Patient Education Handouts: Plain Language That Patients Actually Use

The average American reads at a 7th-grade level, but most patient education materials are written at a 10th-grade level or higher. Low health literacy is directly associated with medication errors, hospital readmissions, and worse outcomes. AI can generate patient-facing materials at a specified reading level, in plain language, in multiple languages — closing a gap that has real clinical consequences.

Patient education prompt

1. 1Aim for 6th-grade Flesch-Kincaid — not because patients are unsophisticated, but because plain language is universally clearer
2. 2Warning signs section is the highest-stakes content — clinician must verify accuracy
3. 3Multi-language versions: request the AI to translate, then have a native speaker spot-check medical terms
4. 4Avoid passive voice — 'take this medication' not 'this medication should be taken'

Cultural competency in patient materials

Plain language is necessary but not sufficient. Patient education materials also need cultural relevance — examples that reflect the patient's likely context, dietary recommendations that align with cultural food practices, family involvement guidance appropriate to the patient's cultural norms. Ask the AI to generate culturally adapted versions and have a clinician from that background review them.

The big idea: patients who understand their instructions follow them. AI writes at the right level; clinicians verify the right content.

Most discharge instructions are written above the average reading level. AI can lower the grade level without losing the clinical meaning, then you verify.

What AI does well here

• Shorten sentences and replace jargon with everyday words.
• Add a 'when to call us' bulleted list at the top.
• Translate units (mg, mL) into household measures when safe.

What AI cannot do

• Decide which clinical details are safe to drop.
• Replace the clinician sign-off on the final document.
• Guarantee the translation captures cultural context.

Forty percent of adults can't read a standard discharge summary. AI can rewrite at 6th-grade level and translate to Spanish, Vietnamese, or Haitian Creole — but a single mistranslated dosage is a readmission.

What AI does well here

• Rewrite jargon-heavy instructions at a 6th-grade level using Flesch-Kincaid as a guardrail.
• Translate to common patient languages with cultural-context notes.
• Generate teach-back questions you can ask the patient before they leave.
• Produce a one-page visual schedule for medication timing.

What AI cannot do

• Catch a wrong medication or dose you typed in error.
• Know which idioms are safe in the target language — 'take with food' translates badly into some.
• Replace a live interpreter for high-stakes informed consent.

A 2025 survey found 35% of patients consulted an LLM before their visit. Many bring printed transcripts. Dismissing 'Dr. Google 2.0' damages the relationship. Engaging it builds one — if you know how.

What AI does well here

• Translate the LLM's overly broad differential into the 2-3 things you'll actually rule out.
• Build trust by validating which parts the model got right.
• Generate a take-home explanation of why you ordered or didn't order a test.
• Pre-empt LLM-driven anxiety with a written 'when to worry' card.

What AI cannot do

• Stop patients from consulting AI — and shouldn't try.
• Replace your physical exam, even if the AI 'did the same workup.'
• Diagnose from the LLM transcript alone.