Can Generative AI Help Simplify Informed Consent Language?

Can AI Make a Research Consent Form Easier to Understand Without Changing What It Means?

You are preparing an informed consent form for a study involving university instructors. The document has already been reviewed for its research procedures, confidentiality arrangements, and participant rights.

However, several paragraphs contain lengthy sentences and technical expressions. The section on data processing refers to pseudonymization, authorized personnel, data retention, and possible withdrawal limitations.

You ask ChatGPT to rewrite the document in simpler language.

The revised version sounds much easier to read. Unfortunately, one sentence now states that all participant information will remain anonymous, even though the original document only promised that direct identifiers would be removed. Another sentence says participants can withdraw their data at any time, although the approved protocol specifies certain limitations after data aggregation.

The AI has improved the language but changed the meaning of important consent conditions.

Can generative AI genuinely help simplify informed consent materials, and how can researchers prevent readability improvements from becoming substantive changes to the information participants receive?

Yes, but Simplification Must Preserve the Approved Meaning

How to Use Generative AI to Simplify Informed Consent Responsibly

Why Does Plain Language Matter in Informed Consent?

Informed consent is intended to support an individual's understanding of a research study and their voluntary decision about participation.

A consent document may contain all the required information yet communicate it in language that prospective participants find difficult to understand.

For example, a statement about the circumstances under which identifiable data may be retained could be written using complex legal terminology and several embedded qualifications.

Researchers may be able to express the same information through shorter sentences and familiar vocabulary without removing the relevant conditions.

The United States Common Rule, 45 CFR 46.116, requires informed consent information in covered research to be presented in language understandable to the participant or legally authorized representative. It also emphasizes presenting information in a way that facilitates comprehension.

These requirements apply within the relevant regulatory framework. Other institutions and jurisdictions may establish their own consent standards, but understandable communication remains an important ethical consideration.

What Can Generative AI Actually Simplify?

Generative AI can suggest changes to several features of consent language without necessarily changing the underlying research procedure.

These changes may improve readability, but their effect on actual comprehension should not be assumed without appropriate evaluation.

Simplification Is Not the Same as Summarization

Researchers should distinguish rewriting information in clearer language from reducing the amount of information provided.

For example, simplifying a confidentiality statement may involve replacing technical expressions while retaining information about access, identifiability, and limits to confidentiality.

Summarizing the same statement as "Your information will be kept safe" may remove qualifications that participants need to understand.

Researchers should not assume that a shorter document necessarily provides better informed consent.

AI should be instructed to preserve material information rather than merely produce the briefest possible version.

Why AI May Change the Meaning While Improving the Wording

Generative AI produces language based on patterns and contextual instructions. It does not automatically distinguish stylistic changes from legally or ethically consequential changes.

For example, a model may replace "reasonable measures will be taken to protect confidentiality" with "your information will remain completely confidential."

The second statement is easier to read, but it introduces an absolute guarantee absent from the original.

Similarly, AI may replace "participants may request withdrawal of their identifiable information before data aggregation" with "participants may withdraw their data whenever they wish."

The revised sentence changes the scope of the withdrawal arrangement.

Researchers should examine whether the AI has strengthened promises, removed conditions, or introduced rights that the study cannot actually provide.

Which Consent Statements Require Particular Care?

Some parts of a consent form are especially sensitive to changes in wording because they describe rights, obligations, or limitations.

These elements should be checked against the approved protocol and applicable institutional requirements rather than evaluated solely for readability.

Can AI Replace Legal or Technical Terms With Everyday Words?

Often, but not every technical term has a simple substitute with exactly the same meaning.

For example, the terms anonymous, anonymized, pseudonymized, and confidential are not interchangeable.

A researcher might explain pseudonymization by stating that names will be replaced with participant codes and that the information connecting those codes to identities will be stored separately.

Such an explanation may be easier to understand than the technical term alone, provided it accurately describes the actual procedure.

However, replacing pseudonymization with anonymity would change the meaning.

Researchers should consider retaining an important technical term alongside a plain-language explanation when no concise substitute is sufficiently accurate.

How Should Researchers Simplify AI Processing Statements?

Consent materials involving generative AI may contain unfamiliar terminology about automated processing, external providers, and data retention.

For example, the original statement might say that de-identified excerpts will undergo automated natural language processing through an institutionally authorized service.

A clearer explanation might state that selected interview excerpts will be processed using an AI system to help researchers organize the responses.

However, the simplified version should retain material information about provider involvement, identifiability, retention, and safeguards where required.

Researchers should not remove these details merely because they make the statement longer.

The broader question of how much participants should be told about AI processing should be resolved before attempting to simplify the language.

Can AI Make Consent Language More Persuasive Than Informative?

Potentially. Generative AI may produce reassuring or promotional wording when asked to make a document more friendly or engaging.

For example, it might describe participation as an exciting opportunity that will contribute significantly to educational innovation.

Such language may create expectations about benefits that the study cannot establish.

Informed consent materials should communicate the research accurately without unnecessarily encouraging participation or minimizing potential risks.

Researchers should avoid asking AI to make consent language more convincing when the intended objective is clarity.

Improving comprehension and increasing recruitment are different goals, and they should not be confused.

What Reading Level Should Researchers Request?

Researchers may ask AI to produce language suitable for a particular reading level or participant population.

However, no single reading level is universally appropriate for every research study.

A study involving university instructors may require different explanations from one involving younger students or participants with limited familiarity with research terminology.

Researchers should consider literacy, language proficiency, familiarity with the topic, and the complexity of the information.

Specifying a target reading level may be useful as a drafting constraint, but it does not establish that participants will understand the document.

The related question of whether AI can check consent readability should be distinguished from evaluating actual comprehension.

Why Readability Scores Are Not Enough

A consent form may receive a favorable readability score while still containing unfamiliar concepts, ambiguous instructions, or misleading statements.

For example, a short sentence stating that data will be pseudonymized may be easy to read grammatically but difficult for participants unfamiliar with the term.

Conversely, a longer explanation may improve understanding by describing what will actually happen to identifying information.

Readability formulas generally focus on measurable features such as sentence length and word characteristics. They do not establish whether a document communicates its intended meaning accurately.

Researchers should therefore combine language review with substantive verification and, where appropriate, evaluation involving representatives of the intended participant population.

Can AI Simplify Consent Materials for Different Participant Groups?

Generative AI can suggest alternative explanations for participants with different educational backgrounds or levels of familiarity with research.

For example, a study involving both university instructors and undergraduate students may require explanations that differ in terminology while preserving the same material information.

However, researchers should avoid assuming that all members of a demographic group have identical literacy levels or communication preferences.

Adaptations should reflect the intended population and be evaluated for clarity and accuracy.

Where multiple versions are used, the research team should ensure that participants receive equivalent material information unless the study legitimately requires different procedures.

AI-generated simplification should not create unequal disclosure of risks or rights across participant groups.

How Should You Prompt AI to Simplify Consent Language Safely?

A useful prompt should explicitly distinguish language editing from substantive revision and require the model to identify possible changes in meaning.

The researcher should review the revised text against the original and verify that all substantive conditions remain accurate.

Where the consent materials contain confidential or sensitive information, researchers should use an appropriately authorized system or nonconfidential drafting material.

Why a Side-by-Side Comparison Is Important

Reviewing only the AI-generated version may make it difficult to notice information that has been omitted or changed.

A side-by-side comparison allows the researcher to examine the original and revised statements together.

The more accurate alternatives are illustrative and still require verification against the actual approved procedures.

The purpose of comparison is to detect substantive changes, not merely determine which sentence sounds more natural.

Can Researchers Simplify Already Approved Consent Forms?

Yes, researchers may identify opportunities to improve clarity after an initial ethics review. However, approved participant materials may be subject to institutional modification procedures.

For example, an institution may require revised consent wording to be submitted before it is used with additional participants.

Even changes intended only to improve readability may need review under the applicable procedures.

Researchers should not assume that a revision is administratively insignificant simply because the research purpose remains unchanged.

The appropriate process depends on institutional requirements and whether the changes affect substantive information.

Should Participants Test the Simplified Language?

Where appropriate, researchers can evaluate revised materials with representatives of the intended participant population.

For example, participants may be asked to explain in their own words what happens to their information, whether participation is voluntary, and when withdrawal is possible.

Such feedback may reveal misunderstandings that are not apparent from readability scores or expert review.

However, consultation should be conducted under appropriate institutional arrangements, particularly when it forms part of formal research activities.

A small comprehension exercise does not establish that every future participant will interpret the document identically.

It can nevertheless provide useful evidence for improving the materials.

Who Is Responsible for the Final Wording?

The research team remains responsible for ensuring that participant information accurately describes the approved study.

AI may suggest clearer language, but it cannot independently authorize changes to consent conditions or establish that a statement complies with applicable requirements.

Researchers should verify revisions concerning participant rights, risks, confidentiality, data processing, and withdrawal.

Where necessary, institutional ethics, legal, or data protection personnel should review the revised wording.

The responsibility for ensuring AI-simplified consent language remains accurate should not be transferred to the system that generated the revision.

Simplifying a Confidentiality Statement Without Promising Anonymity

The final wording is simpler without converting a limited confidentiality safeguard into an absolute promise.

This illustrates why AI-assisted simplification should be treated as a controlled revision process rather than automatic replacement of approved language.

Common Mistakes When Simplifying Informed Consent With AI

Use AI to Improve Clarity Without Changing Participants' Rights or Expectations

When using generative AI to simplify consent language, provide the original wording and clearly instruct the system to preserve substantive meaning.

Review the revision against the approved documents, particularly where it concerns rights, confidentiality, withdrawal, risks, and data processing.

Where appropriate, obtain feedback from representatives of the intended participant population to identify remaining misunderstandings.

The objective is to make the research easier to understand without making promises that the research cannot keep.

Before Using AI-Simplified Informed Consent Language

Frequently Asked Questions About Simplifying Informed Consent With AI

Consent Language Should Become Easier to Understand, Not Easier to Misinterpret

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.