How Peabody levels its books, and why our easiest books were too hard

Every book in Peabody's library sits on a reading level, and the level comes from rules in code, not from a person's guess or an AI's opinion. This post lays out all of those rules, the published formulas behind them with links to where each one comes from, and an honest account of a mistake: our "level 0" books, meant for a child who can't read yet, were too hard for a first grader at home. Here's why that happened and what we're changing. Peabody (peabody.kids) is free through the end of 2026.

The short version

• Our easiest books had short sentences but not easy words. "The Enormous Turnip" at level 0 has only five words on a page, but those words include grandpa, turnip, pulls and mouse. Short is not the same as readable.
• Our "can it be sounded out?" check was set for the end of first grade, not the first week of reading. It counted a word as readable if a child who had finished all of first-grade phonics could sound it out, and it counted all 220 Dolch sight words as known.
• The bottom of the ladder was too coarse. Three levels (0, 1 and 2) covered what school guided-reading systems spread over about ten.
• The fix: a 19-step ladder (levels 0 to 18). The first three levels are true decodable books, where every word can be sounded out from a small, named set of letter-sounds plus a few one-letter or very common words that we teach by name. Today's level 0 books move up to levels 3 and 4. Everything above shifts up to make room.
• We're not adding finer levels above second grade. Up there, the formulas can't honestly tell half a grade apart, so a finer label would be false precision.

How Peabody puts a book on a level today

Every story, whether written by hand for the library or by Peabody for one child, goes through the same code. It costs nothing to run and gives the same answer every time. It measures two kinds of thing.

1. For level 2 and up: published readability formulas

From level 2 upward the level comes from Flesch–Kincaid grade level, the most widely used US formula, which combines average sentence length with syllables per word. It was developed for the US Navy in 1975 (Kincaid et al., 1975).

Beside it, as a check on vocabulary, Peabody counts the share of words that are not on the Dale–Chall list of about 3,000 words most fourth graders know (Dale–Chall formula). Each level has a ceiling on that share.

Grown-ups can also see five more published formulas for any story. Each is shown only where it was built to work, and they are never averaged together:

Peabody also resamples a story's sentences 200 times to get a range, because on one short story any of these formulas can wander by about a grade. A book labelled 3.1 and one labelled 3.4 are, for practical purposes, the same difficulty. The longer explanation is in what "reading level" really means.

2. For the earliest levels: things you can count

Readability formulas don't work for first books. A page that says "I see a cat." scores below kindergarten on Flesch–Kincaid, and so does "Grandpa pulls." Both are short; only one is something a brand-new reader can read. So for levels 0 and 1, Peabody uses counts instead:

• Words per sentence and words per page.
• Sentences per page.
• Decodable share: what fraction of the words a child can sound out, letter by letter, using phonics patterns they've been taught. Peabody checks this against each word's real pronunciation from the Carnegie Mellon pronouncing dictionary, so it can tell bead from head and gave from have, which spelling alone can't.
• Sight words: words on the Dolch lists count as known.

The order of the phonics patterns comes from UFLI Foundations, the University of Florida's K–2 sequence: short vowels, then blends (st, fr), then two letters with one sound (sh, ch), then silent e, endings, two-part words, vowels with r, and vowel teams.

The rules for each level today

What went wrong at the bottom

We measured all 15 books on level 0 and all 43 on level 1 against what a child who is truly just starting knows: the five short vowels, single consonants, and a handful of words like a, I and the.

On level 0, the share of words a true beginner can sound out ranged from 34% to 97%. Only two books, Where Is the Cat? and Mop Can Hop, cleared 90%. Half were under 65%. A child reading at under 90% is guessing at one word in every ten or more, which is frustration, not practice.

On level 1 the range was 29% to 93%, and the books ran from 46 words (I See a Boy) to 200 (The Three Little Pigs), with up to 87 different words in one book.

Three things in our own rules let this happen:

1. The phonics check assumed the end of first grade. For levels 0 and 1 it allowed every pattern up to vowel teams (rain, boat, night). That's most of a year of phonics. Turnip, seed and goat all passed.
2. Every Dolch word counted as known. The Dolch lists run to 220 words through third grade, including laugh, together and because. A first-week reader knows a handful.
3. Nothing capped the number of different words. A book with five words on a page can still use 37 different words, which is 37 things a beginner has to work out.

There's also a gap one step up. Level 1 books top out around 10 words a page; level 2 books start at about 20 and run to 500 words. Nothing sits in between.

What the reading research says a first book should be

• Children start by learning letter-sounds and blending them, not by memorising whole words by shape (Ehri's phases of word reading; the US Institute of Education Sciences practice guide on foundational reading skills).
• A first book should use only the sounds a child has been taught. The classic example is the first Bob Books set: four letters (m, a, t, s) and one sight word, and a book of a dozen pages.
• Decodable and non-decodable books both help, and the two together seem to help most. A 2023 review of reading-intervention studies for kindergarten to third grade (Pugh, Kearns and Hiebert, summarised in Hiebert's 2024 presentation) found similar effects for each, and larger ones for a mix. That's why Peabody keeps read-aloud stories at every level, while the books a child reads alone start fully decodable.
• Guessing from the picture is a habit to avoid. The pictures on Peabody's first pages confirm a word after the child sounds it out; they aren't there to guess from (Shanahan on three-cueing).

How other leveling systems handle the bottom

Every serious system treats the first year of reading differently from everything after it:

• Guided reading levels (Fountas & Pinnell's A to Z) are set by trained people looking at ten features of a book, including how much the words repeat and how well the pictures support them (F&P Text Level Ladder). Commonly shared charts put about four levels in kindergarten (A–D) and about six more in first grade (D–J). Peabody doesn't produce guided reading levels.
• Lexile measures, from MetaMetrics, give the earliest texts a "BR" (beginning reader) code, and since 2017 add early-reading indicators for K–2 texts, including how repetitive and patterned a text is (MetaMetrics on beginning readers). Peabody doesn't measure Lexile and its levels aren't Lexile levels.
• ATOS, the formula behind Accelerated Reader, scans the whole text for sentence length, word length and word difficulty (overview).
• The Common Core's text-complexity model says any number should be read alongside qualitative features and the reader and task (the three-factor model).
• Decodability research measures how well a text matches the phonics lessons a child has had, often called lesson-to-text match (Hiebert, 2024).

The lesson from all of them: for the first year, count what the child has been taught, not how long the sentences are. That's the part we're fixing.

The new ladder: 19 levels, finest at the bottom

The "roughly" column uses commonly shared guided-reading charts for comparison only. Peabody doesn't produce guided reading levels.

So a kindergartner moves through levels 0, 1 and 2, a first grader through 3 to 6 or 7, and from second grade up it's about one level a year, as it is now.

How the new first levels are measured

• Each of levels 0 to 2 has its own named set of letter-sounds, not "everything up to the end of first grade." Level 0's set is eight letters.
• Each level has its own short list of taught words, in teaching order, instead of the whole Dolch list. Level 0 has two: a and I.
• Each level caps the number of different words in a book, and on levels 0 to 2 every word has to come back: three times on level 0, twice on levels 1 and 2.
• A book's level is the lowest one whose rules it passes. Nothing is labelled by hand.

Why not finer levels all the way up?

We looked hard at this, because it's appealing: four or five gently different levels in every grade. At the bottom it works, because the differences are things you can count exactly. One more letter-sound is a real, measurable step.

Above second grade it doesn't. The difference between two books there is mostly vocabulary and background knowledge, and the formulas that measure it have an error of about a grade on a short story. Splitting a grade into four would put the same book on different levels depending on which few sentences you changed. Instead, Peabody keeps one level per grade up there and sorts the books inside each level from easiest to hardest, so a child still gets small steps without a label that claims more precision than it has.

What changes for your child

• If your child can't read yet, they'll start at level 0 with eight letters and books where every word works.
• Children already placed moved to the new number for the same books, except level 1 readers, who moved to level 3 (pattern books), the old level 1's place. A grown-up can move a child to any level in Settings (there is a new first stop, "First books", for the decodables).
• Read-aloud stays at every level. A child on level 0 can still have The Enormous Turnip read to them. They just won't be asked to read it alone yet.

The new levels went live on 27 October 2026. The thirty books for levels 0 to 2 and eight bridge books are in the free library and on the decodable readers shelf, and every older book has its new level. What was level 4, for example, is now level 9. Stories Peabody makes for a child at levels 0 to 2 are written from a closed list of words, the ones that level has taught. The code builds the list, checks every page against it, sends a book back once with the words it got wrong, and if it still misses, gives the child a library book at that level instead of showing a wrong one.

Questions parents ask

What is a decodable book?

A book where nearly every word can be sounded out using letter-sounds the child has already been taught, plus a few common words taught by name. Peabody's levels 0 to 2 are decodable books, and each says which letter-sounds it uses.

Why did Peabody's easiest books feel too hard?

Their sentences were short, but many of their words used phonics patterns a beginner hasn't met yet, like turnip or mouse. Our check for sounding words out was set for the end of first grade instead of the start.

How does Peabody decide a book's level?

With code. For the earliest levels it counts letter-sounds, taught words, different words and words per page. From the bridge level up it uses the Flesch–Kincaid formula plus a vocabulary check. Every rule is in this post.

Is a Peabody level the same as a guided reading or Lexile level?

No. Peabody doesn't measure Lexile or produce guided reading levels. The comparisons in this post are approximate, for orientation only.