Our story

Why We Built Thumbnail AB Test

A free, in-browser tool that scores your covers and titles before you publish, so you can shortlist the strongest candidate under fixed rules instead of relying only on a hunch.

You spend hours on a video and then pick its cover on a hunch. Real audience data only arrives after the video is published, when the launch is already underway.

That creates a practical gap. Creators need a way to compare ideas before publishing, but an offline tool cannot know how real viewers will respond. The useful job is narrower: apply the same rules to every option, show where each one is strong or weak, and make the shortlist easier to defend.

YouTube's own Test & Compare uses real behavior after publication. Thumbnail AB Test works one step earlier, comparing several concepts before publication with a heuristic preflight. Use the preflight to narrow the field, then use YouTube to validate the final options.

Two choices shaped the design. The core tool had to be free and usable without an account. Its scoring also had to run in the browser, so original cover files stay on your device during analysis. A beta account stores only reports you choose to save.

What we care about

The choices behind the tool

  1. Local-first by default

    Core scoring happens in your browser, and original cover files stay local during analysis. Account sync is only for saved reports and has no access to your YouTube channel.

  2. Honest about what the score is

    The preflight score applies fixed heuristics to the options in front of it. It does not observe real viewers, measure YouTube CTR, or claim to predict either one.

  3. Free is the starting point

    The core tool is free with no login. Creator costs $9 per month for linked revision decisions and clean reports; it uses the same score and 2-to-6 input range as Free. Studio team workflows remain research.

In plain terms

What it is, and what it is not

What it is

  • A pre-publish heuristic preflight that compares your covers and titles under the same fixed rules.
  • A 12-dimension breakdown: 5 visual dimensions and 7 title angles, all computed in your browser.
  • A deterministic 10,000-iteration preference model that converts combined scores into relative model share within your current set.
  • Free, with no login required for the core test. A beta account can sync reports you choose to save.

What it is not

  • Live audience research or a forecast of your real click-through rate. The model does not observe people, impressions, or clicks.
  • A guarantee that a high-scoring cover will perform. The score measures whether the cover does its job, not whether the video delivers.
  • A replacement for YouTube's Test & Compare. That runs after you publish on real behavior; this runs before publication on fixed heuristics. They are complementary.

See it on a real thumbnail

Run a free test on two of your own covers and titles. No login required, and every result shows a relative score and the first fix to consider.

Try the tool