See your banner through
the viewer's eye

Predicted attention map, fixation order and a prioritized fix spec — for static ad creatives, before you pay for impressions.

No sign-up · processed in memory, never stored

Live audit

What you get

Attention heatmap

Where the eye lands in the first seconds — predicted by our GazeNet engine, validated on SALICON human-gaze data (AUC 0.948).

Scanpath + zones

Simulated fixation order and attention share for offer, logo, CTA, faces and focus object.

Fix Queue

Ranked problems with measured numbers — each with an action and expected attention gain.

Fix Prompt

An annotated original, a designer brief and a ready prompt for image generators. The creative itself is never altered.

A/B & batch

Compare two variants or audit a whole campaign at once — CSV export for agency workflows.

Benchmark

Percentile against 9,003 real display ads in 8 IAB formats — same-format comparison.

Measured vs predicted vs heuristic

measured

Computed directly from pixels — contrast, text count, readability at feed size.

predicted

Attention-model output — heatmap, fixation order, zone attention.

heuristic

Expert rules — triggers, memory trace, ad-blindness flags. Directions, not guarantees.

Honest limits

  • Attention is not results. Visibility scores predict where the eye goes — not CTR, conversions or sales. Verify with real A/B tests.
  • The model estimates visual attention — not emotions, thoughts or neurotransmitters.
  • Benchmark percentiles compare to our corpus — not the whole market.

Benchmark corpus

9,003real display ads
(AdImageNet, MIT)
8IAB formats,
full corpus
49median Ad-Score
of the corpus
AUC 0.948GazeNet vs SALICON
human gaze