How AI search is changing Amazon product discovery (and why your agency has no plan)

Amazon product discovery is no longer driven by keyword strings. It's driven by AI. The Rufus shopping assistant now fields 274 million daily queries (per Seller Labs), and shoppers who use it are 60% more likely to finish a purchase. If your catalog isn't structured for AI discovery, you're invisible to a growing share of buyers.

Most agencies have no plan for this. They're still tuning for the old search bar while the channel underneath it is being rebuilt. This piece breaks down what's happening, why it matters for your margins, and what to do about it this week.

What is Rufus and why sellers should care

Rufus is Amazon's generative AI shopping assistant. It launched in February 2024, trained on the entire catalog, customer reviews, community Q&As, and information from across the web. Shoppers ask questions like "what's the best protein powder for beginners" or "compare these two strollers," and Rufus generates conversational answers with recommendations baked in.

Scale is already massive. 250 million customers used Rufus in 2025, monthly active users climbed 140% year-over-year, and total interactions grew 210%. Amazon projects the feature will generate $10 billion in annual incremental revenue.

This isn't a side feature. It's becoming the primary path shoppers take to find and compare items.

The economics are simple: Rufus users convert at 60% higher rates. If the assistant doesn't surface your item, someone else gets that sale. And unlike traditional ranked results where you can brute-force placement with ad spend, Rufus picks winners based on content quality, review sentiment, and how well a detail page answers the shopper's actual question.

How Rufus actually decides which items to recommend

Rufus doesn't work like the keyword-based A9/A10 algorithm. It runs on Amazon Bedrock, starting with a custom in-house LLM and expanding to frontier models with real-time model routing. The system picks the best model for each query, balancing cost, latency, and accuracy.

Here's what Rufus evaluates when selecting which items to surface:

• Intent parsing: what is the shopper really asking? Rufus reads natural language, not keyword strings.
• Detail page quality: titles, bullet points, A+ content, and how well they answer likely questions.
• Review tone: does the review content confirm or contradict what the page claims?
• Visual signals: Amazon now indexes text in images and video through visual recognition.
• Q&A depth: items with thorough, helpful Q&A sections get surfaced more often.
• Behavioral patterns: click-through rates, conversion rates, return rates, session duration.

The catch? Rufus has a bias. 83% of its recommendations are Amazon-sold items. Amazon Basics appears in 41% of results despite often being lower quality. Amazon-branded goods are recommended six times more often than market share would justify.

AI shopping assistants match the actual "best product" only 32% of the time. The rest are profit-driven suggestions. Third-party sellers need to work harder to appear, and detail page quality matters more than ever.

Why most agencies have no AI strategy

Here's the uncomfortable truth: the vast majority of Amazon agencies have zero strategy for AI-powered discovery. Not a weak one. None.

We audited the top agencies in the space. Trivium Group has no mention of Rufus, AEO, or AI-based discovery anywhere on their site. Their focus is purely PPC, DSP, and creative. Several other mid-market firms are in the same position.

Why? Three reasons:

1. Amazon provides no Rufus reporting. Sellers have no visibility into how the assistant recommends goods, no Rufus-specific analytics, and no tools from Amazon to improve AI placement. Most agencies won't build for something they can't measure in Seller Central.

2. The playbook doesn't exist yet. Traditional Amazon SEO has been keyword indexing, rank tracking, and PPC bid management. AI-powered discovery requires a different skillset: understanding how LLMs parse content, how conversational queries differ from keyword strings, and how review tone feeds into recommendations.

3. Agencies are reactive by default. They wait for Amazon to release features, then adopt them. The firms that build ahead of the curve are rare.

Seller Labs and a handful of others have started publishing on the topic. Canopy Management has blog content on AI-based discovery. But publishing about it and actually doing it for clients are different things.

At ALFI, we built a Rufus Checker tool that scores pages across seven AI readiness layers because we got tired of waiting for Amazon to provide visibility. (If you're evaluating agencies, our breakdown of the best Amazon PPC agencies covers who actually has a modern approach.) If your agency doesn't have something similar, ask them what their AI discovery plan is. The answer will tell you a lot.

What is AEO for Amazon

AEO is the practice of structuring your content so AI systems can parse it, extract it, and surface it in conversational answers.

On Google, AEO means building for AI Overviews and featured snippets. On Amazon, it means building for Rufus.

The core difference between traditional Amazon SEO and AEO:

• SEO targets keyword indexing. AEO targets question answering.
• SEO cares about rank position. AEO cares about whether Rufus mentions your item in a conversation.
• SEO is built around the search bar. AEO is built around the chat interface.

In practice, AEO for Amazon means:

• Writing bullet points that directly answer common shopper questions, not just listing features.
• Structuring A+ content around use cases and comparisons, not just brand storytelling.
• Making sure review tone matches page claims. If you claim "long battery life" and reviews consistently say the battery dies fast, Rufus will notice.
• Completing your Q&A section with helpful, specific answers.
• Including readable text in infographic images that reinforces your claims, because Amazon indexes image text now.

AEO isn't a replacement for traditional SEO. It's an additional layer. You still need keyword indexing, competitive pricing, and strong conversion rates. But without AEO, you're building for yesterday's discovery channel.

How to audit your pages for AI readiness

You can run a basic AI readiness audit yourself. Here's what to check:

1. Ask Rufus about your category. Open the Amazon app, tap the Rufus icon, and type questions your customers would ask. "What's the best [your category] for [common use case]?" See if your item shows up. If it doesn't, that's your baseline.

2. Review your title structure. Does it read like a natural answer, or is it stuffed with keywords? Rufus favors titles in the 65-80 character range that communicate what the item is and who it's for.

3. Check bullet points for question-answer format. Each bullet should address a specific question a shopper might ask. "How long does the battery last?" should be answerable from your bullets without guessing.

4. Audit review tone. Read your top 20 reviews. Do they confirm or contradict what your page claims? A mismatch is a negative signal for AI-based recommendations.

5. Check Q&A depth. Items with 15+ answered questions and specific, helpful responses get more AI visibility.

6. Verify image text. Your infographic images should contain readable claims that match your copy. Rufus processes these.

For a faster, deeper audit, run your ASINs through ALFI's Rufus Checker. It scores across all seven layers and tells you exactly where gaps exist.

What brands building for Rufus are already seeing

Early movers are getting results. An Adobe survey cited by Seller Labs found that 53% of consumers now use AI tools during their shopping journey. Ecommerce traffic from AI assistants doubled every two months since September 2024, with a 1,300% year-over-year increase in November-December and a 1,950% surge on Cyber Monday.

Brands paying attention are seeing visibility improvements in Rufus responses within 30-45 days of restructuring pages for conversational queries. The exact uplift varies by category and competitive density, but the directional signal is consistent: pages structured for AI get surfaced more.

There's a compounding effect. Rufus learns from behavioral data. If your well-structured page gets surfaced, earns clicks, converts well, and avoids returns, the system feeds that performance data back into its model. Early work creates a feedback loop that grows harder for competitors to break into over time.

Sponsored ads are also entering the picture. Amazon has started embedding sponsored placements inside Rufus conversations, shifting from keyword-driven advertising to intent-driven advertising. Your PPC strategy will eventually need to account for AI placement, not just traditional rank.

The financial stakes are real. Amazon projects $700 million in Rufus operating profit for 2025, growing to $1.2 billion by 2027 including ad revenue from Rufus responses. Amazon is investing heavily in making Rufus the default experience. Sellers who ignore this are betting against the platform's own roadmap.

The future: when AI becomes the default discovery channel

Per Seller Labs, Rufus handled an estimated 13.7% of total Amazon queries by October 2024, projected to reach 35% by end of 2025. That trajectory suggests AI will be the majority discovery channel within two to three years.

58% of consumers already say AI tools are replacing traditional search for product discovery. The shift isn't coming. It's here.

Rufus is also gaining agentic capabilities. It can now set price alerts, highlight better-value alternatives, and is moving toward auto-purchase when items drop below set thresholds. When an AI agent can buy on a shopper's behalf, the page that gets surfaced becomes the page that gets the sale, with no human browsing step in between.

Accuracy problems persist. Rufus hallucinates specs and invents prices, and ChatGPT Shopping hallucinates prices in 28% of queries. These systems are immature. But that's exactly why acting now matters: the brands that feed AI systems accurate, well-structured data will be the ones these systems learn to trust as they improve.

The window is open right now. It won't stay open. Once every brand and agency catches up, the advantage disappears. Move before it becomes table stakes.

What to do this week

1. Open the Amazon app and ask Rufus questions your customers would ask. Note whether your items appear in the responses.
2. Run your top 5 ASINs through ALFI's Rufus Checker to identify specific gaps in AI readiness.
3. Rewrite your hero item's bullet points in question-answer format, addressing the most common shopper concerns.
4. Read your top 20 reviews for each hero ASIN and compare them against your page claims. Fix any mismatch.
5. Add or improve your Q&A section with at least 10 specific, helpful answers per item.
6. Ask your agency what their AI discovery strategy is. If the answer is vague or nonexistent, talk to us.

The shift from keyword search to AI-driven discovery is the biggest change to Amazon since sponsored ads. The agencies and brands that adapt first will own the advantage. The ones that wait will pay to catch up.

If you want a clear picture of where your pages stand, start with the Rufus Checker and go from there.