Guide
A/B testing thumbnails and social images, without the traffic
Pickdot is a free tool to compare two images before you post them. It measures what can be measured — where attention lands, how big your text really renders, whether it has enough contrast — and gives an honest, hedged pick for your audience.
What is A/B testing?
An A/B test shows two versions of something to different slices of the same audience and measures which one does better — more clicks, more watch time, more sign-ups. Version A is the control, version B the challenger. Because both run at the same time for the same kind of people, the difference in results comes from the change, not from luck or timing.
It is the most reliable way to settle "which one is better?" The catch is sample size: to tell a real difference from noise you usually need thousands of impressions per version. A small difference needs even more.
Why creators rarely get to run one
A single YouTube upload, Instagram post or ad often doesn't get enough views to finish a fair test, and paid testing tools cost more than one post is worth. So most thumbnail decisions end up made by gut feeling, a friend's opinion or a poll in a group chat. Pickdot is built to replace the group chat — not the A/B test.
How Pickdot compares two images
- Attention map. A saliency model estimates where the eye lands first, shown side by side for A and B.
- Text size at feed scale. Your headline measured in real pixels at the size the platform shows it — for example a 320px-wide YouTube tile. Under 14px is hard to read.
- Contrast. WCAG contrast between your text and what's behind it. Under 3:1 fails for large text.
- Faces. Whether a face is present, how much of the frame it fills and its expression.
- Clutter and colour. Edge density, palette size, saturation and clipped highlights.
A short, readable rule set turns those numbers into a pick, weighted by the audience you describe once: phone-first viewers weigh legibility more; "close-up faces do well" makes face size count double.
How accurate is it?
The measurements are facts. The verdict is a nudge. Published research on AI predictions of A/B tests finds the direction is right about 70% of the time and the size of the difference is usually overstated. So Pickdot never promises a percentage lift or a score out of 100. It says "more likely", states its confidence in words, and says "too close to call" when the gap is inside noise.
When to run a real A/B test
When you have the traffic, test for real. YouTube Studio's Test & Compare rotates up to three thumbnails on a live video; Meta Ads Manager and Google Ads run split tests on creatives. Use Pickdot first to drop options that fail on legibility or contrast, then test the two strongest. Log what happened in Pickdot afterwards and you build your own hit rate over time.
Frequently asked
- Is it free?
- Yes. No account, no upload, no limit.
- Are my images uploaded?
- No. Everything runs in your browser; the images never leave your device.
- Which platforms does it support?
- YouTube thumbnails, Instagram posts and stories, LinkedIn images and display ads.
- Can it compare more than two images?
- Not yet. Two is the decision most people actually face.