Open to senior and staff roles

Case study / CNN

Preview, not recap: AI summaries at CNN

CNN tested short AI summaries on articles throughout the newsroom. I led the web design for this experiment, from deciding where the summary appears on the page to how it opens.

By Christian Eckels / Web product design lead

Role

Web product design lead

Team

Product, content, engineering, machine learning, analytics and app design partners

Timeline

Live test May to June 2026. Launch targeted for Q4 2026

Platforms

Desktop and mobile web, in the Elevate article

Image to come: summary module in the Elevate article, desktop and mobile

FIG. 1 Summary module in the Elevate article, desktop and mobile.

01

The problem

Many readers leave an article before they know what it’s really about. CNN wanted to give them a short, trustworthy preview so they could decide whether to keep reading, without replacing the reporting itself.

02

What didn’t work

An earlier test in 2025 used long recap summaries that were open by default. They gave readers the story without the reporting, and engagement on the page dropped. Research pointed us in a different direction: preview, don’t recap.

03

The principles

01

Preview, don’t recap. Short bullets that tease the story.

02

Closed until a reader asks. Opening it is the signal.

03

Placement is a business decision, not only a design one.

04

Fit the Elevate article, not the other way around.

05

Every expand and collapse works with a keyboard and a screen reader.

04

Three ways in

On the web, we tested three entry points inside the Elevate article: an accordion below the lead image, a button in the article’s action bar, and a floating button. Each option changed how far people scrolled and what they saw next, so choosing one meant balancing reading, revenue, and subscriptions.

Accordion

Action bar

Floating button

FIG. 2 The three tested placements.

05

A person in the loop

Every summary is reviewed by an editor before readers see it. Editors can edit, regenerate, or hide a summary, and they get a warning if the article changes. I designed the reading side to reflect that care, with clear labeling and nothing that looks more certain than it is.

06

Outcome

Leadership approved a launch with the winning placements: the floating button on mobile web and the accordion on desktop. Readers who opened a summary came back more often and spent more time with CNN.

[Metric]

Return visits among readers who opened a summary

[Metric]

Time per visit

Figures to be added once cleared for public use.

07

What I learned

When using AI in the reading experience, restraint is key. A summary earns trust by staying brief, staying closed until requested, and never pretending to be the full story.