An ad being delivered does not mean it is being seen, and being seen doesn't necessarily mean it's being remembered. Impressions tell you an ad was served, while viewability indicates whether it had an opportunity to be seen. Neither metric alone tells you how much attention the ad received.
This guide explores what attention is, how researchers measure it, how AI can predict visual attention, and what the evidence says about using attention to improve advertising. It also examines the limitations of attention metrics and why they should be evaluated alongside memory and business outcomes.
Read more here: https://www.alpha.one/whitepapers/attention-101
The Attention Gap
Research by Dentsu Aegis Network and attention researcher Karen Nelson-Field examined 17,000 video views from 3,400 panel members across the UK, US and Australia. The study found that ads received direct visual attention for about one-third of the time on average, while approximately 20% received no direct visual attention (Dentsu Aegis Network, 2019).
These figures illustrate an important distinction: delivering an ad does not guarantee that someone looks directly at it. However, direct visual attention is not the only possible route to advertising effectiveness. The study's findings should be understood within its particular research design rather than treated as universal benchmarks for every platform or campaign.
For marketers, the challenge is to understand how attention varies across creative, audiences, platforms and viewing conditions, and whether those differences matter for the outcomes a campaign is intended to achieve.
What Is Attention?
Attention is the process through which people selectively allocate cognitive resources to information in their environment. In visual advertising, researchers can study where people look, how long they look and how their gaze moves between elements.
Cognitive science distinguishes between two interacting influences on attention:
- Bottom-up attention: Stimulus characteristics, such as contrast, movement and visual prominence, can attract attention.
- Top-down attention: A person's goals, expectations, interests and current task influence what they notice.
These processes work together rather than operating as competing explanations (Corbetta & Shulman, 2002; Itti & Koch, 2001).
An ad may attract the eye through its visual design, but whether someone continues looking at it (and what they take away) also depends on their goals and context. Marketers can influence the visual characteristics of an ad and the context in which it appears, but they cannot assume that every viewer will respond in the same way.
How Attention Measurement Works
Advertising measurement includes several different approaches. They are related, but they answer different questions.
- Served: Was the ad delivered? Impressions record delivery, not attention.
- Viewable: Was the ad sufficiently visible on screen to meet a defined viewability standard? Viewability indicates an opportunity to see, not proof that someone looked.
- Measured: Where did people look, and for how long? Eye-tracking studies record gaze behaviour under specified testing conditions.
- Predicted: Where are people likely to look? Predictive models estimate attention patterns based on data and learned relationships.
- Evaluated: How well do attention measures or predictions relate to memory, brand outcomes and business results?
These approaches did not simply replace one another. Each adds a different layer of evidence, and each has limitations.
The growing interest in attention reflects advertisers' need to understand differences in the value of impressions across platforms, formats and creative executions. Eye-tracking research and machine learning have made it increasingly practical to study and predict visual attention at scale, although the reliability of those predictions depends on the data, task and validation method.
How Traditional Attention Testing Works
Traditional visual-attention research often uses eye-tracking equipment, including infrared-based systems, to record participants' gaze as they view advertising. Researchers analyse the resulting fixation data to understand which elements attract attention and how gaze is distributed.
Heatmaps can help visualise these patterns by showing where fixations are concentrated across participants or viewing sessions. However, they should not be interpreted as direct measurements of comprehension, brand recall or purchase intent.
What the Research Says
A landmark study by Pieters and Wedel (2004) analysed 1,363 print ads tested on more than 3,600 consumers. It found that pictorial elements were particularly effective at capturing attention, attention to text increased with its surface size, and brand elements were particularly effective at transferring attention to other elements of an ad.
A separate study investigated advertising originality and familiarity, examining how these characteristics related to brand attention and memory (Pieters et al., 2002). The findings support testing creative distinctiveness, but they do not establish that originality will improve every ad regardless of execution or context.
Eye-tracking also has an important limitation: looking is not the same as remembering. Wedel and Pieters (2000) examined the relationship between eye fixations and brand memory, finding systematic associations between visual attention and memory performance. The relationship depends on what people attend to and how they process the information.
The practical lesson is to measure visual attention as one part of the process, not as a substitute for testing whether people understand or remember the message.
Predicting Attention with AI
AI-based attention prediction uses patterns learned from observed data to estimate where people are likely to look at visual content. Depending on the system, models may analyse features such as contrast, edges, faces, text, objects, layout and the relationship between an ad and its surrounding media context.
Eye-tracking data can provide the human-observation benchmark used to train or evaluate these models. Computer vision and machine learning help identify patterns in that data, while theories of visual attention can inform the choice of features and model design.
Not every predictive system uses the same inputs or methods. Some focus on visual saliency within an image; others explicitly model brand attention or the context in which an ad appears. These differences matter when comparing tools.
Can AI Match Human Eye-Tracking Data?
AI attention prediction aims to approximate patterns observed in human gaze data. The degree of agreement depends on the task, dataset, evaluation metric and variability between human observers.
A score is meaningful only when its methodology is clear. Before accepting a claim that an AI tool matches human attention, ask four questions:
- Reference: Which human eye-tracking dataset was used, and what content did it contain?
- Comparison: Were the model's predictions compared with individual viewers, an aggregate gaze map or another reference?
- Metric: How was agreement calculated, and what does a higher score mean?
- Validation: Was performance tested on independent data that was not used to train or tune the model?
Visual-saliency research uses several evaluation measures, including area under the curve (AUC), normalized scanpath saliency (NSS) and correlation coefficient (CC). Each captures a different aspect of model performance. Research examining common saliency metrics discusses their strengths, limitations and suitability for different evaluation tasks (Bylinskii et al., 2019).
For example, the MIT Saliency Benchmark reports a human-reference AUC of approximately 0.92 for a particular image-saliency benchmark (Bylinskii et al., 2016). This is a benchmark-specific result, not a universal human ceiling for all advertising-attention models.
AUC also has a chance-level baseline of 0.5 in the relevant formulation. A score must therefore be interpreted against its metric's baseline and an appropriate human reference. Scores from different datasets, tasks or implementations should not be compared as if they were interchangeable.
For marketers evaluating predictive attention tools, the key question is not whether a score looks high in isolation. It is whether the model predicts the relevant attention outcome reliably on content similar to the ads being tested.
What Makes an Ad Effective?
Research can inform creative decisions, but it does not provide a universal formula for attention. Four principles offer useful starting points for testing and optimisation.
1. Establish a Clear Focal Point
Make the main visual and message easy to identify. Research on print advertising found that pictorial elements were particularly effective at capturing attention (Pieters & Wedel, 2004).
However, the best focal point depends on the creative objective and the relationship between the visual, message and brand. A clear focal point should help viewers understand the ad rather than merely draw their eyes to the most visually striking element.
2. Make the Brand Identifiable
Attention should have a purpose. If viewers notice an ad but cannot identify the advertiser, the opportunity to build brand associations may be lost.
A 2025 study by VCCP Media and Amplified examined distinctive brand assets in digital video advertising. The research involved more than 20,000 views of 72 ads and explored how brand cues influenced advertising outcomes (Amplified & VCCP Media, 2025). The researchers reported that distinctive brand assets could help brands achieve outcomes during very short periods of attention, while the findings provide a reason to test how clearly brand assets appear in short exposures. They do not guarantee that every ad will produce memory or business results after a fixed number of seconds.
3. Reduce Unnecessary Competition
Remove elements that distract from the intended message, but do not assume that visual simplicity is always better, as research on visual complexity found that different types of complexity can have different effects. Dense perceptual features may hurt brand attention, while more elaborate design can sometimes help attention to the pictorial or the ad as a whole (Pieters et al., 2010). The objective is not to eliminate every visual element. It is to make the creative coherent and easy to process, while retaining the elements that communicate the message and distinguish the brand.
4. Adapt to the Audience and Platform
People encounter ads in different contexts and with different goals, as research on contextual attention prediction highlights why the media environment can matter when estimating where people look. Creative should therefore be tested in the placements and conditions where it is expected to run, rather than evaluated exclusively as a standalone image. A design that attracts attention in one format may not perform the same way in another.
Attention, Memory and Sales
Attention matters because it can create an opportunity for people to process a brand's message. But attention, memory and sales are different outcomes, and the relationship between them is not automatic.
Research by Nelson-Field and Riebe examined the relationship between visual attention and advertising sales impact, using gaze technology and measures of advertising effectiveness. Related work explored how different levels of attention might contribute to sales outcomes (Nelson-Field & Riebe, n.d.; Nelson-Field & Ewens, n.d.). These studies provide evidence that attention can be useful in understanding advertising effectiveness. Their findings should be interpreted within the specific experimental designs and measures used, rather than as proof that every increase in attention causes a predictable increase in sales.
Emotion may also play a role, as ThinkTV commissioned research into emotional responses to television advertising as part of its Benchmark Series. The study reported that ads generating strong emotional responses attracted more attention and had 2.4 times the short-term advertising impact of ads generating weak responses, according to its reported measures (ThinkTV Australia, 2017). They subsequently corrected an earlier version of the sales-impact claim that had described the difference as 30% greater, with the 2.4-times figure being the corrected reported result. The research was industry-commissioned, so its methodology, sample and scope should be considered when interpreting the findings. However, the result should not be treated as a universal estimate of the effect of emotion on sales.
The broader conclusion is that attention and emotion can provide useful signals, but their value depends on the creative, the audience, the media context and the outcome being measured.
How Marketers Can Use Attention
Attention insights can inform four stages of the marketing process.
Plan: Compare channels and formats using attention measures alongside reach, cost, audience quality and campaign objectives. Avoid treating a single attention score as a universal currency across platforms.
Create: Develop creative that makes the message and brand identifiable, then test how viewers distribute their attention across the ad.
Optimise: Use eye-tracking or validated predictive tools to identify potential weaknesses. Test revisions to see whether they improve attention to the intended elements without compromising comprehension or brand recognition.
Report: Evaluate attention alongside brand memory, brand outcomes and business measures. Be clear about which outcomes were directly measured and which were inferred.
What Research Cannot Tell You
Attention science is useful, but it has limits:
Eye-tracking studies often use controlled conditions that differ from real-world feeds, viewing environments and audience behaviour. Findings from print ads may not transfer directly to digital video, and models trained on one type of visual content may not perform equally well on another.
Predictive systems also differ in their definitions of attention. Some estimate visual saliency, others predict gaze duration or brand attention, and still others incorporate contextual information. Their scores are meaningful only in relation to the task and validation method used.
Attention is also just one input to memory and sales, as a highly visible or visually engaging ad can still fail to communicate the intended message, establish a brand association or influence a purchase. Similarly, a viewer may process an ad without sustained direct visual attention.
Finally, some advertising-effectiveness findings cited here come from industry-commissioned research or company-published reports rather than independent, peer-reviewed studies. Such findings can be informative, but their methodology, scope and potential commercial interests should be considered when interpreting the results.
The strongest approach is to use published research to inform hypotheses, validate measurement tools, and test creative in conditions relevant to your own campaign. No single attention score can replace that process.
Conclusion: Make Attention Count
Attention is an important part of advertising effectiveness, but it is not the end goal. An ad can attract the eye without communicating its message, building brand memory or influencing a purchase. The challenge for marketers is to understand not only whether an ad earns attention, but also where that attention goes and what it helps achieve.
Eye-tracking research provides a way to observe how people interact visually with creative, while AI-powered prediction can help marketers evaluate attention patterns more efficiently. Both approaches are most useful when their methods are transparent, their results are validated, and their limitations are understood.
The opportunity is to bring attention into the creative process earlier: use it to identify potential weaknesses, test improvements and learn which elements deserve more focus. Then evaluate those changes against memory, brand and business outcomes rather than assuming that more attention automatically means better performance.
The goal isn't simply to win attention. It is to make attention count.
References
Amplified, & VCCP Media. (2025, May 16). Groundbreaking research from VCCP Media and Amplified on distinctive brand assets. https://www.amplified.co/insight/vccp-research-amplified-distinctive-assets
Bylinskii, Z., Judd, T., Oliva, A., Torralba, A., & Durand, F. (2019). What do different evaluation metrics tell us about saliency models? IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(3), 740–757.
Bylinskii, Z., Recasens, A., Borji, A., Oliva, A., Torralba, A., & Durand, F. (2016). Where should saliency models look next? In Computer Vision – ECCV 2016: 14th European Conference (pp. 809–824). Springer.
Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. Nature Reviews Neuroscience, 3(3), 201–215.
Dentsu Aegis Network. (2019, July 3). Dentsu Aegis Network seeks to change media trading in the Attention Economy. https://www.dentsu.com/cn/en/the-attention-economy
Itti, L., & Koch, C. (2001). Computational modelling of visual attention. Nature Reviews Neuroscience, 2(3), 194–203.
Nelson-Field, K., & Ewens, K. (n.d.). Research on low-attention processing and its relationship to sales. Admap.
Nelson-Field, K., & Riebe, E. (n.d.). Research on attention and sales impact. Admap.
Pieters, R., & Wedel, M. (2004). Attention capture and transfer in advertising: Brand, pictorial, and text-size effects. Journal of Marketing, 68(2), 36–50. https://doi.org/10.1509/jmkg.68.2.36.27794
Pieters, R., Warlop, L., & Wedel, M. (2002). Breaking through the clutter: Benefits of advertisement originality and familiarity for brand attention and memory. Management Science, 48(6), 765–781.
Pieters, R., Wedel, M., & Batra, R. (2010). The stopping power of advertising: Measures and effects of visual complexity. Journal of Marketing, 74(5), 48–60. https://doi.org/10.1509/jmkg.74.5.048
ThinkTV Australia. (2017). The Benchmark Series: Emotions and advertising. https://thinktv.com.au/facts-and-stats/the-benchmark-series-emotions-and-advertising/
Wedel, M., & Pieters, R. (2000). Eye fixations on advertisements and memory for brands: A model and findings. Marketing Science, 19(4), 297–312. https://doi.org/10.1287/mksc.19.4.297.11794
Ye, J., Wedel, M., & Pieters, R. (2026). AdGazer: Improving contextual advertising with theory-informed machine learning. Journal of Marketing, 90(4), 122–144. https://doi.org/10.1177/00222429251396114
