How The Ad Ranker scores copy

The grade is deterministic. You can reproduce it by hand, which is why it is published here rather than described as a proprietary model.

What it measures

Four things, computed from the text of your advert and nothing else.

• Reading ease. A Flesch Reading Ease score, calculated from your average sentence length and average syllables per word. Copy in the middle of the range scores best. Prose harder than its audience loses points, and so does prose so simple it carries no information.
• Benefit language. The count of words that name a gain for the reader. It is capped, so stacking them stops helping — four benefit words do not score twice what two do.
• Urgency language. The count of words that give a reason to act now, capped the same way. Past the cap, urgency reads as pressure and the cap is where we stop rewarding it.
• Length. A penalty for copy too short to say anything, and a smaller penalty for copy long enough that the end will not be read.

What it does not measure

It does not see your offer, your audience, your bid, your creative, your landing page, or whether your claim is true. Those are what decide performance. No tool reading only your copy can predict a click-through rate or a cost per acquisition, and one that claims to is guessing.

It also does not check your copy against any advertising platform's policies or against advertising law in your jurisdiction. A high score is not an indication that an advert will be approved or that it is lawful. That stays with you.

Why publish it

A score you cannot interrogate is not useful. If the grade says your copy is weak, you are entitled to know which of the four components moved and by how much, so you can disagree with it. A rubric kept secret can be changed quietly; one published can only be changed openly.

On benchmark figures

Our industry pages previously showed an average score, a readability band and a “top trigger” per industry. Those numbers were placeholders carried from a draft and were not measured from anything, so they have been removed. When figures return they will carry the sample they were measured on, the method, and the date.