerce.

Ready to grow your business?

Let’s discuss the best approach to meet your brand’s specific needs.

Let's connect

Related resources

Blog

Building a multiphase ADSP & AMC audience strategy for Black Friday Cyber Monday

Written by
Sarah Carter, Julianne MacDonald, & the AMC Team
October 31, 2024
Blog

Everything’s bigger at Amazon’s unBoxed in Austin

Written by
Kristie Schmuck
October 24, 2024