CASPA AI Certification Decoded — What It Actually Bans

CASPA AI Certification Decoded — What It Actually Bans

Short answer: The CASPA Applicant Certification statement — which every PA school applicant must affirmatively sign before submission — is the strictest published rule on substantive AI use in any centralized US health-professions application service. For the 2026-27 cycle, PAEA softened the prior cycle's absolute ban: grammar and spelling correction with AI tools is now explicitly permitted, while substantive drafting, rephrasing, and idea generation remain prohibited. CASPA is still materially stricter than AMCAS, AACOMAS, or TMDSAS on substantive AI use, but it is no longer the absolute blanket ban the 2025-26 language described.

This article is a clause-by-clause read of the verbatim certification statement, a side-by-side comparison with the three other major health-professions application services, and a practical guide for what each clause actually means when you sit down to write your CASPA personal statement.

The verbatim CASPA Applicant Certification (2026-27 cycle)

From the 2026-27 PAEA Admissions Suite of Products Policies & Procedures document (October 2025), pages 15-16, under the heading "Guidance on Generative AI in CASPA User Submissions":

"I certify that all written content in my CASPA application is my own work. This includes, but is not limited to, personal statements, essays, and descriptions of work, educational activities, and events. While consulting personal and professional resources, including artificial intelligence (AI) tools, for non-substantive changes such as correction of spelling and grammar is acceptable, my final submission accurately represents my own writing, work and experiences."

Followed immediately by: "It is my responsibility to review and abide by the institutional policies regarding the use of artificial intelligence tools to any program to which I apply."

There are three distinct things this paragraph does. Each one has implications for what you can and cannot do.

What changed from the 2025-26 certification

The 2025-26 CASPA Policies & Procedures document (October 2024) used materially different — and stricter — language:

"I certify that all written passages within my CASPA application, including but not limited to, personal statements, essays, and descriptions of work and education activities and events, are my own work, and have not been written or modified, in whole or part, by any other person or any generative artificial intelligence platform, technology, system, or process, including but not limited to Chat GPT (collectively, 'Generative AI'). I am strictly prohibited from using Generative AI to create, write and/or modify any content, in whole or part, submitted in CASPA and/or provided to PA programs on my behalf through any means of communication."

Two things are different in 2026-27:

1. The "written or modified, in whole or part" structure is gone. The 25-26 language explicitly prohibited modification "in whole or part." That phrase does not appear in 26-27. In its place is a positive allowance: "non-substantive changes such as correction of spelling and grammar is acceptable."
2. "Chat GPT" is gone from the certification text. The 25-26 cert named Chat GPT as an exemplar. The 26-27 cert does not name any specific AI product in the certification clause itself (though the preamble still lists ChatGPT, Bing, Gemini, and Llama as examples of the broader category).

Structurally, the 26-27 language moves from a rule ("you cannot use AI") to a standard ("your final submission must accurately represent your own writing, work, and experiences"). Grammar-level AI use is now explicitly permitted. Substantive drafting or rephrasing is still prohibited.

Clause 1: The certification is affirmative and unconditional

"I certify that all written content in my CASPA application is my own work."

This is an affirmative attestation. You are not just promising to follow a rule — you are certifying as a matter of fact that every written passage in the application is your own work. If that turns out not to be true, the issue is not just rule violation. It's certification fraud, which can be treated as misrepresentation under the broader CASPA application user agreement.

Three things to notice in the scope:

• "all written content" — not just the personal statement. The Life Experiences essay, the new Situational Decision-Making Question (sometimes called the "AI and Technology essay"), the work and activity descriptions, every prompt response — all of it.
• "including but not limited to" — open-ended scope. If you write it and it goes into CASPA, it's covered.
• "descriptions of work, educational activities, and events" — this is the kicker that catches the 600-character experience descriptions that many applicants treat as low-stakes "fill in the blanks." They are not. Every PCE description, every HCE description, every volunteer line is covered by the same certification.

Practical implication: When you sit down to write the 600-character description of your medical scribe job, the same rule applies as when you write your 5,000-character personal statement. The substantive writing must be your own, even if you use a grammar checker for the final cleanup.

Clause 2: The "non-substantive changes" carve-out — where the line now falls

"While consulting personal and professional resources, including artificial intelligence (AI) tools, for non-substantive changes such as correction of spelling and grammar is acceptable, my final submission accurately represents my own writing, work and experiences."

This is the clause that changed most from 25-26. The old language prohibited "written or modified, in whole or part" — a blanket modification ban. The new language explicitly allows AI tools for "non-substantive changes such as correction of spelling and grammar." Two verbs are doing the work: consulting (for resources, including AI) and non-substantive (describing what kind of change is OK).

What is now explicitly allowed:

• Running your draft through a spelling and grammar checker that uses AI under the hood (Grammarly's base product, Microsoft Editor, Google Docs' grammar suggestions)
• Fixing typos an AI tool flags
• Accepting a comma or subject-verb agreement correction the AI suggests

What is still prohibited (by the "final submission accurately represents my own writing" standard):

• Asking ChatGPT to rewrite a paragraph and pasting the rewrite
• Asking Claude to "make this opening stronger" and incorporating the new opening
• Asking any LLM to rephrase your draft for clarity in a way that changes meaning, voice, or word choice beyond mechanical correction
• Drafting content from scratch with an AI and editing the output
• "Humanizer" tools that disguise AI-generated content (these are substantive and disqualifying)

The line is substantive vs. non-substantive. If the AI's contribution changes meaning, argument, voice, structure, or word choice beyond mechanical correction, it crosses the line. If the AI's contribution is mechanical correction of spelling or grammar, it is now explicitly permitted.

The pragmatic question used to be where the line actually falls. In 25-26, the strictest possible reading of the cert caught a Grammarly comma fix, and only University of Washington MEDEX Northwest had explicitly carved out "non-substantive editing — such as spelling or grammar correction" as permitted. For 26-27, PAEA has adopted essentially the MEDEX framing at the central CASPA level: the Grammarly comma fix is now explicitly in the green zone. MEDEX Northwest was ahead of CASPA on this for a year; we covered the MEDEX Northwest AI policy in detail and the language they use to describe the carve-out is nearly identical to the language PAEA has now adopted.

For everyone applying to PA programs, the safest practice is still to write the substance yourself: do not paste your draft into ChatGPT to "review" it, because once you read the AI's output, the cognitive influence is hard to undo even if you don't copy any of the words verbatim. But the spell-check fear is gone. Run your final pass through a grammar tool with confidence.

The ghost-writing implication

The 26-27 certification's standard — "my final submission accurately represents my own writing, work and experiences" — applies to human help as well as AI help. The 25-26 cert spelled this out by naming "any other person or any generative artificial intelligence platform" in the same breath; the 26-27 standard collapses both into the same authenticity test.

The practical implication: the standard that prohibits substantive AI drafting also prohibits having a human write your essay for you. A consultant can review your draft, suggest changes, push back on weak sections, and help you brainstorm — but they cannot write your draft, and they cannot give you sentences or paragraphs to paste in. The "your own writing" standard applies equally to human and AI assistance.

This matters because some applicants assume the loosened AI rule means "use a human consultant heavily" is fine. It is not. The certification is about your writing, not about whether a human or a robot helped.

Clause 3: PAEA's detection commitment

The 25-26 certification ended with a clause reserving PAEA and PA Programs' right to run AI detection tools on submitted content. The 26-27 P&P drops that reservation language from the certification itself and replaces it with a more applicant-protective formal commitment in the body of the policy:

"Given the significant risk presented by current AI detection software as of the date this policy is being issued, PAEA will not initiate a CASPA investigation where the sole basis for the investigation request is that an AI detection tool tagged a personal statement or evaluation submitted in CASPA."

This is the position PAEA had been advocating informally since its September 2023 advisory (What Your Program Should Know About AI and Admissions). In October 2025 it became binding policy. PAEA will not open an investigation against you on detector output alone, citing the high false-positive and false-negative rates of current detection tools.

PAEA's preferred verification mechanism is in-person essay-writing during interviews — comparing your live writing voice to your submitted essay's voice. This is more reliable than algorithmic detection and is the method most PA programs that test rigorously will use.

But individual programs may go further than PAEA central. MEDEX Northwest is the one PA program in our 20-program survey that explicitly says it will use AI detection tools — out of step now with formally ratified PAEA policy, not just informal advisory. Other programs may use them silently.

For more on the false-positive problem with AI detection — and what to do if your authentic, AI-free essay is incorrectly flagged — see our deep dive on flagxiety. PAEA-funded research now exists with measured false-positive rates on PA application essays specifically — read our breakdown of the PAEA AI detection study and what those numbers mean for CASPA applicants.

How CASPA compares to AMCAS, AACOMAS, and TMDSAS

This is where the difference matters most. PA applicants are operating under the strictest centralized rule of any health-professions applicant in the US. If you are dual-applying to MD programs through AMCAS, you cannot just follow AMCAS's rules — you have to follow CASPA's stricter rules for your CASPA submission. For the undergraduate parallel — the certification language every Common App applicant signs — see our clause-by-clause Common App AI fraud policy decoded.

The four-system comparison in detail is in our Can You Use ChatGPT for Your Medical School Application? article. The takeaway for PA applicants is unambiguous: CASPA's rule is materially stricter than the rule for any other health-professions application. If your friends applying to MD school are using ChatGPT to brainstorm, they are following the AMCAS rule. You cannot follow that rule. CASPA's certification language is explicitly more restrictive.

Practical guidance for PA applicants

If you are about to write your CASPA application, here is what actually complies with the certification.

What's clearly allowed

• Spell check and grammar correction (including AI-powered) — Microsoft Word's spell check, Google Docs' grammar suggestions, Grammarly's base product, and any browser spell checker. As of the 2026-27 CASPA cycle, these are explicitly allowed by the certification language ("non-substantive changes such as correction of spelling and grammar is acceptable").
• Reading example essays from blogs, books, and forums — looking at how others have written CASPA essays is research, not AI use.
• Asking a human PA for feedback on your draft — getting feedback is allowed; what is not allowed is the human writing portions of the draft for you.
• Discussing your draft with a paid admissions consultant — same as above. They can advise; they cannot ghost-write.
• Writing successive drafts by hand or in a plain text editor — this should be the baseline.

What's clearly banned

• Asking ChatGPT, Claude, Gemini, or any other LLM to write your essay — fully banned.
• Asking an LLM to write a paragraph and editing it — still banned. Modification counts.
• Asking an LLM to rephrase your draft for clarity — banned.
• Asking an LLM "is my essay any good" and incorporating any of the suggestions — banned, because the suggestions are AI-generated content.
• Accepting sentence-level AI rewrites from a tool like Grammarly Premium that replace your phrasing with the AI's phrasing — banned, because these are substantive changes (word choice, structure, voice). The 26-27 CASPA cert allows the "non-substantive" tier of Grammarly — spelling, obvious grammar corrections — but not the "rewrite for clarity" tier.
• Using a "humanizer" tool to disguise AI-generated content — banned twice over.

What's contested

• AI-powered autocomplete in your text editor that suggests sentence completions as you type. The cleanest answer: disable it for the duration of your CASPA writing. Most modern editors let you turn AI features off; do that.
• Translation tools for non-native English speakers writing in English. CASPA has not addressed this directly. The safest reading: write in your strongest language first, translate yourself, and use AI translation only as a dictionary lookup, not as a sentence-level translator.
• Voice-to-text dictation tools that may use AI under the hood. Probably allowed if the tool is converting your spoken words to text without modification, but the line is fuzzy and depends on the specific tool.

What to do if you accidentally used AI early in the process

Some applicants brainstorm with ChatGPT early in the writing process, then rewrite their essay from scratch by hand. Is that allowed?

The strict reading of the CASPA certification says no — the AI's output influenced your final draft, and "modify" is broad. The pragmatic reading says it depends on whether any specific words, sentences, or structural choices from the AI made it into your final draft.

The safest practice is not to brainstorm with AI at all. If you already did, the next-safest practice is to set the AI conversation aside, wait at least a week, and then write from scratch without referring back to it. The further your final draft is from what the AI suggested, the less risk. But there is no zero-risk path back from "I asked ChatGPT for help" — the cleanest move is not to start.

What this all means for the new CASPA AI essay

There is a deep irony in CASPA adding a new essay about AI in healthcare to the 2026-2027 cycle — PAEA's official name for it is the Situational Decision-Making Question — right alongside a certification that prohibits substantive AI use in drafting. Applicants are being asked to write thoughtfully about AI as a clinical tool, while CASPA's certification prohibits using AI to draft or rephrase that essay (though grammar checkers are now explicitly OK). We have a complete guide to the new CASPA AI and Technology essay with the verbatim prompt and seven worked angles. The discipline of writing about AI without AI as a co-author is part of the test.

For a broader view of how individual PA programs are interpreting the CASPA central rule, see our survey of 20 PA programs' AI policies. The short version: 18 are silent, 1 has informal advisory language, and 1 (MEDEX Northwest) has its own published rules. The CASPA central certification is the rule that actually binds you. For an always-current list of every PA program's published AI position in one place, see our PA program AI policies aggregator for 2026.

---

Related Reading

Medical school essays hub: Medical School Essays — The Complete Guide to AMCAS, AACOMAS, CASPA & TMDSAS — every medical school essay guide on the site, organized by application system, topic, and applicant profile.

The CASPA + AI policy cluster:

• PA School AI Policies 2026 — Why 18 of 20 Programs Are Silent — survey of 20 PA programs' published policies
• MEDEX Northwest AI Policy — The One PA Program with Rules — the only program with its own published rules
• CASPA AI and Technology Essay 2026-2027 — Prompt + 7 Angles — the new AI essay this cycle
• How Readers Evaluate the CASPA AI and Technology Essay — methodology-grade explanation of what PA admissions readers actually notice across six evaluation dimensions
• Can You Use ChatGPT for Your Medical School Application? AMCAS, AACOMAS, CASPA, TMDSAS Compared — the four-system comparison
• What Is Flagxiety? The AI Detection Anxiety Reshaping How Students Write — false positives and the detection landscape
• Medical School AI Policies 2026: AMCAS Rules & 60+ Schools — our AI policy hub for medical admissions

The CASPA writing cluster:

• Sample CASPA Essay Analysis: How 40+ Applicants Got Into Top PA Programs
• The CASPA Life Experiences Essay: What the New Prompt Actually Asks
• CASPA Personal Statement Topics to Avoid
• CASPA Experience Descriptions: 10 Examples of Strong 600-Character Narratives