---
title: 2025 Optimizely Personalization release notes
slug: personalization/release-notes/2025-optimizely-personalization-release-notes
docTags: 
createdAt: 2026-09-30T15:05:30.384Z
---

Subscribe to updates on this page using the notification and RSS feed options.

## December

- Added improvements to the [AI variation development agent](docId:0aAtuf82Px4hgmN8FL_p-) by letting you describe the change you want Agent Platform to make inside the **Element Change** window, making it easier to access the agent and update your elements faster.

## October

- Released the [AI variation development agent](docId:0aAtuf82Px4hgmN8FL_p-), which helps you modify and update existing website elements, create new ones, and generate and apply enhancement suggestions, while maintaining brand consistency by automatically retrieving page styles. 

## September

- Released the [new version of the Visual Editor](docId\:t7EgfK4gtH2cbr94E1RUS) that uses an overlay instead of an iframe, letting you directly interact with the site itself and adjust the visitor experience for your experiment.

## July

- [Contextual bandits](docId\:HEAGVaimK_ctRVAIDugxA) are now generally available for personalization campaigns. Contextual bandits help you deliver the most personalized variation to each user by dynamically re-allocating traffic based on the primary metric and user attributes.
- Released [ratio metrics](#) for A/B and multivariate tests, which let you select different events for the numerator and the denominator to reflect business-specific key performance indicators, such as revenue per add-to-cart click or feature use per account. See [Create a ratio metric in the metric builder](docId\:li0OJUrrBqSKRSIUKB6cO) for instructions.

## May

- Released [Mark chat](docId\:EJoVYz3iypbVxol5BqppS) for Personalization. Mark automates tasks, surfaces insights, and guides decision-making. 

### Usage and billing update

Effective May 7, 2025, access to Mark features across Content Marketing Platform, Web Experimentation, Feature Experimentation, Personalization, Content Management System (SaaS), Collaboration, and Optimizely Data Platform will transition to a credit-based usage and billing model.

For a full list of Mark features, see [Mark and AI features](docId:51wG_l34ATUNcLespgC-G).

## April

- Added the [Summary page](../build-campaigns/create-a-campaign-in-optimizely-personalization.mdx) to campaigns, which provides a high-level summary of the campaign settings that you can download and share with stakeholders for easier reporting on the campaign you are running.
- Added the ability to [upload images to the visual editor](docId\:li0OJUrrBqSKRSIUKB6cO) from the Optimizely Digital Asset Management (DAM) library. You must have an Optimizely product that has DAM to use it with Optimizely Experimentation.

### Usage and billing update

Effective May 7, 2025, access to Mark features across Content Marketing Platform, Web Experimentation, Feature Experimentation, Personalization, Content Management System (SaaS), Collaboration, and Optimizely Data Platform will transition to a credit-based usage and billing model.

For a full list of Mark features, see [Mark and AI features](docId:51wG_l34ATUNcLespgC-G).

## February

Released AI variation summary. This lets you use[ AI to generate descriptions for your variations](#), summarizing what element or custom code changes were made and providing the variation's purpose. You can better understand how changes affect variation results to decide about your experiments and campaigns.

## January

- Released [experience templates](#) (formerly extensions) for general availability, which lets you streamline campaign and experiment creation and reduce code duplication. You can use the pre-built templates provided by Optimizely to quickly create a campaign or experiment with little to no development or build your own. See [Get started with experience templates](#) for information.
- Added ability to [change the traffic allocation within grouped experiences](../build-campaigns/create-a-campaign-in-optimizely-personalization.mdx) of personalization campaigns.
- Enhanced how Preview mode loads by only querying the relevant pages associated with the experiment. This has led to improvements such as decreasing the load time for some users from 30 seconds to 3 seconds.