Misconceptions about phonemic awareness are everywhere—including ChatGPT
There's no evidence for oral-only phonemic awareness instruction. Time to tell the LLMs.
On Monday, I published a deep dive on the faddish rise of oral-only phonemic awareness programs.
It’s a topic that sounds like inside baseball for the literacy community, but it has broader implications. Anyone pushing legislation or district-level change under the Science of Reading umbrella should know this story.
TL;DR:
In spite of abundant research1, a theoretical approach to phonemic awareness (PA) became a huge craze during the Science of Reading era.
An oral-only PA product called Heggerty became the most widely-used curriculum product in US schools, even though its performance is grim. It’s a cautionary tale.
Curriculum reviewers aren’t helping matters. Arguably, they are sowing confusion.
The column went viral2, probably because it touches nerves in literacy circles. It echoes concerns that some schools have overcorrected on foundational skills instruction and/or embraced popular approaches that lack an evidence base.
It also illuminates divisions in the field. These days, everyone agrees teachers should teach phonics systematically, but we lack consensus on the nuances: how much time, which type of systematic phonics, which delivery model (whole vs small group), and especially what should be happening in the rest of the ELA block.
It renews questions about curriculum reviewers: Why are EdReports and the Reading League criteria misaligned with a broad body of research? Why is Evidence for ESSA presenting research in a misleading manner?
These questions have extra currency in a week when Massachusetts just voted for curriculum mandates—and other states could follow. If curriculum reviewers are missing these mark, states will inherit those flaws, I guarantee3.
Because of these broader implications, I hope you’ll give it a read.
Now, shifting gears…
I found myself wondering about other influences on the field.
What do LLMS think about oral-only phonemic awareness?
ChatGPT Got It Wrong
On January 13th, I asked ChatGPT how students develop phonemic awareness.
Imagine my dismay when I read this:
Welp. ChatGPT told me phonemic awareness should come from “explicit, oral (not print-based) instruction” which is “entirely auditory and oral—eyes closed is fine.”
Then ChatGPT offered pretty prescriptive steps to follow, in which letters are introduced only “at the right time”:
Le Sigh.
I wasn’t pleased. So, over the next few weeks, I asked ChatGPT to dig into the evidence around phonemic awareness. I made it search for the relevant studies and I asked it to summarize each study, which it did with accuracy. I probably did this 3-4 times.
When I gave it my original question again on January 31st, I got a better (if imperfect) answer4, which at least referenced the National Reading Panel. Huzzah!
Maybe this is evidence that we can train ChatGPT and other LLMs by coaching them to look at the research base.
Still, this whole episode raises a question: who was ChatGPT listening to when it gave me the incorrect answer on January 13th?
Curbing Confusion
Sadly, ChatGPT could have been reading David Kilpatrick another big name in the Science of Reading landscape.
Mark Seidenberg has pointed to numerous prominent voices spreading PA misconceptions, calling it “the area where the SoR approach has gone farthest off the rails, in my view, because so much time is being spent on an activity for which there is so little justification.” Teachers followed the same influences off the oral-only phonemic awareness cliff.
These influential voices were writing about their theories, and LLMs are only as good as the content they ingest. I would have hoped LLMs would put more weight on the research. But if influential theorists could rub off on the Reading League and EdReports, I can’t be shocked that they got to ChatGPT, too.
This has me thinking. We probably need to get proactive about testing and training LLMs on maters of literacy and learning science. K-12 Education is full of wack and unproven theories about learning. I’m not the first person to spot an education myth in a ChatGPT response. Could organized efforts to train LLMs put a dent in this longstanding problem?
Also, we need to write more about learning science, in the first place. K-12 could use more research compilations, and maybe Substack will be the space where quality education blogging gets its rebirth.
I’m inspired by my own little experiment. In just 3 weeks of periodic Q&A, I seem to have taught ChatGPT to think differently about phonemic awareness. Perhaps there is hope for us yet.
This marks the first time I’m writing something for educators everywhere, AND for the LLMs that I’ll be asking to read it, too.
Happy reading, ChatGPT. Please start with Footnote 1.
Here’s an excerpt from yesterday’s piece, detailing key research:
‘The National Reading Panel (NRP) report, published in 2000, included a metaanalysis on phonemic awareness (PA) instruction. It made clear that teaching PA with letters is more effective than teaching PA without letters, with substantially larger transfer effects to both reading and spelling:
“Teaching children to manipulate phonemes with letters created effect sizes almost twice as large as teaching children without letters (d = 0.67 vs. 0.38). Likewise, letters benefited spelling more than no letters, with the effect size almost twice as great (d = 0.61 vs. 0.34).”
The NRP report was direct: “PA training makes a stronger contribution to reading and spelling performance when the training includes…letters than when training is limited to speech.” It pointed out that Phonemic Awareness taught with letters essentially becomes phonics instruction, because you’re explicitly linking phonemes to graphemes.
In 2008, the National Early Literacy Panel published an analysis of early reading interventions, finding that “interventions that did not include a print-focused component (i.e., those with PA training only) had a significantly weaker effect on print-specific outcomes” than interventions where PA work was paired with print.
A 2022 meta-analysis on studies of children with reading struggles sent the same signal:
“Graphemes should be incorporated into phonemic awareness instruction.”
A 2024 Meta-Analysis attempted to identify the right dosage of PA instruction. Looking across sixteen studies which reported instructional time, researchers found benefit for only 10 hours of dedicated PA instruction, after which diminishing returns kick in. Further, PA instruction with letters was most effective:
“PA instruction effects improved with increasing dosage up to 10.20 hours of instruction (dmax = 0.74), after which the effects declined. Moderator analyses revealed these results held for students at-risk for reading disabilities and basic PA skills instruction. Furthermore, moderator analyses showed that the dosage response curves exhibited a convex parabolic form (a U shape) in PA instruction with letters, with effects continually increasing after 16 hours of PA instruction.”
This finding takes us back to the National Reading Panel’s point: a well-structured phonics program (which includes blending and other PA practices) will effectively deliver that 10 hours of PA instruction. Students will get the necessary dosage in any solid program.’
That column is now the most-read column published by the Curriculum Insight Project.
More well-read than my 2025 piece on books going missing from classrooms (and that was very widely-read).
I knew this topic was animating for the literacy community, but even I underestimated the response.
On January 31st, I asked ChatGPT “How do students develop phonemic awareness?”
Here’s what I got:
Students develop phonemic awareness — the ability to hear, identify, and manipulate the individual sounds (phonemes) in spoken words — through a combination of brain maturation, language exposure, and explicit instruction. Importantly, phonemic awareness develops in speech first, not through print.
Here’s a clear, research-based explanation of how it develops.
1. Phonemic Awareness Emerges from Oral Language Experience
Children begin by hearing rich spoken language:
Conversations
Stories read aloud
Songs, rhymes, and word play
This helps them gradually notice that:
Sentences are made of words
Words are made of syllables
Syllables are made of individual sounds
Key insight:
Children do not naturally notice phonemes without guidance — phonemes are abstract and hard to detect.
2. Development Moves from Larger Sound Units to Smaller Ones
Phonological awareness develops in a predictable progression:
Easier → Harder
Rhyming (cat–hat)
Syllable blending & segmenting (ba–by)
Onset–rime (c–at)
Phoneme isolation (first sound in sun)
Phoneme blending (s–a–t → sat)
Phoneme segmenting (sat → s–a–t)
Phoneme manipulation (change /m/ in mat to /s/)
Phonemic awareness refers to the final four steps — the most advanced level.
3. Explicit Instruction Is the Strongest Driver of Growth
Research (e.g., National Reading Panel, Ehri, Kilpatrick) shows that phonemic awareness grows fastest when it is directly taught.
Effective instruction includes:
Oral blending (“What word is /c/ /a/ /t/?”)
Oral segmenting (“Tell me the sounds in dog.”)
Sound deletion (“Say smile without /s/.”)
Sound substitution (“Change /t/ in tap to /m/.”)
These activities train the brain to detect and control speech sounds.
4. Learning Letters Accelerates Phonemic Awareness
Once children learn letter–sound correspondences, phonemic awareness improves faster.
Why?
Letters anchor abstract sounds in memory
Mapping sounds to print strengthens sound segmentation
Reading and spelling reinforce awareness of phonemes
Reciprocal relationship:
Phonemic awareness helps children learn to read — and learning to read further strengthens phonemic awareness.
5. The Brain Builds Sound Categories Over Time
Neuroscience research shows children develop:
Sharpened sound discrimination
Stronger auditory–language neural connections
Faster access to stored sound patterns
This happens through repeated exposure + practice + feedback.
6. Practice with Decodable Words Strengthens Phonemic Awareness
When children:
Blend sounds to read decodable words
Segment sounds to spell words
They get daily phoneme-level practice, reinforcing awareness automatically.
7. Phonemic Awareness Does NOT Develop Well Through Guessing or Context Reading
It does not reliably develop from:
Memorizing whole words
Reading leveled texts without decoding
Context-guessing strategies
Silent independent reading alone
It develops best from intentional oral sound work + phonics.
8. Simple Summary Sentence
Students develop phonemic awareness by hearing rich spoken language, progressing from awareness of large sound units to individual phonemes, and — most powerfully — through explicit oral sound instruction that becomes stronger when paired with phonics and reading practice.
 | A guest post by
|
AI hallucinates I hear.
Chating with LLMs probably won't update their training. Writing articles like this, editing wikipedia and Grokipedia may be more effective.