Readable by Law

Three states grade insurance policies with a formula

Most plain-language rules ask for clear writing and leave it there. Insurance law in these three states sets a number. A policy form filed in Florida or New York must score at least 45 on the Flesch Reading Ease test. Texas sets 40 for life, annuity and health forms. Higher means easier.

The laws also say how to count, and the counting differs from ordinary Flesch tools:

• A semicolon or a colon ends a sentence, just like a period.
• A contraction, a hyphenated word, or a run of numbers and letters counts as one word.
• Syllables come from "an accepted dictionary".
• Headings, captions, tables and the insurer's name and address don't count.

These rules apply to policy forms. They don't apply to statutes, consumer guides or study material. So every comparison below asks one question: if this text were a policy form, would it pass? Nothing here says any document breaks a law.

We built a scorer that follows the statutes' counting rules, with syllables from the CMU Pronouncing Dictionary, and scored every document with it. Next to each score we report the ordinary Flesch score that this site's readability explainer gives the same text.

3 of the 5 readability provisions pass their own test

We started with the rules themselves, scored by the state's own method against the state's own bar.

Florida's rule scores 51.8 and clears its own 45. New York's doesn't: § 3102 scores 39.1, with longer sentences (20.9 words against Florida's 17.6) and longer words. The provisions that would fail are N.Y. Ins. Law § 3102 and Tex. Ins. Code § 1501.260. Texas's two rules, the ones that actually set its score, both pass.

Statutes, guides and lessons, state by state

Each mark below is one set of text scored as a whole, the way a regulator scores a whole policy form. The dashed line is the state's bar.

The law behind the licensing exam fails the test in every state

To sell life or health insurance you pass a state exam, and much of it tests the state's insurance code. We scored every section of the insurance code that one exam-prep provider lists as the source material for its lessons: 2,471 sections and 736K words. Sections of 150 words or more (966 of them) were also scored one by one. 150 words is the cut-off the Texas rule itself uses: it doesn't score a rider shorter than that.

New York's insurance law is the hardest to read: a pooled score of 16.0, with 37.2 words per sentence even after every semicolon counts as a full stop. Not one of New York's 84 longer sections would pass New York's own test. Texas does best, at 36.2, partly because its code is cut into many short sections. Even there, 74% of the longer sections fall below the bar.

The regulators' own plain-English guides don't all pass, and none of New York's do

Each regulator publishes guides that explain life insurance, annuities and health insurance to consumers. These are the regulator writing plainly in its own voice, for the public, with no lawyer's drafting to blame. We scored three per state.

Texas writes the way its rule asks: a pooled score of 60.5 and 13.3 words per sentence. New York's three guides score 41.5, 35.1 and 42.8. Florida's life and annuity guides clear 45, and its 13,000-word health guide, at 38.9, does not.

The exam-prep lessons clear the bar in all three states, narrowly in Texas

The fourth set is the 290 lessons MilesToPass publishes for the Florida, New York and Texas life and health exams: 184K words, scored as they appear on the public lesson pages on 2026-10-15.

Before scoring, we defined "meets the bar" in a state as both a pooled score at or above the bar and at least half of the lessons at or above it. We predicted the lessons would fall short in at least two states. They didn't.

Texas is the close call. The pooled score is 41.6 against a bar of 40, and the share of lessons at the bar is 55% with an interval that runs from 40% to 73%. In Texas the regulator's guides are far easier than the lessons. In Florida and New York the lessons score above the guides.

Some lessons list one of these consumer guides among their sources, so lessons and guides are not fully independent. Leaving those lessons out changes the pooled scores by less than a point: Florida 46.9, New York 46.7, Texas 41.1. The hardest lessons are about regulation itself, for example Policy or contract replacement duties (20.7) and Group health insurance and COBRA (26.0).

Passing a readability formula is not the same as being plain, and we don't claim it is. A formula counts sentence and word length and nothing else. See how readability formulas fail, and the section below on what this one misses.

One rule, three ways: the grace period

Every life policy has a grace period, the days after a missed premium when the policy stays in force. Here is the first paragraph on it in each state's statutes, its consumer guide and the lessons. Each score is for the whole paragraph.

We matched 23 concepts in Florida and 20 in New York this way: the first paragraph that uses a glossary term in each of the three sources. On average the lesson paragraph scores 34.7 points above the statute paragraph in Florida [23.7, 47.0] and 30.7 in New York [14.3, 50.6]. The guide paragraph sits between them.

These pairs are matched by word, not by meaning, and a single paragraph is a small sample for a formula. Some matches use the word in another sense: a statute paragraph about a "replacement license" turned up for replacement. Texas had too few concepts in all three sources (9, against a threshold of 20 we set in advance) and is left out.

What a law that counts semicolons is really measuring

We expected the semicolon rule to rescue the statutes, since legal drafting strings clauses together with semicolons. It doesn't come close. On its own the rule adds 3.6 points to the average statute section, and the statutes are 17.8 to 29.0 points short in Florida and New York. Most statutes already put each clause on its own line, and a line break ends a sentence under any counting.

The syllable count matters about as much. The statutes say to use "an accepted dictionary". Most software, including this site's, estimates syllables from spelling. We registered a rule in advance: if the dictionary moved any set of text by more than 2 points on average, the dictionary would be the headline. It moved statute sections by 3.0 points on average and the guides by 2.5, on average downwards, so it is. With the site's own counter, the Texas lessons would score about 1.5 points higher. That is enough to flip a close call, from software that never reads a word.

The formula can also be fooled. One Florida section, § 624.488, is a list of other section numbers. Every number counts as a one-syllable word and every semicolon ends a sentence, so it scores 91, "very easy". Nobody who reads it would agree.

What the formula can't see shows up in the site's other diagnostics:

Statutes use far more nominalizations (verbs turned into nouns) than either the guides or the lessons. The formula only sees that as longer words. This site's plain language checker flags each one, which a formula can't do.

What we expected, and what happened: 5 of 15 predictions held

We wrote these down and committed them before any document was scored. Most were wrong. The guides scored lower than we expected, the lessons higher, and the semicolon rule mattered less.

How this was measured

Sources. Statutes come from each legislature's own site: Florida's Online Sunshine, the Texas Legislature's statute host, and the New York Senate's law PDFs. The two Texas rules come from Cornell's Legal Information Institute, because the Texas register's viewer returns no rule text to an automated request. The guides come from Florida's Department of Financial Services, the Texas Department of Insurance and New York's Department of Financial Services. The Texas guides were saved from a browser, because the department's site turns away research crawlers, and each saved copy was checked against a hash of the page. The lessons are the public MilesToPass objective pages. Everything was collected on 2026-10-15.

Why MilesToPass. It's the exam-prep site we run, and the only one whose full content we could analyze. If grading other sites proves feasible, we'll extend the study to them.

Which statutes. "The law behind the exam" means the insurance-code sections MilesToPass lists as source material for its lessons, a list it built before this study and for another purpose. The state exam outlines name topics, not sections, so some list had to be chosen, and we say whose it is. Other codes on that list (labor, health, tax), regulations and federal law are left out.

What counts as text. The same rules apply to every source. Headings, captions, tables, subsection numbers like "(a)", history notes, navigation and in-text legal citations are removed. Defined terms and medical terms stay in, even though the laws let an insurer exclude them on certification, because we scored text as published.

Pre-registration. The corpus, the rules, the measures, the bar, what "meets the bar" means, the analysis and the predictions were all committed before scoring. After scoring we fixed two extraction faults, both found by reading the longest sentences. A checklist in a Florida guide had underline blanks that stopped its questions from ending sentences, and New York's guide pages had bold run-in headings glued to the first sentence of a paragraph. Both fixes raised the guides' scores, by 0.1 and 0.2 points pooled, and neither changed a conclusion. The breakdown table in the section above was also added after scoring and is labelled that way.

What this cannot tell you

• It is not a compliance finding. The rules apply to policy forms. Nothing scored here is a policy form, and nothing here says any document breaks a law.
• A formula is not a reader. Nobody was tested on understanding. A score measures sentence and word length, and the section above shows how far that is from measuring plainness.
• The statute list was chosen by an exam-prep provider, and the lesson scores come from the provider that chose it.
• PDF and web extraction is imperfect. A few bold run-in headings in the Florida guide PDFs remain attached to the sentence after them, and a two-column table of agent credentials reads as one long sentence. We found these by reading the output and left them, because each affects a handful of sentences out of thousands.
• Some intervals are rough. Florida's statute sections come from six chapter pages, so its interval clusters on six units. Three guides per state is not a sample, so lessons and guides are compared without intervals.

Check our work

Every number above is computed from these files. Each row names its source, so you can fetch the original text and score it yourself, with any tool.

• Every scored document — 2,775 rows: statute sections, rules, guides and lessons, with words, sentences, syllables (estimated and from the dictionary), both scores and the site's diagnostics
• Pooled scores — one row per state and set of text
• Side-by-side passages — the concept matches, with scores

Published October 15, 2026. Statutes and rules are public law. Excerpts from the consumer guides are quoted briefly for comment. Corrections are welcome. Get in touch.