Researchers assessing 300 Facebook image ads found that liking, sharing and remembering an image were related to different aspects of the viewer’s response.
Ulf Böckenholt and Kunpeng Zhang examined advertisements posted by three fast-food chains and used human judgments to study how the images were evaluated. Interest was especially relevant to memorability, while both interest and enjoyment were associated with liking and sharing.
The research, available online in the International Journal of Research in Marketing in October 2026, offers a way to think about creative testing before an ad runs. Its sample covers a particular format and industry, so it cannot establish a universal formula for effective advertising.
Five ways to assess an image
The researchers examined typicality, complexity, ease of understanding, enjoyment and interest. Typicality concerns how familiar or conventional an image appears. Complexity describes how much visual information it contains.
Ease of understanding concerns how readily viewers interpret what they see. Enjoyment and interest capture different reactions: an image can be pleasant without holding attention, or intriguing without being especially enjoyable.
The team related these judgments to liking, sharing and a measure of whether viewers could later retrieve or recognize the image. A measurement-error-corrected statistical model was used to examine relationships at the image level.
Interest was particularly associated with the memory measure. Enjoyment and interest were both relevant to liking and sharing. The findings point to more than one way an image can succeed, depending on what the advertiser wants it to do.
What a marketer can test
A campaign designed to introduce a brand might ask whether people can later recognize the image and identify the advertiser. A social campaign might also track whether people choose to share it. A retailer seeking immediate orders needs a separate measure of sales and the cost of generating them.
Those outcomes need not move together. A memorable image may attract attention without explaining the offer, while a well-liked design may be easy to forget. Testing several questions can expose a weak point that a single score misses.
Our earlier coverage of a small AI video advertising study considered a related problem: whether viewers understand the connection between a creative idea and the service being advertised.
Computer scores and their limits
The researchers also used computational measures of image features to approximate the human judgments. In a supplementary check, a synthetic panel based on modeled consumer personas agreed more closely with human assessments of typicality and complexity than with assessments of enjoyment and interest.
That weaker agreement is a reason to retain human feedback when decisions depend on how an audience feels. The work does not demonstrate that simulated respondents can replace customer research or forecast sales from a new campaign.
The 300 ads came from McDonald’s, Wendy’s and Chipotle. Other product categories, audiences and advertising formats may respond differently. Marketers can use the framework to form questions for their own tests, then compare results against the business outcome they actually need.
