Do Plagiarism and AI-Detection Tools discriminate against people with disabilities?

Credit pix4free.org

Plagiarism and AI-detection tools have now become part of the underlying structure of both the education and employment systems. They determine which assignments are flagged, which applications are sieved, and which individuals need to explain their writing methods.

These systems are often viewed as neutral instruments that maintain integrity and efficiency. However, for many people with disabilities, they form a barrier to entry rather than act as a protection against misconduct.

This is not about dishonesty. It’s about accessibility, unintentional discrimination, and design decisions.

From Safeguards to Gatekeeping

Originally, plagiarism checkers were only used in specific circumstances, as a way of verifying suspicions. Nowadays, they are used on a larger scale:

· Automatically scanning student papers

· Highlighting similarity rates with no specific information

· Recognizing “AI-generated” content

· Pre-screening employment applications before being reviewed by a human

In both educational and professional contexts, these tools have become gatekeepers to opportunities.

When they make errors, the results are not theoretical. They are intimate, stressful, and sometimes even life-altering.

Disabled Authors can Bear the Brunt

Many authors with disabilities depend on assistive and supportive technology in order to write. These include devices and tools like:

· Speech-to-text and dictation

· Word prediction and completion of phrases

· Grammar and sentence support

· Built-in AI writing tools within accessibility devices

The issue isn’t their existence; it’s the fact that these detection systems are beginning to treat the outcomes of these tools as suspicious.

A text that is coherent, consistent, or linguistically “sophisticated” can be identified as problematic, not because it’s dishonest, but because it was created using supporting technology.

To put it differently, accessibility is being read as misconduct. As these examples demonstrate.

NPR told the story of Shaker Heights junior Zi Shi, whose first language is Mandarin. He explained that his writing style can sometimes look like AI “because of the repetition of words I use. I feel like it’s because of how limited my vocabulary is.”

https://www.npr.org/2025/12/16/nx-s1-5492397/ai-schools-teachers-students

Shi went on to explain how an English class assignment was flagged by an AI detector as possibly AI-generated. His teacher suggested that his use of “Grammarly” may have triggered the detection software. Grammarly uses AI to correct grammar and, if prompted, generate text, and is widely used by people with Dyslexia to accommodate the challenges they face in producing high-quality text.

The issue was highlighted by Amanda Hirsch in 2024. She noted that Neurodiverse students are also more likely to be falsely flagged for AI-generated writing, and research has revealed biases against linguistic patterns and dialects in AI detectors. The disproportionate rate of false positives against already marginalised student groups could have chilling effects in education and beyond. For instance, it could foster an atmosphere of distrust between faculty and students, discourage academic participation and engagement, and undermine the perception of fairness in assessment and disciplinary processes.”

https://citl.news.niu.edu/2024/12/12/ai-detectors-an-ethical-minefield/#:~:text=Neurodiverse%20students%20are%20also%20more,Title%20VI%2C%20the%20ADA.

Furthermore, Timothy Markley from Kaltman Law added that “Students with Autism Students on the autism spectrum often exhibit structured, literal, or repetitive writing styles that may resemble AI-generated text. The case of Moira Olmsted, detailed by Davalos and Yin (2024), highlights this risk. Olmsted, a college student with autism, was falsely accused of cheating based solely on AI detector output. Despite explaining her communication style, which is shaped by her neurodivergence, she received a zero and a disciplinary warning. Such cases underscore how AI detectors can criminalise disability-related differences in expression.”

https://www.kaltmanlaw.com/post/ai-detectors-academic-integrity-bias#:~:text=Disproportionate%20Impact%20on%20Marginalized%20Students,disability%2Drelated%20differences%20in%20expression.

Uniform Language Is Not Dishonest

Accessible writing typically uses:

· Simple language

· Frameworks for sentences

· Structures that are relatively easy to predict

· Phrases that are easy to reproduce and reduce cognitive effort

These methods are generally used in:

· Supporting individuals with autism

· Educating people with intellectual disabilities

· Cognitive and linguistic processing variations

Cheating systems are programmed to think that sameness means copying. However, in the realm of accessibility, similarity generally means utilizing supportive communication.

When these systems fail to grasp this distinction, disabled authors are more likely to be flagged, questioned, or punished.

The AI-Detection Intersection

As AI starts to be a part of typical tools, the line between “assistive technology” and “AI writing” is becoming increasingly blurred.

Many accessibility tools now:

· Use AI in an automatic mode

· Adjust language according to the user

· Predict or create text in order to minimize effort.

At the same time, AI detection tools are beginning to view any kind of AI mediation as suspicious.

The result is a harmful intersection: AI-detection very quickly becomes a ban against any reasonable modifications.

This isn’t some rare incident. This failure in design could have been anticipated.

The True Cost of False Positives

False positives don’t affect everybody in the same way.

Disabled authors might:

· Re-use phrases that work for them

· Show less diversity in style and content over an extended period of time

· Prioritize clarity over originality

These characteristics are not indicative of cheating. They are indicative of accessibility approaches.

However, within detection systems, these factors can lead to higher sameness scores, more scrutiny, and greater emotional and administrative effort.

Being accused of cheating is stressful for everyone. For disabled individuals, it often means:

· Re-disclosing a disability

· Proving that their allowances are legitimate

· Coping with appeals that are inaccessible

After some time, this will begin to have a chilling effect. People will avoid using devices that they are absolutely entitled to. They write less or they produce poorer outputs in order to avoid coming under suspicion. Some individuals may even stop writing altogether.

Employment Screening — Beyond Academic Plagiarism

The dangers faced in recruitment are even greater.

Plagiarism and AI-detection tools may very well be used to:

· Automatically filter job applications

· Reject candidates before they even have a chance to disclose

· Flag them as non-compliant without due process

There is no appeal, feedback, or any viable way to comprehend what went wrong.

For candidates with a disability, this might mean that they are systematically weeded out, not because they lack the requisite skills, but because the system has no way to recognize that they need access.

This Isn’t the Fault of Malicious Individuals

Most of the perils aren’t caused by malicious motives. They stem from factors such as:

· Systems that have been developed with no input from disabled users

· Models that are designed based on limited interpretations of what constitutes “normal” writing

· Use of automation where human judgment used to be the standard

That makes this a failure on the accessibility front, not a case of misuse.

And when it comes to structural problems, you need structural solutions along with a different approach to policy-making, rather than isolated quick fixes.

What Operational Platforms Should Look Like

Plagiarism and AI-detection platforms don’t have to work like they do now.

Operational systems:

· Should view assistive technology as being perfectly acceptable by default

· Need to separate assistance from detection of potential misconduct

· Should have a human review the situation before any adverse conclusions are drawn

· The results of the tests should be explainable, put in the right context, and not viewed as being the final word

· Ensure that reasonable adjustments and route for appeals are available

These systems should be developed and tested with individuals with disabilities.

Above everything else, they need to clearly state what they are designed to do and what they can’t do.