What is an attention map?

How is an attention map built?

Two different calculations ship under the same name. Check which one your tool runs before interpreting a single warm region, because the two produce different maps from the same traffic.

Cursor-based (move map, hover map). Samples pointer position several times a second and accumulates dwell time per region. Warm areas are where the cursor rested. This is the classic implementation and the weaker one.

Viewport-time-based. Measures how long each region of the page remained inside the viewport, weighted by scroll speed. Warm areas are where visitors slowed down or stopped. It makes no claim about the cursor, and time in viewport is directly observable rather than inferred, which makes it the defensible variant.

Vendors differ on which calculation they ship, on what they call it, and on whether they offer one at all, and the label alone will not tell you which of the two you are looking at. Read your tool's own documentation for the calculation behind the map before you interpret a single warm region.

How accurate is cursor position as a proxy for gaze?

Weakly, and inconsistently across people, which is the core problem.

Three distinct behaviors exist in any traffic sample:

• Cursor followers, who move the pointer along as they read. For these visitors the correlation is real.
• Cursor parkers, who move the pointer somewhere neutral and leave it there for the entire session. Their attention map contribution is a hot spot on a random patch of whitespace.
• Cursor ignorers, who never touch the mouse while reading and only move it to act.

An aggregate blends all three with no way to separate them afterward, which is the defect that matters. The map does not measure attention. It measures cursor behavior, from which attention can be inferred for an unknown share of your visitors. On touch devices the concept degrades further, since there is no persistent pointer at all and the map is built from taps and scroll pauses.

When is an attention map still worth reading?

Three situations where it earns its place, all of them comparative rather than absolute.

Comparing two regions on the same page. Absolute values are unreliable, and the difference between two sections in one sample carries the same bias on both sides, so it partially cancels. The ordering is more trustworthy than either number.

Detecting scroll-through zones. Long stretches of cold on the viewport-time variant identify content people scrolled past without slowing. Useful for finding sections earning their vertical space and sections that are not.

Corroborating a finding from another source. When a scroll map shows a cliff and the attention map shows cold immediately above it, the two together are stronger than either alone.

How does it differ from real eye tracking?

Eye tracking uses hardware, an infrared camera calibrated per participant, to record where the eye actually fixated and for how long. It is a measurement. Attention maps are an inference from a different signal entirely.

They trade off in opposite directions. Eye tracking is accurate on a small recruited sample that may not represent your buyers. Attention maps are inaccurate on your entire real audience. Neither substitutes for the other, and published eye tracking research is usually a better input to a layout decision than your own attention map, because it was measured rather than inferred.

What should you do instead?

For the question "did visitors engage with this section", better instruments exist.

1. Element visibility events. An IntersectionObserver firing when a specific element enters the viewport, optionally with a dwell threshold. Directly answers whether a section was seen, with no inference.
2. Scroll maps. More reliable for reach, and cheaper to interpret. See scroll map.
3. Session recordings. Show scroll speed and pauses in sequence, which is what an attention map is trying to approximate.
4. A test. If the question is whether a section earns its position, moving it and measuring the outcome answers definitively.

Attention maps sit at the bottom of the behavioral evidence stack. Read them last, weight them least, and never move a section on their evidence alone.

How do Webflow interactions affect attention maps?

Scroll-triggered interactions, which Webflow sites use heavily, change what is on screen as visitors move down the page. The attention map renders over a single static screenshot, so regions animated into view get mapped against a page state that never matched what any visitor saw. On an interaction-heavy Webflow page the map is unusable rather than merely weak, and the check takes seconds: open the page with animations disabled and compare it to the screenshot the tool captured.

Related terms

Website heatmap · Scroll map · Click map · Scroll depth · Session recording

Service: Conversion Rate Optimization.

FAQ

Is an attention map the same as a move map?

Usually yes. Move map and hover map are the older names for the cursor-based version. Some tools use "attention map" for a viewport-time calculation that does not involve the cursor at all, so check the vendor's definition before interpreting.

How many sessions does an attention map need?

More than any other heatmap type, around 500 minimum and 1,000 or more before drawing conclusions, because cursor data is noisier than click or scroll data. Even at high volume, treat the output as directional.

Can attention maps work on mobile?

Not in the cursor sense, since touch devices have no persistent pointer. Mobile attention maps are built from taps and scroll pauses, which makes them closer to a slow-scroll map than an attention measurement.

Do attention maps prove people read your content?

No. They suggest where cursors rested or where scrolling slowed. Reading is not observable from either signal. Content comprehension requires user testing or comprehension questions, not heatmaps.