INTERNATIONAL RESEARCH / PEER-REVIEWED ARTICLE

Do personalized AI video ads increase engagement?

A real WhatsApp advertising experiment compared AI video, personalized images and generic video. We explain what the six-to-nine-point engagement difference does and does not show.

ACADEMIC WORK REVIEWED

Frontiers: Generative AI and Personalized Video Advertisements

ResearchersAnuj Kapoor and Madhav Kumar.

Marketing Science · 8 May 2025 · 44(4):733–747 · Peer-reviewed field experiment · Source language: English

Open the original publication Find on Google Scholar

ENGLISH REVIEW AND COMMENTARY: DIJITALPI

We explain the paper in plain English, discuss what it may mean for organisations and identify our own examples separately.

Read the explanation Go to 3 business examples

English review: 10 September 2026 · The source date appears in the citation above.

Published review

Source and editorial checks are complete. DijitalPi did not conduct a customer experiment for this paper.

LET'S READ THE RESEARCH TOGETHER

We first explain the researchers' question, method and findings. We then discuss how to interpret the results, clearly separating DijitalPi's commentary from the source.

01 / WHAT DID THE RESEARCHERS WANT TO UNDERSTAND?

Do personalized AI video ads create more engagement than images or generic video?

Kapoor and Kumar tested AI-generated video personalized with purchase history in a real mobile-advertising environment.

Explaining the context · DijitalPi commentary

Video format and personalization are different changes. The study used two baselines, comparing AI video with a personalized static image and a generic video.

The measured outcome was engagement. A response to a message is not the same as a purchase or profit.

02 / HOW WAS THE RESEARCH CONDUCTED?

How did they test the question?

Customers of a direct-to-consumer sustainable-products brand were randomly assigned on WhatsApp to personalized AI video, personalized image or generic video advertising.

Understanding the method · DijitalPi commentary

A field experiment captures behavior in a real channel and brand. One brand and one channel still limit generalization to other countries, categories and offers.

The published paper provides approximate cost and productivity calculations. They are not measured net-profit outcomes.

03 / RESEARCH FINDINGS

AI video produced six to nine percentage points more engagement than the two baselines.

Personalized AI video ads generated six to nine percentage points more engagement than personalized images and generic videos. The broad pattern remained across gender and location groups.

Engagement difference: +6–9 pointsPercentage-point difference from the two baselines; not a sales increase.
Comparison: 3 conditionsAI video, personalized image and generic video.

Source: Marketing Science publication, abstract and DOI ↗

04 / DIJITALPI'S EXPLANATION

How should we interpret these findings?

The combination of personalization and moving video may create value. A local test needs separate comparisons to identify which part contributes.

Using purchase history is a data-governance question as well as a performance choice. Better engagement does not permit unlimited data use.

Do not confuse percentage points with relative percentages in campaign reporting.

05 / CONCLUSION AND OPEN QUESTIONS

What did we learn, and what do we still not know?

Personalized AI video produced more engagement than two meaningful baselines in this field experiment. One brand, one channel and an engagement measure prevent treating the result as a general sales rule.

Would the same difference appear in Turkish messages, other price points and purchases? That requires a local control group and end-to-end conversion measurement.

Return to the researchers' original publication

Marketing Science publication, abstract and DOI

DIJITALPI'S APPLICATION COMMENTARY

What could this look like in your organisation?

We created these scenarios to make the topic concrete. They are not cases from the paper or measured client results.

EXAMPLE 01

Clicks rose, orders are unknown

The AI video generated more message engagement, but purchase data was not connected to the campaign report.

Would you report this as a sales success?

Open the recommendation for example 01

Report engagement by its proper name.

Track opens, clicks, product-page visits and orders as separate steps. Do not describe an increase at one stage as proof that the next stage occurred.

EXAMPLE 02

Three things changed at once

The video format, personalization rule and discount all changed in the same campaign.

Can you attribute the difference to AI video?

Open the recommendation for example 02

Test one primary variable at a time.

Keep the offer, audience and delivery time stable. Compare video with image, or personalized with generic, in separate tests.

EXAMPLE 03

Purchase history drives the message

A customer receives video based on a previous category purchase, but the campaign does not clearly explain how the data is used.

Does performance remove data responsibility?

Open the recommendation for example 03

Record the data source and its limits.

Define the fields used, legal basis, retention period and opt-out path before launch. Do not send unnecessary personal data to the model.

TRY IT WITH YOUR TEAM

Connect a video test to the sales funnel.

Create three versions for the same product. Choose the primary measure in advance, then report every step from engagement to sales separately.

Explore our social media advertising service

Source and ownership

Anuj Kapoor and Madhav Kumar.
University of Missouri, Robert J. Trulaske Sr. College of Business; MIT Sloan School of Management.

Marketing Science · 8 May 2025 · 44(4):733–747 · Peer-reviewed field experiment. Open the original publication · Find on Google Scholar

The academic work belongs to the researchers named above. This page contains DijitalPi's explanatory review and original business examples; it is not a full translation of the paper.

← Return to all research articles