Amazon Personalize

What it does

Amazon Personalize takes in your user interaction data, such as clicks and views, along with item and user data, uploaded in batches from Amazon S3 or streamed in real time. It trains custom recommendation models on that data and returns personalized item recommendations, personalized re-rankings of item lists, and customer segmentation. Recommendations can be served in real time or generated in batch. It also pairs with Amazon Bedrock models for content generation, including a Content Generator that creates themes for batch recommendations.

Teams use it to add recommendation carousels to websites and apps, rank streaming or travel content, surface trending retail products, and build recommendation-driven marketing emails. AWS manages the infrastructure for training and serving, and recommendations adapt to how users are behaving at the moment rather than relying only on historical rules. Two newer recipes, User-Personalization-v2 and Personalized-Ranking-v2, use a Transformer-based architecture, alongside older custom recipes.

It suits developers and businesses that already have interaction data and want recommendations without building their own engine. An AWS account is required, and work starts in the AWS Management Console. Costs scale with data ingested, training, and requests. Active real-time campaigns are billed at a minimum of 1 request per second even when idle, which matters for low-traffic uses.

What it can do

• Train custom recommendation models on your own user interaction data
• Serve real-time recommendations that adapt to current user behavior
• Generate batch recommendations for use in email and other campaigns
• Re-rank item lists per user and surface trending or similar items
• Generate themes for batch recommendations with the Content Generator

Works with

• Amazon Bedrock
• Amazon S3

Example workflows

Homepage recommendation carousel

Inputs: item catalog and user click history uploaded to Amazon Personalize. Steps: ingest the data, train a User-Personalization-v2 model, create a campaign. Output: real-time personalized item recommendations for each visitor, returned per request.

Personalized email campaign

Inputs: interaction data and a list of users. Steps: train a custom solution, then run a batch recommendation job for those users. Output: per-user item recommendations from batch inference, which can be used in marketing emails.

Before you choose

• Real-time campaigns are billed at a minimum of 1 request per second even with no traffic, so idle campaigns still cost money
• Training and inference costs vary by recipe, and training hours can exceed clock time
• Quality depends on having sufficient interaction data of your own

Listed under

• Marketing
• Ecommerce
• Email and automation
• Developers
• Machine learning
• Media streaming
• Personalization
• Real-time

Get started

Create an AWS account and open the Personalize console

1. Create or sign in to an AWS account
2. Open Amazon Personalize in the AWS Management Console
3. Upload interaction, item and user data, then train a model and create a campaign or batch job

Pay per use with no upfront commitments. The free allowance lasts two months, so it is not a permanent free tier. Costs include data ingestion, training and recommendation requests, with a 1 TPS minimum on real-time campaigns. See the pricing page for full rate tables and examples.

60-day free trial.

Source: aws.amazon.com/personalize/pricing, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Recommender System – Amazon Personalize – Amazon Web Services ↗aws.amazon.com
• Amazon Personalize Pricing - Amazon Web Services ↗aws.amazon.com

Company website

aws.amazon.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.