[{"data":1,"prerenderedAt":1855},["ShallowReactive",2],{"insights":3},[4,407,633,1022,1280,1560],{"_path":5,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":9,"description":10,"date":11,"author":12,"readTime":13,"category":14,"body":15,"_type":401,"_id":402,"_source":403,"_file":404,"_stem":405,"_extension":406},"\u002Fresearch\u002Finsights\u002Fhow-webcam-eye-tracking-works","insights",false,"","How webcam eye tracking works — and how we validated it","A look at the science behind browser-based eye tracking, how it compares to laboratory hardware, and what makes it viable for commercial research.","2026-04-20","Caspar Goeke","9 min","Methodology",{"type":16,"children":17,"toc":391},"root",[18,26,33,38,84,89,101,112,117,151,163,169,174,179,202,207,213,218,251,263,269,274,279,300,306,370],{"type":19,"tag":20,"props":21,"children":22},"element","p",{},[23],{"type":24,"value":25},"text","Eye tracking has traditionally required expensive hardware — dedicated infrared cameras, chin rests, and controlled lab environments. Webcam-based eye tracking changes that equation entirely. The interesting question is no longer whether it works, but how close it gets to the laboratory standard, and what kinds of research questions it can now reliably answer.",{"type":19,"tag":27,"props":28,"children":30},"h2",{"id":29},"how-it-works",[31],{"type":24,"value":32},"How it works",{"type":19,"tag":20,"props":34,"children":35},{},[36],{"type":24,"value":37},"Modern webcam eye tracking uses the front-facing camera on a participant's laptop or desktop to estimate where they are looking on the screen. The pipeline has four steps:",{"type":19,"tag":39,"props":40,"children":41},"ol",{},[42,54,64,74],{"type":19,"tag":43,"props":44,"children":45},"li",{},[46,52],{"type":19,"tag":47,"props":48,"children":49},"strong",{},[50],{"type":24,"value":51},"Face detection",{"type":24,"value":53}," — locate the face and eyes in the camera feed.",{"type":19,"tag":43,"props":55,"children":56},{},[57,62],{"type":19,"tag":47,"props":58,"children":59},{},[60],{"type":24,"value":61},"Feature extraction",{"type":24,"value":63}," — identify key landmarks (pupil position, eye corners, head pose).",{"type":19,"tag":43,"props":65,"children":66},{},[67,72],{"type":19,"tag":47,"props":68,"children":69},{},[70],{"type":24,"value":71},"Gaze estimation",{"type":24,"value":73}," — map eye features to screen coordinates using a calibrated model.",{"type":19,"tag":43,"props":75,"children":76},{},[77,82],{"type":19,"tag":47,"props":78,"children":79},{},[80],{"type":24,"value":81},"Calibration",{"type":24,"value":83}," — each participant completes a brief calibration sequence to account for their individual setup, camera position, and viewing distance.",{"type":19,"tag":20,"props":85,"children":86},{},[87],{"type":24,"value":88},"There is no hardware to ship, no software to install. The participant opens a link in their normal browser; the calibration takes about a minute; the study begins.",{"type":19,"tag":27,"props":90,"children":92},{"id":91},"validated-against-eyelink-1000-in-behavior-research-methods",[93,95],{"type":24,"value":94},"Validated against EyeLink 1000 in ",{"type":19,"tag":96,"props":97,"children":98},"em",{},[99],{"type":24,"value":100},"Behavior Research Methods",{"type":19,"tag":20,"props":102,"children":103},{},[104,106,110],{"type":24,"value":105},"The harder claim — that this is good enough to draw conclusions from — is supported by a 2023 peer-reviewed validation study. Kaduk, Goeke, Finger, and König published in ",{"type":19,"tag":96,"props":107,"children":108},{},[109],{"type":24,"value":100},{"type":24,"value":111}," a head-to-head comparison of the Labvanced webcam eye-tracking system (the same system that powers Pagegazer studies) against the EyeLink 1000, the most widely-used research-grade infrared eye tracker. The two systems recorded simultaneously while participants completed five different tasks — a Large Grid calibration sweep, smooth-pursuit eye movements, viewing of natural images, and two head-movement conditions.",{"type":19,"tag":20,"props":113,"children":114},{},[115],{"type":24,"value":116},"The headline numbers from that paper:",{"type":19,"tag":118,"props":119,"children":120},"ul",{},[121,131,141],{"type":19,"tag":43,"props":122,"children":123},{},[124,129],{"type":19,"tag":47,"props":125,"children":126},{},[127],{"type":24,"value":128},"Accuracy: 1.4° (overall), 1.3° at central targets.",{"type":24,"value":130}," The EyeLink achieved 0.91° on the same tasks. The webcam error is roughly 0.5° larger than the laboratory system — and represents about a 300% improvement over earlier webcam solutions.",{"type":19,"tag":43,"props":132,"children":133},{},[134,139],{"type":19,"tag":47,"props":135,"children":136},{},[137],{"type":24,"value":138},"Precision: 1.1°.",{"type":24,"value":140}," Approximately 0.5° looser than EyeLink.",{"type":19,"tag":43,"props":142,"children":143},{},[144,149],{"type":19,"tag":47,"props":145,"children":146},{},[147],{"type":24,"value":148},"Correlation between systems on raw gaze samples: 90% on the Large Grid task, 80% on Free View and Smooth Pursuit.",{"type":24,"value":150}," In other words, on a moment-by-moment basis, the two systems agree about the direction of gaze a large majority of the time.",{"type":19,"tag":20,"props":152,"children":153},{},[154,156,161],{"type":24,"value":155},"The authors conclude that webcam eye tracking now performs ",{"type":19,"tag":96,"props":157,"children":158},{},[159],{"type":24,"value":160},"roughly on par with mobile eye-tracking devices",{"type":24,"value":162}," — not as good as a desktop EyeLink, but in the same regime, and good enough for the questions most non-laboratory research is trying to answer.",{"type":19,"tag":27,"props":164,"children":166},{"id":165},"what-14-means-in-practice",[167],{"type":24,"value":168},"What 1.4° means in practice",{"type":19,"tag":20,"props":170,"children":171},{},[172],{"type":24,"value":173},"At a typical viewing distance of 60 cm from a 24-inch monitor, 1.4° of visual angle corresponds to roughly 1.5 cm on the screen.",{"type":19,"tag":20,"props":175,"children":176},{},[177],{"type":24,"value":178},"That is precise enough to:",{"type":19,"tag":118,"props":180,"children":181},{},[182,187,192,197],{"type":19,"tag":43,"props":183,"children":184},{},[185],{"type":24,"value":186},"distinguish between adjacent paragraphs of text;",{"type":19,"tag":43,"props":188,"children":189},{},[190],{"type":24,"value":191},"identify which product in a row is being fixated;",{"type":19,"tag":43,"props":193,"children":194},{},[195],{"type":24,"value":196},"determine whether a specific button, headline, or call-to-action is being noticed;",{"type":19,"tag":43,"props":198,"children":199},{},[200],{"type":24,"value":201},"measure dwell time on a packaging element, a video frame region, or a UI control.",{"type":19,"tag":20,"props":203,"children":204},{},[205],{"type":24,"value":206},"It is not precise enough to distinguish individual letters within a word, or to do certain kinds of micro-saccade research that an infrared system supports. But it covers the great majority of commercial research questions, and the recent peer-reviewed literature confirms this. Serrano-Carot, Angele, Xu, and Vasilev (2025) showed that webcam eye tracking can reproduce the established laboratory effects of word frequency and skipping cost during reading — the most demanding visual task on a screen.",{"type":19,"tag":27,"props":208,"children":210},{"id":209},"when-it-works-less-well-and-what-we-do-about-it",[211],{"type":24,"value":212},"When it works less well — and what we do about it",{"type":19,"tag":20,"props":214,"children":215},{},[216],{"type":24,"value":217},"Webcam eye tracking is more sensitive to environment than the laboratory equivalent. It works less well in low-light conditions, with significant head movement, or with certain eyeglass configurations. We address this in three ways:",{"type":19,"tag":118,"props":219,"children":220},{},[221,231,241],{"type":19,"tag":43,"props":222,"children":223},{},[224,229],{"type":19,"tag":47,"props":225,"children":226},{},[227],{"type":24,"value":228},"Calibration validation.",{"type":24,"value":230}," Every session ends the calibration sequence with a verification step. If gaze error exceeds threshold, the participant re-calibrates or is excluded.",{"type":19,"tag":43,"props":232,"children":233},{},[234,239],{"type":19,"tag":47,"props":235,"children":236},{},[237],{"type":24,"value":238},"Environmental checks.",{"type":24,"value":240}," Lighting and webcam quality are screened before data collection begins.",{"type":19,"tag":43,"props":242,"children":243},{},[244,249],{"type":19,"tag":47,"props":245,"children":246},{},[247],{"type":24,"value":248},"Quality-driven exclusion.",{"type":24,"value":250}," Sessions where data quality is insufficient are excluded from analysis. Typical exclusion rates are 5–10%, which is comparable to attrition in traditional online research.",{"type":19,"tag":20,"props":252,"children":253},{},[254,256,261],{"type":24,"value":255},"Patterson, Nicklin, and Vitta (2025) consolidated these and other recommendations into a methodological scoping review in ",{"type":19,"tag":96,"props":257,"children":258},{},[259],{"type":24,"value":260},"Research Methods in Applied Linguistics",{"type":24,"value":262}," — now a reference for running webcam eye-tracking studies at scale.",{"type":19,"tag":27,"props":264,"children":266},{"id":265},"why-this-matters-for-commercial-research",[267],{"type":24,"value":268},"Why this matters for commercial research",{"type":19,"tag":20,"props":270,"children":271},{},[272],{"type":24,"value":273},"The practical implication is large. Instead of bringing 20 participants to a lab, you can test 200 participants in their own environments — on the actual stimuli they will encounter in the real world. Sample sizes are bigger, turnaround is faster, the setting is closer to ecologically valid, and the methodology has now been published and replicated independently across reading research, attention research, packaging research, and clinical decision-making research.",{"type":19,"tag":20,"props":275,"children":276},{},[277],{"type":24,"value":278},"The hardware-eye-tracker era did not end. It became one tool among several. For the questions Pagegazer answers — where does attention go, what is noticed and what is missed, where does engagement drop — the browser is now the right instrument.",{"type":19,"tag":280,"props":281,"children":282},"blockquote",{},[283,292],{"type":19,"tag":20,"props":284,"children":285},{},[286],{"type":19,"tag":287,"props":288,"children":291},"img",{"alt":289,"src":290},"Placeholder — figure to add","\u002Fimages\u002Fplaceholders\u002Fhero-heatmap.svg",[],{"type":19,"tag":20,"props":293,"children":294},{},[295],{"type":19,"tag":96,"props":296,"children":297},{},[298],{"type":24,"value":299},"A figure from Kaduk et al. (2023) — for example, the comparison of accuracy across the screen between the webcam system and EyeLink — would fit well here. The paper is published under CC-BY 4.0 and figures can be reused with proper attribution.",{"type":19,"tag":27,"props":301,"children":303},{"id":302},"citations",[304],{"type":24,"value":305},"Citations",{"type":19,"tag":118,"props":307,"children":308},{},[309,332,351],{"type":19,"tag":43,"props":310,"children":311},{},[312,314,319,321,330],{"type":24,"value":313},"Kaduk, T., Goeke, C., Finger, H., & König, P. (2023). ",{"type":19,"tag":96,"props":315,"children":316},{},[317],{"type":24,"value":318},"Webcam eye tracking close to laboratory standards: Comparing a new webcam-based system and the EyeLink 1000.",{"type":24,"value":320}," Behavior Research Methods, 56(5), 5002–5022. ",{"type":19,"tag":322,"props":323,"children":327},"a",{"href":324,"rel":325},"https:\u002F\u002Fdoi.org\u002F10.3758\u002Fs13428-023-02237-8",[326],"nofollow",[328],{"type":24,"value":329},"doi.org\u002F10.3758\u002Fs13428-023-02237-8",{"type":24,"value":331}," (CC-BY 4.0)",{"type":19,"tag":43,"props":333,"children":334},{},[335,337,342,344],{"type":24,"value":336},"Serrano-Carot, M., Angele, B., Xu, H., & Vasilev, M. R. (2025). ",{"type":19,"tag":96,"props":338,"children":339},{},[340],{"type":24,"value":341},"Webcams Can Be Used to Study Eye Movements during Reading.",{"type":24,"value":343}," PsyArXiv. ",{"type":19,"tag":322,"props":345,"children":348},{"href":346,"rel":347},"https:\u002F\u002Fdoi.org\u002F10.31234\u002Fosf.io\u002Fbzt2h_v1",[326],[349],{"type":24,"value":350},"doi.org\u002F10.31234\u002Fosf.io\u002Fbzt2h_v1",{"type":19,"tag":43,"props":352,"children":353},{},[354,356,361,363],{"type":24,"value":355},"Patterson, A. S., Nicklin, C., & Vitta, J. P. (2025). ",{"type":19,"tag":96,"props":357,"children":358},{},[359],{"type":24,"value":360},"Methodological recommendations for webcam-based eye tracking: A scoping review.",{"type":24,"value":362}," Research Methods in Applied Linguistics. ",{"type":19,"tag":322,"props":364,"children":367},{"href":365,"rel":366},"https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.rmal.2025.100244",[326],[368],{"type":24,"value":369},"doi.org\u002F10.1016\u002Fj.rmal.2025.100244",{"type":19,"tag":20,"props":371,"children":372},{},[373,375,381,383,389],{"type":24,"value":374},"For an overview of webcam eye tracking and how Pagegazer applies it as a service, see ",{"type":19,"tag":322,"props":376,"children":378},{"href":377},"\u002Fwebcam-eye-tracking",[379],{"type":24,"value":380},"webcam eye tracking for consumer research",{"type":24,"value":382},". For a fuller list of peer-reviewed work using the same measurement platform, see ",{"type":19,"tag":322,"props":384,"children":386},{"href":385},"\u002Fresearch\u002Fpublished",[387],{"type":24,"value":388},"published research",{"type":24,"value":390},".",{"title":8,"searchDepth":392,"depth":392,"links":393},2,[394,395,397,398,399,400],{"id":29,"depth":392,"text":32},{"id":91,"depth":392,"text":396},"Validated against EyeLink 1000 in Behavior Research Methods",{"id":165,"depth":392,"text":168},{"id":209,"depth":392,"text":212},{"id":265,"depth":392,"text":268},{"id":302,"depth":392,"text":305},"markdown","content:research:insights:how-webcam-eye-tracking-works.md","content","research\u002Finsights\u002Fhow-webcam-eye-tracking-works.md","research\u002Finsights\u002Fhow-webcam-eye-tracking-works","md",{"_path":408,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":409,"description":410,"date":411,"author":412,"readTime":413,"category":414,"area":415,"body":416,"_type":401,"_id":630,"_source":403,"_file":631,"_stem":632,"_extension":406},"\u002Fresearch\u002Finsights\u002Feye-tracking-the-online-reading-experience","Methodology spotlight: webcam eye tracking is good enough to study reading","Recent peer-reviewed work shows webcam-based eye tracking can resolve fine-grained reading behaviour. The implication for digital experience research is direct.","2026-04-15","Pagegazer team","7 min","Methodology spotlight","Digital Experience",{"type":16,"children":417,"toc":623},[418,423,428,434,459,465,477,482,488,500,506,511,516,539,544,548,613],{"type":19,"tag":20,"props":419,"children":420},{},[421],{"type":24,"value":422},"Reading is the hardest visual task to measure with eye tracking. Letters are small, fixations are short (around 250 milliseconds), and the eye moves in fast, irregular jumps (saccades) that need precise timing to capture. If a methodology can resolve reading behaviour, it can resolve almost anything else on a screen.",{"type":19,"tag":20,"props":424,"children":425},{},[426],{"type":24,"value":427},"Three recent peer-reviewed papers — using the same platform that powers Pagegazer — establish where webcam eye tracking now sits relative to laboratory hardware, and what kind of digital experience research it can support.",{"type":19,"tag":27,"props":429,"children":431},{"id":430},"the-validation-against-eyelink-1000",[432],{"type":24,"value":433},"The validation against EyeLink 1000",{"type":19,"tag":20,"props":435,"children":436},{},[437,439,443,445,450,452,457],{"type":24,"value":438},"The foundation is Kaduk, Goeke, Finger, and König (2023), published in ",{"type":19,"tag":96,"props":440,"children":441},{},[442],{"type":24,"value":100},{"type":24,"value":444},". They ran the Labvanced webcam eye tracker and the EyeLink 1000 simultaneously across five tasks — large-grid calibration, smooth-pursuit eye movements, viewing of natural images, and two head-movement conditions. The webcam system reached ",{"type":19,"tag":47,"props":446,"children":447},{},[448],{"type":24,"value":449},"1.4° spatial accuracy",{"type":24,"value":451}," (1.3° at central targets), with the EyeLink at 0.91°. Raw gaze samples between the two systems agreed about ",{"type":19,"tag":47,"props":453,"children":454},{},[455],{"type":24,"value":456},"90% of the time on the Large Grid task and 80% on Free View and Smooth Pursuit",{"type":24,"value":458},". The webcam was approximately 0.5° looser overall — a meaningful gap for some research, but small in the context of attention measurement on a typical screen layout. The authors describe the result as a roughly 300% improvement over earlier webcam solutions and \"roughly on par with mobile eye-tracking devices.\"",{"type":19,"tag":27,"props":460,"children":462},{"id":461},"reproducing-the-established-laboratory-findings-on-reading",[463],{"type":24,"value":464},"Reproducing the established laboratory findings on reading",{"type":19,"tag":20,"props":466,"children":467},{},[468,470,475],{"type":24,"value":469},"Serrano-Carot, Angele, Xu, and Vasilev (2025) tested whether a webcam can recover the ",{"type":19,"tag":96,"props":471,"children":472},{},[473],{"type":24,"value":474},"signature effects",{"type":24,"value":476}," the reading-research community has spent decades developing — fixation duration as a function of word frequency, the cost of skipping a word, and predictability effects on gaze duration. The webcam data reproduced these effects in the right direction and with the right structure. The reading community has long been the strictest audience for eye-tracking accuracy claims; the paper concludes that webcams now meet the bar for reading research.",{"type":19,"tag":20,"props":478,"children":479},{},[480],{"type":24,"value":481},"In a complementary review, Patterson, Nicklin, and Vitta (2025) consolidated methodological recommendations for webcam-based eye tracking — calibration, data quality control, sample size, and exclusion criteria — drawing on the growing body of online reading studies. The review has become the working reference for running a methodologically defensible webcam eye-tracking study at scale.",{"type":19,"tag":27,"props":483,"children":485},{"id":484},"direct-evidence-on-digital-reading-behaviour",[486],{"type":24,"value":487},"Direct evidence on digital reading behaviour",{"type":19,"tag":20,"props":489,"children":490},{},[491,493,498],{"type":24,"value":492},"Krenca, Taylor, and Deacon (2024), in the ",{"type":19,"tag":96,"props":494,"children":495},{},[496],{"type":24,"value":497},"Journal of Research in Reading",{"type":24,"value":499},", used the same platform to study how scrolling and hyperlinks — two of the most basic features of digital text — affect children's comprehension. The methodology behind that finding (eye tracking + behavioural measures + comprehension testing in a real online reading environment) is the same methodology used to measure adult attention on a marketing page or a documentation site.",{"type":19,"tag":27,"props":501,"children":503},{"id":502},"what-this-means-for-digital-experience-research",[504],{"type":24,"value":505},"What this means for digital experience research",{"type":19,"tag":20,"props":507,"children":508},{},[509],{"type":24,"value":510},"Reading is the demanding case. Studying where attention goes on a website, app, or marketing landing page is a substantially easier measurement problem — the regions of interest are larger, the dwell times longer, and the questions less about millisecond-level precision than about which elements get noticed and which don't.",{"type":19,"tag":20,"props":512,"children":513},{},[514],{"type":24,"value":515},"The same approach can answer commercial questions like:",{"type":19,"tag":118,"props":517,"children":518},{},[519,524,529,534],{"type":19,"tag":43,"props":520,"children":521},{},[522],{"type":24,"value":523},"Where does attention drop on a long landing page, and is the primary CTA inside or outside the high-attention zone?",{"type":19,"tag":43,"props":525,"children":526},{},[527],{"type":24,"value":528},"When users scan a comparison table, which columns do they fixate on first — and how does that change after a redesign?",{"type":19,"tag":43,"props":530,"children":531},{},[532],{"type":24,"value":533},"On a checkout flow, which fields produce hesitation (long pre-input dwell with no input), and where does attention leave the funnel?",{"type":19,"tag":43,"props":535,"children":536},{},[537],{"type":24,"value":538},"How does scrolling design or in-page navigation affect comprehension of a long-form article?",{"type":19,"tag":20,"props":540,"children":541},{},[542],{"type":24,"value":543},"These were the kind of questions the academic reading-research literature spent decades developing instruments for. The same instruments now run in a participant's own browser.",{"type":19,"tag":27,"props":545,"children":546},{"id":302},[547],{"type":24,"value":305},{"type":19,"tag":118,"props":549,"children":550},{},[551,566,580,594],{"type":19,"tag":43,"props":552,"children":553},{},[554,555,559,560,565],{"type":24,"value":313},{"type":19,"tag":96,"props":556,"children":557},{},[558],{"type":24,"value":318},{"type":24,"value":320},{"type":19,"tag":322,"props":561,"children":563},{"href":324,"rel":562},[326],[564],{"type":24,"value":329},{"type":24,"value":331},{"type":19,"tag":43,"props":567,"children":568},{},[569,570,574,575],{"type":24,"value":336},{"type":19,"tag":96,"props":571,"children":572},{},[573],{"type":24,"value":341},{"type":24,"value":343},{"type":19,"tag":322,"props":576,"children":578},{"href":346,"rel":577},[326],[579],{"type":24,"value":350},{"type":19,"tag":43,"props":581,"children":582},{},[583,584,588,589],{"type":24,"value":355},{"type":19,"tag":96,"props":585,"children":586},{},[587],{"type":24,"value":360},{"type":24,"value":362},{"type":19,"tag":322,"props":590,"children":592},{"href":365,"rel":591},[326],[593],{"type":24,"value":369},{"type":19,"tag":43,"props":595,"children":596},{},[597,599,604,606],{"type":24,"value":598},"Krenca, K., Taylor, E., & Deacon, S. H. (2024). ",{"type":19,"tag":96,"props":600,"children":601},{},[602],{"type":24,"value":603},"Scrolling and hyperlinks: The effects of two prevalent digital features on children's digital reading comprehension.",{"type":24,"value":605}," Journal of Research in Reading. ",{"type":19,"tag":322,"props":607,"children":610},{"href":608,"rel":609},"https:\u002F\u002Fdoi.org\u002F10.1111\u002F1467-9817.12468",[326],[611],{"type":24,"value":612},"doi.org\u002F10.1111\u002F1467-9817.12468",{"type":19,"tag":20,"props":614,"children":615},{},[616,618,622],{"type":24,"value":617},"For the full list of peer-reviewed work using the Pagegazer measurement platform, see ",{"type":19,"tag":322,"props":619,"children":620},{"href":385},[621],{"type":24,"value":388},{"type":24,"value":390},{"title":8,"searchDepth":392,"depth":392,"links":624},[625,626,627,628,629],{"id":430,"depth":392,"text":433},{"id":461,"depth":392,"text":464},{"id":484,"depth":392,"text":487},{"id":502,"depth":392,"text":505},{"id":302,"depth":392,"text":305},"content:research:insights:eye-tracking-the-online-reading-experience.md","research\u002Finsights\u002Feye-tracking-the-online-reading-experience.md","research\u002Finsights\u002Feye-tracking-the-online-reading-experience",{"_path":634,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":635,"description":636,"date":411,"author":12,"readTime":413,"category":14,"area":637,"body":638,"_type":401,"_id":1019,"_source":403,"_file":1020,"_stem":1021,"_extension":406},"\u002Fresearch\u002Finsights\u002Fwhat-packaging-research-looks-like","What packaging attention research actually looks like","A practical walkthrough of how a shelf-attention study is designed, what it measures, and what published research has shown about packaging cues.","Packaging & Visual",{"type":16,"children":639,"toc":1007},[640,652,658,663,675,681,686,739,765,771,776,783,802,807,813,825,830,836,841,847,852,895,900,919,923,997],{"type":19,"tag":20,"props":641,"children":642},{},[643,645,650],{"type":24,"value":644},"When a CPG brand commissions packaging research, the question is usually some version of ",{"type":19,"tag":96,"props":646,"children":647},{},[648],{"type":24,"value":649},"\"which design gets noticed first on the shelf?\"",{"type":24,"value":651}," The answer is rarely as simple as the question — and that is what makes attention measurement worth doing in the first place.",{"type":19,"tag":27,"props":653,"children":655},{"id":654},"the-setup",[656],{"type":24,"value":657},"The setup",{"type":19,"tag":20,"props":659,"children":660},{},[661],{"type":24,"value":662},"We simulate a shelf environment — either recreated from a planogram or photographed from an actual retail setting. The client's product is placed alongside its real competitors in realistic positions, with realistic spacing and lighting.",{"type":19,"tag":20,"props":664,"children":665},{},[666,668,673],{"type":24,"value":667},"Participants view the shelf on their own screen and complete a realistic shopping task — for example, ",{"type":19,"tag":96,"props":669,"children":670},{},[671],{"type":24,"value":672},"\"choose the breakfast cereal you would buy this week\"",{"type":24,"value":674}," — without being told which product is the focus of the research. Their eye movements are recorded continuously throughout. A short post-task interview captures stated preference and recall, which we compare against the observed attention data.",{"type":19,"tag":27,"props":676,"children":678},{"id":677},"what-we-actually-measure",[679],{"type":24,"value":680},"What we actually measure",{"type":19,"tag":20,"props":682,"children":683},{},[684],{"type":24,"value":685},"Raw eye-tracking data gives us fixation coordinates and durations. From those, we compute:",{"type":19,"tag":118,"props":687,"children":688},{},[689,699,709,719,729],{"type":19,"tag":43,"props":690,"children":691},{},[692,697],{"type":19,"tag":47,"props":693,"children":694},{},[695],{"type":24,"value":696},"Time to first fixation",{"type":24,"value":698}," — how quickly participants notice the product among distractors.",{"type":19,"tag":43,"props":700,"children":701},{},[702,707],{"type":19,"tag":47,"props":703,"children":704},{},[705],{"type":24,"value":706},"Dwell time",{"type":24,"value":708}," — how long they spend looking at each product.",{"type":19,"tag":43,"props":710,"children":711},{},[712,717],{"type":19,"tag":47,"props":713,"children":714},{},[715],{"type":24,"value":716},"Scan patterns",{"type":24,"value":718}," — the typical path the eye takes across the shelf.",{"type":19,"tag":43,"props":720,"children":721},{},[722,727],{"type":19,"tag":47,"props":723,"children":724},{},[725],{"type":24,"value":726},"Areas of interest (AOI)",{"type":24,"value":728}," — attention on specific design elements (logo, product window, claim, variant cue).",{"type":19,"tag":43,"props":730,"children":731},{},[732,737],{"type":19,"tag":47,"props":733,"children":734},{},[735],{"type":24,"value":736},"Brand-recognition latency",{"type":24,"value":738}," — the gap between fixating the pack and registering the brand.",{"type":19,"tag":20,"props":740,"children":741},{},[742,744,749,751,756,758,763],{"type":24,"value":743},"These metrics, taken together, separate the question of ",{"type":19,"tag":96,"props":745,"children":746},{},[747],{"type":24,"value":748},"visibility",{"type":24,"value":750}," (is the pack noticed at all?) from ",{"type":19,"tag":96,"props":752,"children":753},{},[754],{"type":24,"value":755},"recognition",{"type":24,"value":757}," (is the brand identified?) from ",{"type":19,"tag":96,"props":759,"children":760},{},[761],{"type":24,"value":762},"preference",{"type":24,"value":764}," (when noticed and recognised, is it chosen?).",{"type":19,"tag":27,"props":766,"children":768},{"id":767},"what-recent-published-research-has-shown",[769],{"type":24,"value":770},"What recent published research has shown",{"type":19,"tag":20,"props":772,"children":773},{},[774],{"type":24,"value":775},"The methodology behind shelf-attention studies is the subject of an active peer-reviewed literature using the same measurement platform. Three recent findings worth highlighting:",{"type":19,"tag":777,"props":778,"children":780},"h3",{"id":779},"_1-nutritional-labels-slow-product-detection",[781],{"type":24,"value":782},"1. Nutritional labels slow product detection",{"type":19,"tag":20,"props":784,"children":785},{},[786,788,793,795,800],{"type":24,"value":787},"González, Ojedo, Ruiz, and de Brugada (2025), published in ",{"type":19,"tag":96,"props":789,"children":790},{},[791],{"type":24,"value":792},"Food Quality and Preference",{"type":24,"value":794},", tested how mandatory nutritional warning labels — the kind used on packaging in many EU and Latin American markets — affect ",{"type":19,"tag":96,"props":796,"children":797},{},[798],{"type":24,"value":799},"product detection",{"type":24,"value":801}," speed. Products carrying a \"high in salt\" or \"high in saturated fat\" warning were detected more slowly than identical products without the warning. The finding goes beyond preference: the label actively interferes with a product's ability to capture attention on a busy shelf, before the shopper has formed any judgement at all.",{"type":19,"tag":20,"props":803,"children":804},{},[805],{"type":24,"value":806},"For a brand, this changes the question. It is no longer \"do shoppers prefer the labelled or unlabelled version?\" — it is \"by how much is shelf standout reduced, and how can other design choices compensate?\"",{"type":19,"tag":777,"props":808,"children":810},{"id":809},"_2-brand-assets-and-product-imagery-behave-differently-under-shopper-state",[811],{"type":24,"value":812},"2. Brand assets and product imagery behave differently under shopper state",{"type":19,"tag":20,"props":814,"children":815},{},[816,818,823],{"type":24,"value":817},"Ruiz, González, and de Brugada (2025) ran a complementary study on attentional capture under satiation. Hungry shoppers and recently-fed shoppers differ markedly in how strongly food-related imagery captures their attention — but they do ",{"type":19,"tag":96,"props":819,"children":820},{},[821],{"type":24,"value":822},"not",{"type":24,"value":824}," differ in how strongly food-brand logos do. Brand recognition appears to operate as a stable layer of attentional currency, robust to short-term shifts in shopper state, while product imagery rises and falls with appetite.",{"type":19,"tag":20,"props":826,"children":827},{},[828],{"type":24,"value":829},"Practically: a brand's logo is a more reliable shelf asset than its product photography. Logo prominence in a redesign is therefore a higher-stakes decision than it might appear.",{"type":19,"tag":777,"props":831,"children":833},{"id":832},"_3-categorisation-strategies-vary-across-consumer-groups",[834],{"type":24,"value":835},"3. Categorisation strategies vary across consumer groups",{"type":19,"tag":20,"props":837,"children":838},{},[839],{"type":24,"value":840},"Lakritz, Iceta, and Lafraire (2024) studied how individuals categorise food-related visual information, and how these strategies differ across populations. The relevance for packaging research is that two segments of the same target audience may not be looking at the same elements of the same pack — the cue that drives recognition for one group may be different from the cue that drives recognition for another. Sample composition matters; one number averaged across groups can hide the structure of the finding.",{"type":19,"tag":27,"props":842,"children":844},{"id":843},"what-the-findings-deliver",[845],{"type":24,"value":846},"What the findings deliver",{"type":19,"tag":20,"props":848,"children":849},{},[850],{"type":24,"value":851},"A typical packaging readout produces:",{"type":19,"tag":118,"props":853,"children":854},{},[855,865,875,885],{"type":19,"tag":43,"props":856,"children":857},{},[858,863],{"type":19,"tag":47,"props":859,"children":860},{},[861],{"type":24,"value":862},"Heatmaps",{"type":24,"value":864}," showing attention distribution across the shelf and across each pack, identifying which elements draw the eye and which are passed over.",{"type":19,"tag":43,"props":866,"children":867},{},[868,873],{"type":19,"tag":47,"props":869,"children":870},{},[871],{"type":24,"value":872},"Comparative metrics",{"type":24,"value":874}," for each design variant — time-to-first-fixation, total fixation count, dwell time, brand-recognition latency.",{"type":19,"tag":43,"props":876,"children":877},{},[878,883],{"type":19,"tag":47,"props":879,"children":880},{},[881],{"type":24,"value":882},"A separation of visibility from preference.",{"type":24,"value":884}," A pack can be highly preferred when shown in isolation but invisible on the shelf, or vice versa. Most of the time the actionable finding is one of these.",{"type":19,"tag":43,"props":886,"children":887},{},[888,893],{"type":19,"tag":47,"props":889,"children":890},{},[891],{"type":24,"value":892},"Specific design-element recommendations",{"type":24,"value":894}," — which cues to scale up, which to retain, which to retest. Recommendations that the design team can take into the next iteration without further interpretation.",{"type":19,"tag":20,"props":896,"children":897},{},[898],{"type":24,"value":899},"The value of attention measurement in packaging is not that it replaces preference testing — it is that it answers a question preference testing cannot ask.",{"type":19,"tag":280,"props":901,"children":902},{},[903,911],{"type":19,"tag":20,"props":904,"children":905},{},[906],{"type":19,"tag":287,"props":907,"children":910},{"alt":908,"src":909},"Placeholder — example shelf-attention heatmap","\u002Fimages\u002Fplaceholders\u002Fcasestudy-heatmap.svg",[],{"type":19,"tag":20,"props":912,"children":913},{},[914],{"type":19,"tag":96,"props":915,"children":916},{},[917],{"type":24,"value":918},"An anonymised heatmap from one of our studies, or a published figure under an open licence, would fit well here. We will replace this placeholder with a real example as engagements complete or as we identify a suitable open-access figure to license.",{"type":19,"tag":27,"props":920,"children":921},{"id":302},[922],{"type":24,"value":305},{"type":19,"tag":118,"props":924,"children":925},{},[926,945,964,983],{"type":19,"tag":43,"props":927,"children":928},{},[929,931,936,938],{"type":24,"value":930},"González, A., Ojedo, F., Ruiz, I., & de Brugada, I. (2025). ",{"type":19,"tag":96,"props":932,"children":933},{},[934],{"type":24,"value":935},"The effect of nutritional labels on the facilitation of Food Image Detection.",{"type":24,"value":937}," Food Quality and Preference. ",{"type":19,"tag":322,"props":939,"children":942},{"href":940,"rel":941},"https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.foodqual.2025.105547",[326],[943],{"type":24,"value":944},"doi.org\u002F10.1016\u002Fj.foodqual.2025.105547",{"type":19,"tag":43,"props":946,"children":947},{},[948,950,955,957],{"type":24,"value":949},"Ruiz, I., González, A., & de Brugada, I. (2025). ",{"type":19,"tag":96,"props":951,"children":952},{},[953],{"type":24,"value":954},"Satiation Modulates Attentional Capture by Food-Related Images But Not Food-Brand Logos.",{"type":24,"value":956}," SSRN. ",{"type":19,"tag":322,"props":958,"children":961},{"href":959,"rel":960},"https:\u002F\u002Fdoi.org\u002F10.2139\u002Fssrn.5115225",[326],[962],{"type":24,"value":963},"doi.org\u002F10.2139\u002Fssrn.5115225",{"type":19,"tag":43,"props":965,"children":966},{},[967,969,974,976],{"type":24,"value":968},"Lakritz, C., Iceta, S., & Lafraire, J. (2024). ",{"type":19,"tag":96,"props":970,"children":971},{},[972],{"type":24,"value":973},"Food Categorization Performance and Strategies in Orthorexia Nervosa.",{"type":24,"value":975}," Cognitive Therapy and Research. ",{"type":19,"tag":322,"props":977,"children":980},{"href":978,"rel":979},"https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs10608-024-10495-9",[326],[981],{"type":24,"value":982},"doi.org\u002F10.1007\u002Fs10608-024-10495-9",{"type":19,"tag":43,"props":984,"children":985},{},[986,987,991,992],{"type":24,"value":313},{"type":19,"tag":96,"props":988,"children":989},{},[990],{"type":24,"value":318},{"type":24,"value":320},{"type":19,"tag":322,"props":993,"children":995},{"href":324,"rel":994},[326],[996],{"type":24,"value":329},{"type":19,"tag":20,"props":998,"children":999},{},[1000,1002,1006],{"type":24,"value":1001},"For a fuller list of peer-reviewed work using the same measurement platform, see ",{"type":19,"tag":322,"props":1003,"children":1004},{"href":385},[1005],{"type":24,"value":388},{"type":24,"value":390},{"title":8,"searchDepth":392,"depth":392,"links":1008},[1009,1010,1011,1017,1018],{"id":654,"depth":392,"text":657},{"id":677,"depth":392,"text":680},{"id":767,"depth":392,"text":770,"children":1012},[1013,1015,1016],{"id":779,"depth":1014,"text":782},3,{"id":809,"depth":1014,"text":812},{"id":832,"depth":1014,"text":835},{"id":843,"depth":392,"text":846},{"id":302,"depth":392,"text":305},"content:research:insights:what-packaging-research-looks-like.md","research\u002Finsights\u002Fwhat-packaging-research-looks-like.md","research\u002Finsights\u002Fwhat-packaging-research-looks-like",{"_path":1023,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":1024,"description":1025,"date":1026,"author":412,"readTime":413,"category":414,"area":637,"body":1027,"_type":401,"_id":1277,"_source":403,"_file":1278,"_stem":1279,"_extension":406},"\u002Fresearch\u002Finsights\u002Fattention-to-food-labels-and-brand-cues","Methodology spotlight: how labels and brand cues compete for shopper attention","Recent peer-reviewed work has measured exactly which packaging cues capture attention — and how that capture changes with the consumer's state and context.","2026-04-10",{"type":16,"children":1028,"toc":1269},[1029,1034,1053,1059,1077,1088,1094,1112,1117,1123,1135,1141,1153,1159,1164,1187,1192,1196,1260],{"type":19,"tag":20,"props":1030,"children":1031},{},[1032],{"type":24,"value":1033},"A common assumption in packaging research is that \"preference equals performance\" — show shoppers two designs, ask which they prefer, ship the winner. But preference tests miss what happens at shelf, where the product has to win attention in under two seconds, often from a shopper who isn't in a buying mode.",{"type":19,"tag":20,"props":1035,"children":1036},{},[1037,1039,1044,1046,1051],{"type":24,"value":1038},"Four recent peer-reviewed studies — run on the platform that powers Pagegazer — separated ",{"type":19,"tag":96,"props":1040,"children":1041},{},[1042],{"type":24,"value":1043},"which cues capture attention",{"type":24,"value":1045}," from ",{"type":19,"tag":96,"props":1047,"children":1048},{},[1049],{"type":24,"value":1050},"which cues are preferred",{"type":24,"value":1052},", and showed how that capture shifts with shopper state, label content, and emotional context. The implications for packaging research are practical.",{"type":19,"tag":27,"props":1054,"children":1056},{"id":1055},"labels-change-which-products-get-noticed",[1057],{"type":24,"value":1058},"Labels change which products get noticed",{"type":19,"tag":20,"props":1060,"children":1061},{},[1062,1064,1068,1070,1075],{"type":24,"value":1063},"González, Ojedo, Ruiz, and de Brugada (2025), in ",{"type":19,"tag":96,"props":1065,"children":1066},{},[1067],{"type":24,"value":792},{"type":24,"value":1069},", tested how nutritional labels — the warning-style markers used on food packaging in many EU and Latin American markets — affect ",{"type":19,"tag":96,"props":1071,"children":1072},{},[1073],{"type":24,"value":1074},"image detection",{"type":24,"value":1076},": how quickly a shopper notices a product among distractors. Products carrying a \"high in salt\" or \"high in saturated fat\" warning were detected more slowly than identical products without the warning. The label, in other words, didn't just signal a value judgement at the point of choice — it actively interfered with attention capture earlier in the visual pipeline.",{"type":19,"tag":20,"props":1078,"children":1079},{},[1080,1082,1087],{"type":24,"value":1081},"For a brand, this is a very different finding from \"do consumers prefer the labelled or unlabelled version?\" It is a measurable handicap to ",{"type":19,"tag":96,"props":1083,"children":1084},{},[1085],{"type":24,"value":1086},"being seen on shelf at all",{"type":24,"value":390},{"type":19,"tag":27,"props":1089,"children":1091},{"id":1090},"brand-logos-behave-differently-from-product-imagery",[1092],{"type":24,"value":1093},"Brand logos behave differently from product imagery",{"type":19,"tag":20,"props":1095,"children":1096},{},[1097,1099,1104,1106,1110],{"type":24,"value":1098},"Ruiz, González, and de Brugada (2025) ran a complementary study on how a shopper's ",{"type":19,"tag":96,"props":1100,"children":1101},{},[1102],{"type":24,"value":1103},"internal state",{"type":24,"value":1105}," — being satiated rather than hungry — shifts attentional capture. Satiation reduced attention to food-related images (the things shoppers want when hungry). It did ",{"type":19,"tag":96,"props":1107,"children":1108},{},[1109],{"type":24,"value":822},{"type":24,"value":1111}," reduce attention to food-brand logos.",{"type":19,"tag":20,"props":1113,"children":1114},{},[1115],{"type":24,"value":1116},"In commercial terms: when a shopper has just eaten, the appetite-driven attentional pull toward product imagery weakens — but their brand recognition does not. Brand assets behave like a stable layer of attentional currency that survives shifts in the shopper's state. This makes logo prominence in a redesign a higher-stakes design decision than it might appear.",{"type":19,"tag":27,"props":1118,"children":1120},{"id":1119},"categorisation-strategies-vary-across-consumer-segments",[1121],{"type":24,"value":1122},"Categorisation strategies vary across consumer segments",{"type":19,"tag":20,"props":1124,"children":1125},{},[1126,1128,1133],{"type":24,"value":1127},"Lakritz, Iceta, and Lafraire (2024), in ",{"type":19,"tag":96,"props":1129,"children":1130},{},[1131],{"type":24,"value":1132},"Cognitive Therapy and Research",{"type":24,"value":1134},", studied how individuals categorise food-related visual information using a structured Go\u002FNo-Go paradigm. The relevance for packaging research is structural: two segments of the same target audience may not be looking at the same elements of the same pack. The cue that drives recognition for one group may be different from the cue that drives recognition for another. Sample composition matters — one number averaged across groups can hide the structure of the finding.",{"type":19,"tag":27,"props":1136,"children":1138},{"id":1137},"emotion-and-disgust-shape-what-registers",[1139],{"type":24,"value":1140},"Emotion and disgust shape what registers",{"type":19,"tag":20,"props":1142,"children":1143},{},[1144,1146,1151],{"type":24,"value":1145},"Gagliardi, Borghini, and Lafraire (2025), in ",{"type":19,"tag":96,"props":1147,"children":1148},{},[1149],{"type":24,"value":1150},"Appetite",{"type":24,"value":1152},", examined disgust reactions and their justifications in the context of meat. The finding most relevant to packaging is methodological: emotional reactions to food imagery are not noise sitting on top of a \"preference\" measurement — they are a measurable, structured part of how the product is processed. Studies that ignore the emotional channel underestimate the effect of imagery, copy, and context on consumer response.",{"type":19,"tag":27,"props":1154,"children":1156},{"id":1155},"what-this-means-for-packaging-research",[1157],{"type":24,"value":1158},"What this means for packaging research",{"type":19,"tag":20,"props":1160,"children":1161},{},[1162],{"type":24,"value":1163},"The same paradigms — combined with eye tracking on realistic shelf renderings — can answer questions a brand actually faces before a redesign ships:",{"type":19,"tag":118,"props":1165,"children":1166},{},[1167,1172,1177,1182],{"type":19,"tag":43,"props":1168,"children":1169},{},[1170],{"type":24,"value":1171},"Does the new pack get fixated faster than the version it replaces, on a competitive shelf?",{"type":19,"tag":43,"props":1173,"children":1174},{},[1175],{"type":24,"value":1176},"How does a regulatory label change shelf standout — is it offset by other design choices, or compounded?",{"type":19,"tag":43,"props":1178,"children":1179},{},[1180],{"type":24,"value":1181},"When the consumer arrives in different states (hungry vs. satiated, time-pressed vs. browsing), which design carries through?",{"type":19,"tag":43,"props":1183,"children":1184},{},[1185],{"type":24,"value":1186},"Does the imagery on the front of the pack provoke the intended emotional response, or does it cross into territory the segment finds off-putting?",{"type":19,"tag":20,"props":1188,"children":1189},{},[1190],{"type":24,"value":1191},"A \"preference\" test answers a question consumers can verbalise. Attention measurement, combined with implicit and emotional channels, reaches the moment that decides whether the product is even in the consideration set.",{"type":19,"tag":27,"props":1193,"children":1194},{"id":302},[1195],{"type":24,"value":305},{"type":19,"tag":118,"props":1197,"children":1198},{},[1199,1213,1227,1241],{"type":19,"tag":43,"props":1200,"children":1201},{},[1202,1203,1207,1208],{"type":24,"value":930},{"type":19,"tag":96,"props":1204,"children":1205},{},[1206],{"type":24,"value":935},{"type":24,"value":937},{"type":19,"tag":322,"props":1209,"children":1211},{"href":940,"rel":1210},[326],[1212],{"type":24,"value":944},{"type":19,"tag":43,"props":1214,"children":1215},{},[1216,1217,1221,1222],{"type":24,"value":949},{"type":19,"tag":96,"props":1218,"children":1219},{},[1220],{"type":24,"value":954},{"type":24,"value":956},{"type":19,"tag":322,"props":1223,"children":1225},{"href":959,"rel":1224},[326],[1226],{"type":24,"value":963},{"type":19,"tag":43,"props":1228,"children":1229},{},[1230,1231,1235,1236],{"type":24,"value":968},{"type":19,"tag":96,"props":1232,"children":1233},{},[1234],{"type":24,"value":973},{"type":24,"value":975},{"type":19,"tag":322,"props":1237,"children":1239},{"href":978,"rel":1238},[326],[1240],{"type":24,"value":982},{"type":19,"tag":43,"props":1242,"children":1243},{},[1244,1246,1251,1253],{"type":24,"value":1245},"Gagliardi, L., Borghini, A., & Lafraire, J. (2025). ",{"type":19,"tag":96,"props":1247,"children":1248},{},[1249],{"type":24,"value":1250},"Disgust reactions and their justifications: The case of meat.",{"type":24,"value":1252}," Appetite. ",{"type":19,"tag":322,"props":1254,"children":1257},{"href":1255,"rel":1256},"https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.appet.2025.108083",[326],[1258],{"type":24,"value":1259},"doi.org\u002F10.1016\u002Fj.appet.2025.108083",{"type":19,"tag":20,"props":1261,"children":1262},{},[1263,1264,1268],{"type":24,"value":617},{"type":19,"tag":322,"props":1265,"children":1266},{"href":385},[1267],{"type":24,"value":388},{"type":24,"value":390},{"title":8,"searchDepth":392,"depth":392,"links":1270},[1271,1272,1273,1274,1275,1276],{"id":1055,"depth":392,"text":1058},{"id":1090,"depth":392,"text":1093},{"id":1119,"depth":392,"text":1122},{"id":1137,"depth":392,"text":1140},{"id":1155,"depth":392,"text":1158},{"id":302,"depth":392,"text":305},"content:research:insights:attention-to-food-labels-and-brand-cues.md","research\u002Finsights\u002Fattention-to-food-labels-and-brand-cues.md","research\u002Finsights\u002Fattention-to-food-labels-and-brand-cues",{"_path":1281,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":1282,"description":1283,"date":1284,"author":412,"readTime":413,"category":414,"area":1285,"body":1286,"_type":401,"_id":1557,"_source":403,"_file":1558,"_stem":1559,"_extension":406},"\u002Fresearch\u002Finsights\u002Fphysiology-of-engagement-with-creative","Methodology spotlight: separating curiosity from understanding from liking","Recent research shows that distinct physiological signatures track different stages of engagement with creative work — far richer than a 'did you like it' score.","2026-04-05","Ad & Video",{"type":16,"children":1287,"toc":1549},[1288,1293,1298,1304,1316,1321,1327,1339,1344,1350,1376,1381,1387,1399,1405,1410,1433,1438,1442,1540],{"type":19,"tag":20,"props":1289,"children":1290},{},[1291],{"type":24,"value":1292},"Ad pre-testing usually ends with a likability score and some recall questions. The scores are easy to read and easy to defend — but they collapse a complex mental process (\"I noticed it… I'm trying to figure it out… I get it now… I like it\") into a single number measured at the wrong moment.",{"type":19,"tag":20,"props":1294,"children":1295},{},[1296],{"type":24,"value":1297},"Recent peer-reviewed work — run on the platform that powers Pagegazer — demonstrates that the underlying experience can be pulled apart, in real time, using readily available signals: gaze, pupillometry, heart rate, and continuous response.",{"type":19,"tag":27,"props":1299,"children":1301},{"id":1300},"curiosity-insight-understanding-and-liking-are-different-states",[1302],{"type":24,"value":1303},"Curiosity, insight, understanding, and liking are different states",{"type":19,"tag":20,"props":1305,"children":1306},{},[1307,1309,1314],{"type":24,"value":1308},"Welke and Vessel (2025) tracked viewers continuously during interactions with visual art and identified ",{"type":19,"tag":96,"props":1310,"children":1311},{},[1312],{"type":24,"value":1313},"distinct physiological signatures",{"type":24,"value":1315}," for four stages of an aesthetic encounter: curiosity (the moment of approach), insight (the moment of \"getting it\"), understanding (sustained comprehension), and liking (evaluative response). Heart-rate dynamics, pupil dilation, and gaze patterns each contributed differently to identifying which stage the viewer was in at any given moment.",{"type":19,"tag":20,"props":1317,"children":1318},{},[1319],{"type":24,"value":1320},"In a commercial creative context — a TV spot, a social video, a long-form ad — these stages happen in seconds, often in the same order. A viewer can be highly curious about an ad, never reach insight, and rate it as \"fine\" at the end. The likability score reads as flat. The physiological time-series shows exactly where the experience broke down.",{"type":19,"tag":27,"props":1322,"children":1324},{"id":1323},"music-and-visuals-dont-combine-the-way-intuition-suggests",[1325],{"type":24,"value":1326},"Music and visuals don't combine the way intuition suggests",{"type":19,"tag":20,"props":1328,"children":1329},{},[1330,1332,1337],{"type":24,"value":1331},"Fink, Fiehn, and Wald-Fuhrmann (2024), in ",{"type":19,"tag":96,"props":1333,"children":1334},{},[1335],{"type":24,"value":1336},"Scientific Reports",{"type":24,"value":1338},", measured what happens when music is paired with visual art — congruent versus incongruent matches. The aesthetic effect of audiovisual pairing was not a simple sum of the two channels' separate effects. Match quality changed how the visuals were processed, including where attention went and how engaged viewers were.",{"type":19,"tag":20,"props":1340,"children":1341},{},[1342],{"type":24,"value":1343},"For ad and video research, this finding maps directly: the music bed, voiceover tone, and sound design are not background; they actively shape how the visuals are received. A re-edit that swaps the audio while leaving the cut intact can change attention patterns measurably.",{"type":19,"tag":27,"props":1345,"children":1347},{"id":1346},"voice-carries-more-than-people-realise",[1348],{"type":24,"value":1349},"Voice carries more than people realise",{"type":19,"tag":20,"props":1351,"children":1352},{},[1353,1355,1360,1362,1367,1369,1374],{"type":24,"value":1354},"Two related studies extend the finding into voice. Bruder, Frieler, and Larrouy-Maestri (2024), in ",{"type":19,"tag":96,"props":1356,"children":1357},{},[1358],{"type":24,"value":1359},"Royal Society Open Science",{"type":24,"value":1361},", showed that appreciation of singing and speaking voices is ",{"type":19,"tag":47,"props":1363,"children":1364},{},[1365],{"type":24,"value":1366},"highly idiosyncratic",{"type":24,"value":1368}," — a much smaller share of voice preference is shared across listeners than industry assumptions suggest. Bruder, Breda, and Larrouy-Maestri (2025) extended the work to synthetic voices in ",{"type":19,"tag":96,"props":1370,"children":1371},{},[1372],{"type":24,"value":1373},"Computers in Human Behavior: Artificial Humans",{"type":24,"value":1375},", with implications for any brand using AI-generated narration.",{"type":19,"tag":20,"props":1377,"children":1378},{},[1379],{"type":24,"value":1380},"For a marketer, the practical takeaway is that \"we tested the voice and people liked it\" is a weaker claim than it sounds. Voice preference has a large individual-difference component; segment-level analysis is required, and a voice that wins on average can lose decisively in important sub-populations.",{"type":19,"tag":27,"props":1382,"children":1384},{"id":1383},"attention-to-the-brand-reveal-at-pnas-level-rigour",[1385],{"type":24,"value":1386},"Attention to the brand reveal — at PNAS-level rigour",{"type":19,"tag":20,"props":1388,"children":1389},{},[1390,1392,1397],{"type":24,"value":1391},"A recent study at the highest tier of the literature illustrates how robustly these methods can resolve subtle effects. Canessa-Pollard, Anikin, and Reby (2025), in ",{"type":19,"tag":96,"props":1393,"children":1394},{},[1395],{"type":24,"value":1396},"Proceedings of the National Academy of Sciences",{"type":24,"value":1398},", identified shared acoustic features across chant traditions from seven cultures that consistently produce subjective relaxation. The methodology — controlled stimulus presentation, continuous physiological measurement, and structured response collection in the participant's own browser — is the same instrumentation used to test whether an ad's audio bed lands the intended emotional effect.",{"type":19,"tag":27,"props":1400,"children":1402},{"id":1401},"what-this-means-for-ad-and-video-research",[1403],{"type":24,"value":1404},"What this means for ad and video research",{"type":19,"tag":20,"props":1406,"children":1407},{},[1408],{"type":24,"value":1409},"The methodology behind these studies — eye tracking, heart-rate measurement (via webcam rPPG), facial-expression analysis, all in the participant's own browser — is what Pagegazer runs for ad pre-testing. The questions it answers, in commercial terms:",{"type":19,"tag":118,"props":1411,"children":1412},{},[1413,1418,1423,1428],{"type":19,"tag":43,"props":1414,"children":1415},{},[1416],{"type":24,"value":1417},"Where in the cut does attention drop, and is the brand reveal inside or outside the high-attention window?",{"type":19,"tag":43,"props":1419,"children":1420},{},[1421],{"type":24,"value":1422},"Does the music edit support or fight the intended emotional arc?",{"type":19,"tag":43,"props":1424,"children":1425},{},[1426],{"type":24,"value":1427},"Are viewers reaching \"understanding\" before the call-to-action, or is the spot landing the message after they have disengaged?",{"type":19,"tag":43,"props":1429,"children":1430},{},[1431],{"type":24,"value":1432},"Does a voice-over choice work across audience segments, or only in some?",{"type":19,"tag":20,"props":1434,"children":1435},{},[1436],{"type":24,"value":1437},"A likability score is a single readout. Continuous physiology and attention provide the timeline.",{"type":19,"tag":27,"props":1439,"children":1440},{"id":302},[1441],{"type":24,"value":305},{"type":19,"tag":118,"props":1443,"children":1444},{},[1445,1464,1483,1502,1521],{"type":19,"tag":43,"props":1446,"children":1447},{},[1448,1450,1455,1457],{"type":24,"value":1449},"Welke, D., & Vessel, E. A. (2025). ",{"type":19,"tag":96,"props":1451,"children":1452},{},[1453],{"type":24,"value":1454},"Tracing the Epistemic Arc: Distinct Physiological Signatures for Curiosity, Insight, Understanding and Liking during Interactions with Visual Art.",{"type":24,"value":1456}," bioRxiv. ",{"type":19,"tag":322,"props":1458,"children":1461},{"href":1459,"rel":1460},"https:\u002F\u002Fdoi.org\u002F10.1101\u002F2025.05.15.654230",[326],[1462],{"type":24,"value":1463},"doi.org\u002F10.1101\u002F2025.05.15.654230",{"type":19,"tag":43,"props":1465,"children":1466},{},[1467,1469,1474,1476],{"type":24,"value":1468},"Fink, L., Fiehn, H., & Wald-Fuhrmann, M. (2024). ",{"type":19,"tag":96,"props":1470,"children":1471},{},[1472],{"type":24,"value":1473},"The role of audiovisual congruence in perception and aesthetic appreciation of contemporary music and visual art.",{"type":24,"value":1475}," Scientific Reports. ",{"type":19,"tag":322,"props":1477,"children":1480},{"href":1478,"rel":1479},"https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41598-024-71399-y",[326],[1481],{"type":24,"value":1482},"doi.org\u002F10.1038\u002Fs41598-024-71399-y",{"type":19,"tag":43,"props":1484,"children":1485},{},[1486,1488,1493,1495],{"type":24,"value":1487},"Bruder, C., Frieler, K., & Larrouy-Maestri, P. (2024). ",{"type":19,"tag":96,"props":1489,"children":1490},{},[1491],{"type":24,"value":1492},"Appreciation of singing and speaking voices is highly idiosyncratic.",{"type":24,"value":1494}," Royal Society Open Science. ",{"type":19,"tag":322,"props":1496,"children":1499},{"href":1497,"rel":1498},"https:\u002F\u002Fdoi.org\u002F10.1098\u002Frsos.241623",[326],[1500],{"type":24,"value":1501},"doi.org\u002F10.1098\u002Frsos.241623",{"type":19,"tag":43,"props":1503,"children":1504},{},[1505,1507,1512,1514],{"type":24,"value":1506},"Bruder, C., Breda, P., & Larrouy-Maestri, P. (2025). ",{"type":19,"tag":96,"props":1508,"children":1509},{},[1510],{"type":24,"value":1511},"Attractive synthetic voices.",{"type":24,"value":1513}," Computers in Human Behavior: Artificial Humans. ",{"type":19,"tag":322,"props":1515,"children":1518},{"href":1516,"rel":1517},"https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.chbah.2025.100211",[326],[1519],{"type":24,"value":1520},"doi.org\u002F10.1016\u002Fj.chbah.2025.100211",{"type":19,"tag":43,"props":1522,"children":1523},{},[1524,1526,1531,1533],{"type":24,"value":1525},"Canessa-Pollard, V., Anikin, A., & Reby, D. (2025). ",{"type":19,"tag":96,"props":1527,"children":1528},{},[1529],{"type":24,"value":1530},"Chants across seven traditions share acoustic traits that enhance subjective relaxation.",{"type":24,"value":1532}," Proceedings of the National Academy of Sciences. ",{"type":19,"tag":322,"props":1534,"children":1537},{"href":1535,"rel":1536},"https:\u002F\u002Fdoi.org\u002F10.1073\u002Fpnas.2506480122",[326],[1538],{"type":24,"value":1539},"doi.org\u002F10.1073\u002Fpnas.2506480122",{"type":19,"tag":20,"props":1541,"children":1542},{},[1543,1544,1548],{"type":24,"value":617},{"type":19,"tag":322,"props":1545,"children":1546},{"href":385},[1547],{"type":24,"value":388},{"type":24,"value":390},{"title":8,"searchDepth":392,"depth":392,"links":1550},[1551,1552,1553,1554,1555,1556],{"id":1300,"depth":392,"text":1303},{"id":1323,"depth":392,"text":1326},{"id":1346,"depth":392,"text":1349},{"id":1383,"depth":392,"text":1386},{"id":1401,"depth":392,"text":1404},{"id":302,"depth":392,"text":305},"content:research:insights:physiology-of-engagement-with-creative.md","research\u002Finsights\u002Fphysiology-of-engagement-with-creative.md","research\u002Finsights\u002Fphysiology-of-engagement-with-creative",{"_path":1561,"_dir":6,"_draft":7,"_partial":7,"_locale":8,"title":1562,"description":1563,"date":1564,"author":412,"readTime":1565,"category":414,"area":1566,"body":1567,"_type":401,"_id":1852,"_source":403,"_file":1853,"_stem":1854,"_extension":406},"\u002Fresearch\u002Finsights\u002Fmeasuring-trust-in-ai-tools-at-work","Methodology spotlight: how knowledge workers actually trust AI tools","A series of peer-reviewed studies measured pharmacists using AI-assisted decision tools — and what they found generalises directly to enterprise software UX.","2026-03-30","8 min","Behavioral & Custom",{"type":16,"children":1568,"toc":1843},[1569,1574,1579,1585,1597,1602,1608,1620,1632,1638,1664,1670,1689,1695,1707,1713,1718,1746,1751,1755,1834],{"type":19,"tag":20,"props":1570,"children":1571},{},[1572],{"type":24,"value":1573},"Trust in AI is the question of the moment for enterprise software. Every B2B SaaS roadmap in 2026 includes an AI feature, a copilot, an agentic step. The question every vendor is asking — and very few are measuring well — is whether their users actually trust the AI's output enough to act on it, or are quietly working around it.",{"type":19,"tag":20,"props":1575,"children":1576},{},[1577],{"type":24,"value":1578},"A connected series of peer-reviewed studies — run on the platform that powers Pagegazer — measured exactly this. The studied population was pharmacists doing medication verification with AI assistance. It is a high-stakes, real-task domain, but the methods and findings transfer directly to general enterprise UX research on AI features.",{"type":19,"tag":27,"props":1580,"children":1582},{"id":1581},"trust-is-not-uniform-and-depends-on-demographics-and-context",[1583],{"type":24,"value":1584},"Trust is not uniform — and depends on demographics and context",{"type":19,"tag":20,"props":1586,"children":1587},{},[1588,1590,1595],{"type":24,"value":1589},"Whitaker, Rowell, Kim, Al Kontar, Yang, and Lester (2026), in the ",{"type":19,"tag":96,"props":1591,"children":1592},{},[1593],{"type":24,"value":1594},"Journal of the American Pharmacists Association",{"type":24,"value":1596},", measured pharmacists' propensity to trust automated technologies as a function of demographic and contextual factors. The headline result is that trust is not a single value users carry into an interaction — it varies systematically by experience level, prior exposure to automation, and the kind of task. Two pharmacists looking at the same AI suggestion bring different priors and behave differently.",{"type":19,"tag":20,"props":1598,"children":1599},{},[1600],{"type":24,"value":1601},"The implication for enterprise UX research is that population-level satisfaction scores hide the variance that matters. A 7.4 on a trust scale can mean \"everyone trusts it 7.4\" or \"half are 9, half are 5 and quietly ignoring it.\" The latter is a product problem; the former isn't.",{"type":19,"tag":27,"props":1603,"children":1605},{"id":1604},"showing-ai-uncertainty-changes-what-knowledge-workers-do",[1606],{"type":24,"value":1607},"Showing AI uncertainty changes what knowledge workers do",{"type":19,"tag":20,"props":1609,"children":1610},{},[1611,1613,1618],{"type":24,"value":1612},"Lester, Rowell, Zheng, Co, Marshall, Kim, Chen, Kontar, and Yang (2025) ran a randomized controlled trial measuring how ",{"type":19,"tag":96,"props":1614,"children":1615},{},[1616],{"type":24,"value":1617},"uncertainty-aware AI models",{"type":24,"value":1619}," — AI that displays its confidence — change pharmacists' reaction time and decision-making in a mock medication verification task. Showing uncertainty did not just shift trust ratings; it changed the behaviour itself: how long pharmacists spent looking at the AI's suggestion, when they overrode it, and how they paced themselves through verification.",{"type":19,"tag":20,"props":1621,"children":1622},{},[1623,1625,1630],{"type":24,"value":1624},"This was published in ",{"type":19,"tag":96,"props":1626,"children":1627},{},[1628],{"type":24,"value":1629},"JMIR Medical Informatics",{"type":24,"value":1631}," — a journal where claims have to clear methodological bars not unlike a clinical trial. The study was randomised, controlled, and large enough to distinguish behavioural change from noise. The implication is that displaying confidence is not a UI detail. It is part of the decision system the user is operating in.",{"type":19,"tag":27,"props":1633,"children":1635},{"id":1634},"helpfulness-and-uncertainty-interact",[1636],{"type":24,"value":1637},"Helpfulness and uncertainty interact",{"type":19,"tag":20,"props":1639,"children":1640},{},[1641,1643,1648,1650,1655,1657,1662],{"type":24,"value":1642},"In a related study, Tsai et al. (2025) extended the question to how ",{"type":19,"tag":96,"props":1644,"children":1645},{},[1646],{"type":24,"value":1647},"AI helpfulness",{"type":24,"value":1649}," and ",{"type":19,"tag":96,"props":1651,"children":1652},{},[1653],{"type":24,"value":1654},"AI uncertainty",{"type":24,"value":1656}," interact in shaping pharmacist cognition, published in the ",{"type":19,"tag":96,"props":1658,"children":1659},{},[1660],{"type":24,"value":1661},"Journal of Medical Internet Research",{"type":24,"value":1663},". Showing high helpfulness without uncertainty produced different patterns of cognitive engagement than showing helpfulness with uncertainty cues. The combination of the two is the design surface — not either independently.",{"type":19,"tag":27,"props":1665,"children":1667},{"id":1666},"earlier-stage-exploration-ai-with-different-uncertainty-representations",[1668],{"type":24,"value":1669},"Earlier-stage exploration: AI with different uncertainty representations",{"type":19,"tag":20,"props":1671,"children":1672},{},[1673,1675,1680,1682,1687],{"type":24,"value":1674},"A related earlier study by Kim and colleagues (2025), in ",{"type":19,"tag":96,"props":1676,"children":1677},{},[1678],{"type":24,"value":1679},"JMIR Human Factors",{"type":24,"value":1681},", presented different forms of AI uncertainty information (probabilistic, categorical, qualitative) and measured pharmacists' trust response across them. The finding was specific: ",{"type":19,"tag":96,"props":1683,"children":1684},{},[1685],{"type":24,"value":1686},"how",{"type":24,"value":1688}," uncertainty is shown changes the trust response measurably, and the optimal representation depends on the user's prior experience with automation. This is actionable for any enterprise team designing how their AI feature surfaces confidence.",{"type":19,"tag":27,"props":1690,"children":1692},{"id":1691},"self-reported-trust-diverges-from-observed-behaviour",[1693],{"type":24,"value":1694},"Self-reported trust diverges from observed behaviour",{"type":19,"tag":20,"props":1696,"children":1697},{},[1698,1700,1705],{"type":24,"value":1699},"The methodological lesson across all four studies is that ",{"type":19,"tag":47,"props":1701,"children":1702},{},[1703],{"type":24,"value":1704},"self-reported trust and observed trust behaviour are not the same construct.",{"type":24,"value":1706}," Saying \"I trust this\" in a survey is informative; pausing on the AI suggestion for 1.4× longer than baseline before accepting it is a different kind of informative. Both should be measured. When they diverge, the observed behaviour is usually the better predictor of what will happen at scale.",{"type":19,"tag":27,"props":1708,"children":1710},{"id":1709},"what-this-means-for-enterprise-ux-research",[1711],{"type":24,"value":1712},"What this means for enterprise UX research",{"type":19,"tag":20,"props":1714,"children":1715},{},[1716],{"type":24,"value":1717},"The same methods — eye tracking on the AI suggestion field, reaction-time measurement, decision tracking, in the user's own browser — answer the questions enterprise software teams actually have:",{"type":19,"tag":118,"props":1719,"children":1720},{},[1721,1726,1731,1736,1741],{"type":19,"tag":43,"props":1722,"children":1723},{},[1724],{"type":24,"value":1725},"Where does the user's eye go first when the AI's suggestion appears? Do they read it, or skip past?",{"type":19,"tag":43,"props":1727,"children":1728},{},[1729],{"type":24,"value":1730},"How long do they spend evaluating it before accepting or overriding?",{"type":19,"tag":43,"props":1732,"children":1733},{},[1734],{"type":24,"value":1735},"Does adding a confidence indicator change accept rates — and does it change them in the right direction (more rejection of low-confidence outputs, sustained acceptance of high-confidence ones)?",{"type":19,"tag":43,"props":1737,"children":1738},{},[1739],{"type":24,"value":1740},"Which segments of users adopt the AI feature and which work around it, and is that segmentation visible in the behaviour before it shows up in feature-adoption analytics?",{"type":19,"tag":43,"props":1742,"children":1743},{},[1744],{"type":24,"value":1745},"Which uncertainty representation works best for which segment of your user base?",{"type":19,"tag":20,"props":1747,"children":1748},{},[1749],{"type":24,"value":1750},"The pharmacist studies are the clearest demonstration in the published literature that these questions can be measured precisely, with methods that scale to a representative sample and don't require lab visits. The same instrumentation runs for any knowledge worker using any software in any browser.",{"type":19,"tag":27,"props":1752,"children":1753},{"id":302},[1754],{"type":24,"value":305},{"type":19,"tag":118,"props":1756,"children":1757},{},[1758,1777,1796,1815],{"type":19,"tag":43,"props":1759,"children":1760},{},[1761,1763,1768,1770],{"type":24,"value":1762},"Whitaker, M., Rowell, B., Kim, J. Y., Al Kontar, R., Yang, X. J., & Lester, C. A. (2026). ",{"type":19,"tag":96,"props":1764,"children":1765},{},[1766],{"type":24,"value":1767},"Pharmacists' propensity to trust automated technologies: A demographic analysis.",{"type":24,"value":1769}," Journal of the American Pharmacists Association. ",{"type":19,"tag":322,"props":1771,"children":1774},{"href":1772,"rel":1773},"https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.japh.2025.103011",[326],[1775],{"type":24,"value":1776},"doi.org\u002F10.1016\u002Fj.japh.2025.103011",{"type":19,"tag":43,"props":1778,"children":1779},{},[1780,1782,1787,1789],{"type":24,"value":1781},"Lester, C., Rowell, B., Zheng, Y., Co, Z., Marshall, V., Kim, J. Y., Chen, Q., Kontar, R., & Yang, X. J. (2025). ",{"type":19,"tag":96,"props":1783,"children":1784},{},[1785],{"type":24,"value":1786},"Effect of uncertainty-aware AI models on pharmacists' reaction time and decision-making in a web-based mock medication verification task.",{"type":24,"value":1788}," JMIR Medical Informatics. ",{"type":19,"tag":322,"props":1790,"children":1793},{"href":1791,"rel":1792},"https:\u002F\u002Fdoi.org\u002F10.2196\u002F64902",[326],[1794],{"type":24,"value":1795},"doi.org\u002F10.2196\u002F64902",{"type":19,"tag":43,"props":1797,"children":1798},{},[1799,1801,1806,1808],{"type":24,"value":1800},"Tsai, C. C., Kim, J. Y., Chen, Q., Rowell, B., Yang, X. J., Kontar, R., et al. (2025). ",{"type":19,"tag":96,"props":1802,"children":1803},{},[1804],{"type":24,"value":1805},"Effect of Artificial Intelligence Helpfulness and Uncertainty on Cognitive Interactions with Pharmacists.",{"type":24,"value":1807}," Journal of Medical Internet Research. ",{"type":19,"tag":322,"props":1809,"children":1812},{"href":1810,"rel":1811},"https:\u002F\u002Fdoi.org\u002F10.2196\u002F59946",[326],[1813],{"type":24,"value":1814},"doi.org\u002F10.2196\u002F59946",{"type":19,"tag":43,"props":1816,"children":1817},{},[1818,1820,1825,1827],{"type":24,"value":1819},"Kim, J. Y., Marshall, V. D., Rowell, B., Chen, Q., Zheng, Y., Lee, J. D., Al Kontar, R., Lester, C., & Yang, X. J. (2025). ",{"type":19,"tag":96,"props":1821,"children":1822},{},[1823],{"type":24,"value":1824},"The Effects of Presenting AI Uncertainty Information on Pharmacists' Trust in Automated Pill Recognition Technology.",{"type":24,"value":1826}," JMIR Human Factors. ",{"type":19,"tag":322,"props":1828,"children":1831},{"href":1829,"rel":1830},"https:\u002F\u002Fdoi.org\u002F10.2196\u002F60273",[326],[1832],{"type":24,"value":1833},"doi.org\u002F10.2196\u002F60273",{"type":19,"tag":20,"props":1835,"children":1836},{},[1837,1838,1842],{"type":24,"value":617},{"type":19,"tag":322,"props":1839,"children":1840},{"href":385},[1841],{"type":24,"value":388},{"type":24,"value":390},{"title":8,"searchDepth":392,"depth":392,"links":1844},[1845,1846,1847,1848,1849,1850,1851],{"id":1581,"depth":392,"text":1584},{"id":1604,"depth":392,"text":1607},{"id":1634,"depth":392,"text":1637},{"id":1666,"depth":392,"text":1669},{"id":1691,"depth":392,"text":1694},{"id":1709,"depth":392,"text":1712},{"id":302,"depth":392,"text":305},"content:research:insights:measuring-trust-in-ai-tools-at-work.md","research\u002Finsights\u002Fmeasuring-trust-in-ai-tools-at-work.md","research\u002Finsights\u002Fmeasuring-trust-in-ai-tools-at-work",1780554429670]