MarketingBench Research

Research for marketing in the age of AI.

Independent benchmarks and field studies on how AI performs marketing work, recommends software, and changes how marketing teams operate.

Featured research

The first MarketingBench research note

Research Note 001Reviewed exploratory study

What AI Actually Recommends for Lifecycle Marketing

Leading AI models have a strong spontaneous default—but their preferences change materially after equal exposure to scenario-appropriate platforms.

Read the reportBrowse the evidence

44.2%

Braze in unaided recall

91 of 206 accepted open responses

52.6%

Adobe after equal exposure

60 of 114 eligible controlled runs

The central finding

AI visibility and scenario fit are related, but they are not the same signal.

432

Responses

24

Buyer scenarios

3

Leading models

96.5%

Accepted

Research agenda

Four questions guide the work

01

Model performance

Where leading models produce stronger marketing work—and where quality breaks down.

02

Recommendation behavior

Which products AI recalls, prefers after exposure, and associates with specific buyer situations.

03

Marketing workflows

How teams can use AI across strategy, creative, lifecycle, analytics, and operations.

04

AI adoption

How marketing roles, practices, and buying decisions change as AI becomes part of the stack.

How we publish

Evidence before certainty

MarketingBench research is designed to be inspectable. Every note states what was tested, what was excluded, and how far the findings can reasonably travel.

01

Versioned scenarios and model identities

02

Published denominators and exclusions

03

Separate unaided and controlled results

04

Browsable evidence behind the charts

Open research

See the data behind the conclusion.

Read the methodology, filter all 417 accepted observations, or download the public dataset.

Open the data