How the caption readability check works
What it measures
Flesch reading ease and Flesch-Kincaid grade, average and longest sentence length, a passive voice count, adverb share, emoji per 100 words, ALL CAPS words, hashtag share, link count and a four-point hook checklist for the first 10 words.
It then ranks the problems by how much they cost a reader and lists the top three as plain instructions.
Where the rules come from
Both readability formulas are the published Flesch (1948) and Flesch-Kincaid (1975) formulas. The thresholds used for advice, such as 25 words for a long sentence, 4% adverbs and 15 emoji per 100 words, are editing rules of thumb, not network rules, and nothing here is an official platform limit.
Limitations
The formulas were designed for English prose. Captions in other languages get a score, but it is not meaningful. Short captions of one or two sentences swing the score a lot, so read the suggestions more than the number.
Frequently asked questions
What is a good Flesch reading ease score for a caption?
Aim for 60 or more, which the Flesch scale calls plain English. Short sentences and short words push the score up. Below 50 a caption starts to read like a report.
How is the score calculated?
Reading ease is 206.835 minus 1.015 times the average words per sentence, minus 84.6 times the average syllables per word. The grade level uses the Flesch-Kincaid formula. Syllables are estimated from vowel groups, so a rare word can be off by one.
Why are my hashtags and emoji ignored in the score?
They are not words in a sentence, so counting them would distort the average. They are measured separately as a hashtag ratio and an emoji density per 100 words.
How does the passive voice count work?
It is a heuristic. It flags a form of "to be" followed by a word ending in "ed" or a common irregular participle, such as "was scheduled" or "is written". It can miss unusual phrasing and rarely flags an active sentence by mistake.
What makes a strong caption hook?
The first 10 words decide whether people tap "more". A question, a specific number, the word "you" and one concrete noun each help, and the checklist shows which of the four your opening has.
Does it use AI to judge my caption?
No. Every number comes from a fixed formula or a word-pattern rule that runs in your browser. The same caption always gets the same result, and nothing is sent anywhere.