The 2026-27 CASPA AI Rules: What's Allowed, What Isn't, and How to Stay on the Right Side

Starting with the 2026-27 CASPA cycle, applicants will write a mandatory essay on AI and technology in healthcare, and CASPA has explicitly stated which AI uses are permitted and which are prohibited. The rule is clearer than the panic about it. This post walks through what the rule actually says, what counts as allowed, what does not, and how to write a strong AI essay without overthinking the technology.

Two changes landed for the 2026-27 CASPA cycle, and both are worth understanding before you start writing your application materials.

First, CASPA added a mandatory essay on artificial intelligence and technology in healthcare. Every 2026-27 applicant will write this in addition to the standard personal statement. The prompt is open enough that you have real room to write something specific to your experience.

Second, CASPA explicitly clarified that applicants may use AI tools for non-substantive edits - grammar and spelling. While continuing to prohibit AI for generating content. PAEA Also publicly stated it will not initiate investigations based solely on AI-detector flags, because detectors are unreliable.

The line is simpler than the conversation about it has been. If you wrote it, AI assistance for polish is allowed. If AI wrote it, that is academic dishonesty. The hard part is making sure your process keeps you cleanly on the allowed side, and that your final essay reads like writing you actually did.

What the CASPA rule actually says

Drawing directly from the CASPA 2026-27 cycle guidance (with the language paraphrased for readability):

• Generative AI for content creation is prohibited. You cannot ask ChatGPT, Claude, Gemini, or any equivalent tool to write your personal statement or your AI essay for you, then submit the output as your own. This is treated as an academic-honesty violation.
• AI tools for non-substantive edits are permitted. Grammar checks, spelling correction, and similar polish on writing you produced is explicitly allowed.
• PAEA will not investigate based solely on AI-detector results. Detectors flag false positives at a real rate - sometimes flagging text written by humans whose style happens to look statistically similar to model output. PAEA stated publicly it will not act on detector flags alone.

The line CASPA drew is the line that matters: You wrote it, polish is fine. You did not write it, that is not your work. Tools that Review Your writing against a rubric - like the Dashboard’s statement checker - sit squarely inside the permitted zone, because they evaluate writing you produced rather than producing writing for you.

Why this rule exists

The fall 2025 application cycle made the AI question impossible to ignore. The 2026-27 cycle landed it inside the application itself, on the timeline walked through in the CASPA 2026-27 application timeline. Programs reported a noticeable rise in essays that read flat, polished in a generic way, and structurally similar across hundreds of applicants. Some essays clearly used AI; others were just well-edited but human. Detectors could not reliably tell the difference. Programs were left adjudicating cases on inconsistent evidence.

The 2026-27 rule is CASPA's response to that mess. By drawing a clear line between content generation (prohibited) and polish (allowed), and by signaling that detector flags alone will not trigger investigation, CASPA is trying to get applicants and programs back to a stable working norm.

The new mandatory AI essay is a related move. By forcing every applicant to write directly about AI in healthcare - a topic that is unavoidable at this point in the technology's adoption, and one programs are also asking about directly in interviews. Programs get to see how each applicant thinks about the technology. Strong answers will read as thoughtful and grounded; thin answers will read as either alarmist or starry-eyed. The essay is a signal of judgment as much as a signal of knowledge.

How to think about the AI essay prompt

The exact wording will be in your CASPA application, but the spirit of the prompt is: how do you think about the role of AI and technology in clinical practice, and how does that thinking shape the kind of PA you want to be?

A measured answer is what programs want, and "measured" here does not mean noncommittal. It means an answer that takes a real position grounded in something you have actually observed. The two failure modes I see most often:

• Over-praising the technology. Essays that frame AI as the savior of healthcare read as naive. Programs know the literature on diagnostic-AI miscalibration, on bias in training data, on the ways automation reshapes clinical judgment, both for better and for worse. An applicant who has not engaged with the downsides comes across as not having engaged with the topic.
• Over-warning against the technology. Essays that frame AI as a coming threat read as out of touch. PAs already work with AI-assisted tools in radiology, in EHR documentation, in triage. The technology is not the future; it is the present. An applicant who frames it as a hypothetical risk has not been paying attention.

The middle path is the strongest: a specific clinical or shadowing observation, a clear position on what the observation taught you, and a brief acknowledgment of what is uncertain. That structure reads as the kind of clinical reasoning programs want to see in their incoming class.

A working structure for the AI essay

If you want a scaffold to start from:

1. Open with a specific observation - a moment in clinical, shadowing, or PCE work where you saw AI or technology used (or saw its absence). Two to three sentences.
2. Name what you took from it. What did the moment make you think about? Avoid platitudes. Be specific to your role, your patient, your environment.
3. Take a position. Where do you land on the role of this technology in PA practice? It is fine if your position is nuanced - most strong positions are. It is not fine if it is unspecified.
4. Acknowledge what is uncertain. Genuine uncertainty (not performed humility) about how the technology evolves, where its limits sit, what the patient impact will be over time.
5. Close with how this shapes the kind of PA you want to be. Tie it back to clinical judgment, to patient care, to your work with the team.

This is not a formula to copy verbatim. It is a structure for thinking about what to say. The voice that makes the essay land is yours.

What "non-substantive edits" actually covers

If you want to know exactly where the line is, here is my reading of it:

• Allowed: Grammar correction, spelling correction, suggestions for cleaner phrasing where the underlying idea was already in your draft, identifying passive constructions you wrote and want to fix, flagging awkward sentences for rewording.
• Allowed: Rubric-based review tools that evaluate your draft against admissions criteria. The Dashboard’s statement checker Is in this category - it scores your draft, flags PA-specific failure modes, and returns a graded PDF, but it does not write any of your essay for you.
• Edge case (probably allowed but be careful): Asking AI to rephrase a single sentence you wrote. This sits closer to the line. Doing it once or twice is fine. Doing it thirty times in an essay is functionally generation.
• Prohibited: Drafting any portion of the personal statement with AI. Generating an opener and editing it does not make it yours. The same applies to body paragraphs, conclusions, and the AI essay itself.
• Prohibited: Asking AI for "a personal statement about a [insert your background] applying to PA school" and using any of the output, even reworded.

The cleanest mental test: if you removed every contribution made by an AI tool from your essay, would the essay still be intact? If the answer is yes - because the AI only fixed grammar. You are fine. If the answer is no - because the AI wrote sentences you used. You are over the line.

What about AI detectors

This is the part of the rule applicants worry about most, and the part with the least teeth. PAEA stated publicly that AI-detector flags alone will not trigger an investigation, because detectors have a known false-positive rate that makes them unreliable as the sole basis for an academic-honesty case.

What this means in practice:

• If you wrote your essay and an AI detector flags it as machine-generated, you are not in trouble. The detector is wrong, and PAEA has acknowledged that detectors are wrong often enough to not act on them.
• If your essay reads structurally like AI output - flat tone, generic vocabulary, predictable parallelism, no specific personal voice. That is a different problem. It will not get you flagged, but it will get you read past, which is functionally worse than being flagged.
• The defense against both flagging and being read past is the same: write in your actual voice, with specific scenes and specific reasoning, and let the polish be polish rather than re-architecture.

What I tell applicants who ask me about this rule

Three things, in order:

1. Write the draft cold. No AI assistance, no rubric tool, no review service. Just you, the prompt, and the page. Whatever comes out is the foundation. You can fix it later.
2. Polish in passes. Run a grammar pass. Read it out loud. Fix awkward sentences yourself. If you want a rubric review, that is the moment for it - after the draft exists in your voice. The statement checker fits naturally here.
3. Do not run the final draft through an AI rewriter. Even with the best intent, a rewriter shifts the writing toward generic, polished, AI-adjacent prose. Programs read that and the essay loses you. Resist the temptation to over-polish. A clean draft in your voice beats a polished draft in nobody's.

What I didn't realize until I started reviewing essays

I have read hundreds of PA personal statements between my own application cycles, the friends I have helped, and the calibration runs I do for the statement checker. The ones that land are almost never the most polished. They are the ones with specific scenes, real voice, and a narrative the reader can follow. Polish is not the differentiator I thought it was when I was writing my own.

Programs are reading hundreds of essays. The applicants who get noticed are the ones who sound like a person, not the ones who sound like a perfect candidate. AI is good at making writing sound like a perfect candidate. That is exactly the wrong direction to push your essay.

Write it yourself. Use AI for grammar. Use a rubric tool for structural feedback. Submit something that sounds like you on a clear day. The rule allows that, the readers reward it, and the cycle gets easier.

Frequently asked questions

Can I use AI to write my CASPA personal statement?

No. CASPA explicitly prohibits using generative AI to create content for application materials. Submitting AI-generated writing as your own is treated as an academic-honesty violation.

Can I use AI to check grammar and spelling on my CASPA essay?

Yes. CASPA 2026-27 explicitly permits AI tools for non-substantive edits like grammar and spelling. The rule covers polish on writing you produced.

Will CASPA's AI detector flag my essay if I used a review tool?

PAEA has publicly stated it will not initiate investigations based solely on AI-detector flags, because detectors are unreliable. A rubric-based review tool that evaluates your draft (rather than generating content) is inside the permitted zone.

What is the new mandatory CASPA AI essay for 2026-27?

Starting with the 2026-27 cycle, every CASPA applicant writes a mandatory essay on AI and technology in healthcare, separate from the personal statement. Programs use it to assess how applicants think about the role of AI in clinical practice.

How should I structure my CASPA AI essay?

Open with a specific observation from clinical, shadowing, or PCE work; name what the moment taught you; take a clear position on the role of AI in PA practice; acknowledge genuine uncertainty; close with how this shapes the kind of PA you want to be. Avoid both over-praising and over-warning against the technology.