Amazon AI Search 2026: Rufus, COSMO & AI Shopping Prompts — What Sellers Need to Know Now

As of March 25, 2026, Amazon's AI Shopping Prompts are billable. At the same time, Rufus processes over 13 % of all Amazon search queries, the COSMO algorithm replaces classic keyword matching with intent recognition, and the new Amazon Ads MCP Server enables AI-driven campaign creation in a single prompt.

1. What Changed in March 2026?

In the first quarter of 2026, Amazon introduced three fundamental changes that are reshaping the interplay between organic search and advertising:

• March 25, 2026: AI Shopping Prompts are officially billable. Every click on an AI-generated product recommendation within Rufus responses is charged as a Sponsored Products click — through existing campaigns, with no opt-out.
• Rufus rollout complete: The AI assistant is now available in all marketplaces and processes over 13 % of all Amazon search queries. Trend: rising.
• COSMO expanding: The Customer Obsession Shaping Model is now active in 18 product categories, replacing classic keyword matching with intent-based results.

On top of that, the Amazon Ads MCP Server — a standardized API interface through which AI agents can create campaigns, set bids, and retrieve reports. This is the technical foundation for fully autonomous campaign management.

2. Rufus: Amazon's AI Shopping Assistant in Detail

Rufus is Amazon's conversational shopping assistant, trained on the entire product catalog, customer reviews, and publicly available web content. Shoppers ask questions like "What's the best coffee for French Press?" — and Rufus delivers a curated answer with product recommendations.

Why Rufus Matters for Sellers

• 13 % search volume: Over one-eighth of all Amazon searches already run through Rufus. These users don't see a classic search results page — they see AI-curated recommendations.
• Content quality decides: Rufus favors listings with naturally written, complete information — especially A+ Content, meaningful bullet points, and backend keywords.
• Reviews as a ranking signal: Rufus analyzes customer reviews semantically. Products with consistently positive reviews on relevant purchase criteria are recommended more often.

What You Should Optimize for Rufus

1. Complete your A+ Content: Rufus extracts information from A+ modules. Empty or minimal A+ pages are referenced less frequently.
2. Optimize bullet points for questions: Frame your bullet points as answers to typical buyer questions in your category.
3. Expand backend keywords: Use synonyms, use cases, and contextual terms — not just exact match keywords.

3. COSMO: From Keyword Matching to Intent Recognition

COSMO stands for Customer Obsession Shaping Model. It's Amazon's next step away from pure keyword matching toward a search that understands why someone is searching — not just what.

How COSMO Works

Classic A9 works simply: A buyer searches "non-slip yoga mat," and Amazon shows products containing those keywords in their listing — weighted by conversion rate and relevance. COSMO goes further:

• Intent recognition: COSMO detects whether someone searching "non-slip yoga mat" is a beginner (→ affordable, thick, with bag) or an advanced yogi (→ thin, natural rubber, premium).
• Contextual signals: Seasonality, past purchase behavior, review trends, and even time of day factor into result ranking.
• Category-specific: COSMO is currently active in 18 categories, including Sports & Outdoors, Kitchen, Electronics, and Beauty. Expansion to more categories is underway.

What This Means for Your PPC Strategy

In COSMO categories, pure keyword campaigns lose effectiveness. An exact match on "non-slip yoga mat" may only reach a portion of relevant buyers, because COSMO clusters search results by intent — and assigns your product only to the matching cluster.

"If you're still relying purely on keyword match in 2026, you're optimizing for an algorithm that's being replaced."

4. AI Shopping Prompts: How the New Ad Format Works

AI Shopping Prompts are Amazon's newest ad format. They appear within Rufus answers and on AI-curated search result pages. Since March 25, 2026, they are billable — charged as CPC through existing Sponsored Products campaigns.

How Billing Works

• Automatic enrollment: All existing Sponsored Products campaigns are automatically enrolled in AI Shopping Prompts. There is currently no opt-out.
• CPC billing: Every click on a prompt recommendation is charged like a regular Sponsored Products click.
• Prompts report: The Ads Console now has a new Prompts tab that separately reports clicks, conversions, and ACoS for AI-generated placements.

Early Insights from Practice

Conversion rates from AI prompt clicks sometimes differ significantly from regular search clicks. In some categories they're higher (buyers with clear purchase intent), in others lower (exploratory searches via Rufus). That's why separate analysis in the Prompts report is essential.

5. Amazon Ads MCP Server: AI Agents Running Campaigns

The Amazon Ads MCP Server (Model Context Protocol) is a standardized API interface through which AI agents — like Claude, GPT, or specialized PPC bots — can interact directly with Amazon's Ads infrastructure.

What the MCP Server Enables

• Campaign creation via prompt: "Create a Sponsored Products campaign for ASIN B0XXXX with €50 daily budget and 20% ACoS target." — One prompt, one campaign.
• Automatic bid adjustments: AI agents can adjust bids in real time based on performance data, inventory levels, and competitive landscape.
• Report queries: Instead of manually navigating the Ads Console, AI agents retrieve structured reports and act on them.

What This Means for the Future

The MCP Server is Amazon's signal that the future of campaign management is automated. Tools that leverage this interface can optimize campaigns at a speed and granularity that's impossible to achieve manually.

6. What You Should Do Now: 7-Point Action Plan

The changes are complex, but the next steps are clear. Here's your 7-point action plan:

1. Analyze the Prompts Report

Open the Ads Console and check the new Prompts tab. Compare CPC, conversion rate, and ACoS from AI prompt clicks with your regular search clicks. Identify under- and overperforming campaigns.

2. Reset Your ACoS Benchmark

Your current ACoS benchmark is based on a world without AI prompt costs. Set a new benchmark that accounts for the additional click costs — and measure separately from March 25 onward.

3. Optimize Listings for Intent

Rework your top 20 listings: titles, bullet points, and A+ Content should address buying intent, not just contain keywords. Think in usage scenarios, not search terms.

4. Expand Backend Keywords

Add contextual terms, synonyms, and usage scenarios to your backend keywords. COSMO and Rufus analyze these fields semantically — not just as keyword matches.

5. Check Review Quality

Rufus analyzes reviews by content. Products whose reviews consistently contain positive statements about relevant purchase criteria are recommended more often. Focus your post-purchase communication on the right aspects.

6. Identify COSMO Categories

Check if your main products fall into one of the 18 COSMO categories. If so: keyword-only campaigns are no longer sufficient. You need an intent-based strategy.

7. Evaluate Predictive Tools

Reactive tools optimize on yesterday's data — but COSMO, Rufus, and AI Prompts are changing the rules in real time. Predictive AI that optimizes 5 days ahead across all 15 dimensions is the logical answer.

Last updated: April 6, 2026