Banner blindness

The name comes from Benway (1998), and her “banners” were not advertising: they were large, deliberately highlighted links to the very information participants had been asked to find, and they were missed anyway. Geometry and task predicted the misses. Banners placed higher on the page and further from the other links were missed more often than banners lower and closer to them, they were missed more often on pages listing categories than on pages listing specific items, and users “saw banners hardly at all when clicking a banner was not required to accomplish a task”. The user who most wants the highlighted information is the one least likely to find it.

Task orientation is the clearest moderator. Pagendarm and Schaumburg (2001) ran the same banner-carrying site under browsing and searching instructions, then tested memory. Aimless browsers reproduced a mean of 1.00 advertised companies, products or services against 0.06 for the goal-directed searchers, and 1.87 against 0.63 pictorial motifs and colours. Same page, same banners: whether the user is mid-task decides whether the message is retained at all.

Position, not styling, is the variable you can move. Burke, Hornof, Nilsen and Gorman (2005) eye-tracked participants searching headline layouts: banners were fixated in 11.7% of trials at a rate that did not depend on appearance, with plain gray placeholder rectangles drawing gazes in 8.8% of trials, static commercial banners in 13.0% and animated ones in 13.2%, p = .29. Predictability mattered instead — the banner pinned to a fixed slot at the top of the screen was fixated significantly less often than the same banner dropped into a random row inside the content, p = .02. Resnick and Albert (2014) found blindness strongest for banners in the right rail and under goal-directed tasks, with neither rated visual appeal nor page familiarity explaining it. On search pages, Schultheiß and Lewandowski (2021) tracked 100 people over 2,000 tasks: the first labelled text ad above the organic results took 24.2% of desktop fixations and 32.4% on a smartphone, while the same kind of ad at the foot of the page drew 2.5–8.6% and was clicked zero times. Ads took 25% of desktop and 29% of smartphone clicks on navigational tasks, falling to 3–4% on informational ones.

Fixation and memory come apart in both directions, so pick your measure deliberately. In Burke's study recognition was at chance — a 20.1% hit rate against a 20.2% false-alarm rate — and 93.9% of the banners participants did correctly recognise had never received a direct gaze. Hervet, Guérard, Tremblay and Chtourou (2011) note that the avoidance conclusion rested “mostly on indirect evidence that participants do not remember the ad content”, and found most participants fixated the ads at least once per visit. Simonetti and Bigne (2024) put three banners in the content column of a news page: while reading a story, 93% of 100 participants looked at all three yet gave them 5.5% of their dwell time, and in a “find where to click next” task only 55% looked at all three but ad dwell share rose to 11.9%. The eye lands; it just does not stay, and neither impressions nor gaze tells you a message landed.

Turning up the volume does not reverse it and costs the host task. Static banners in Burke's study were remembered above chance (d′ = .667) while animated look-alikes were not (d′ = −.07). Commercial banners slowed easy search by roughly 6–7% against gray rectangles (2,169 ms static and 2,193 ms animated against 2,040 ms) and not at all on a harder semantic task, but banners of flashing cyan text raised measured workload and a target sandwiched between two of them took 75% — about 4.3 seconds — longer to find. Sapronov and Gorbunova (2022) varied banner arousal and valence in goal-directed search and changed nothing: 28 and 29 of 40 participants reported seeing high- and low-arousal banners, the counts across positive, neutral and negative banners were 31, 32 and 29, and neutral banners were recognised better than emotional ones.

Being skipped is not the same as having no effect. Drèze and Hussherr (2003) confirmed that surfers avoid looking at banners — their reason for calling click-through the wrong metric — and still found repetition moving unaided recall, brand recognition and brand awareness. Lewis and Reiley randomised 1,577,256 matched retail customers on Yahoo!: click-through was 0.28% and only 7.2% of treated users ever clicked, yet sales rose by R$0.166 per treated person over three weeks, 93% of it in physical stores, and non-clickers accounted for 78% of the total effect. An element nobody reports seeing can still be working.

Two limits. Every interface study here is lab work with 24 to 120 participants on mock pages built around 2000s layouts of top bars and right rails, and the measures disagree — chance-level recognition in one study, 93% of readers fixating every banner in another. And nothing here shows that looking commercial is what gets an element skipped: the condition skipped just as thoroughly was a plain gray rectangle in the same slot. No study in this literature tested the advice teams draw from the principle, that a genuine product notice styled like a promo will be missed. Treat that as a hypothesis for your own instrumentation; the supported moves are positional and static.

The anti-pattern is regulated, not merely rude. The FTC's Native Advertising: A Guide for Businesses treats a format as deceptive where it materially misleads about the commercial nature of the content, wants the disclosure in front of or above the headline in wording such as “Ad” or “Paid advertisement” rather than “Promoted”, and says company logos and names alone are not likely adequate. In the UK, paragraph 12 of Schedule 20 of the Digital Markets, Competition and Consumers Act 2024, in force 6 April 2025, makes undisclosed paid-for editorial content unfair in all circumstances, with no deception test to argue about.