PTCAS AI Policy — What the Certification Actually Bans

PTCAS AI Policy — What the Certification Actually Bans

Short answer: PTCAS does have an explicit AI rule for the personal statement, and it names a specific commercial chatbot as its example. It is not in the essay instructions you read while drafting. It sits in the PTCAS Certification and Authorization Statement on apta.org — the block of terms you accept at submit time. The rule has two separate halves: one about human help, where feedback on your drafts is the single named exception and editing by another person is not, and one about AI text generators "or other online sources," where no content from either may appear in your statement. Higher up the same page, in plain prose, APTA also expects applicants to "compose a personal essay without assistance from others" — that one carries no exception at all, and we take both sentences up below.

Every quote below was confirmed against the live pages on August 10, 2026.

The verbatim rule

From the PTCAS Certification and Authorization Statement on APTA's PTCAS Application Preparation page, retrieved August 10, 2026:

"I certify that my the [sic] content of my personal statement (essay) is original and composed in its entirety solely by me with no assistance from other individuals beyond providing feedback on my drafts. Further, I certify that I have not used any artificial intelligence (AI) text generators (e.g. [a named commercial chatbot]) or other online sources for the content of my personal statement."

Two notes on that quotation. The doubled "my the" is in the original; we have left it and marked it, because that is how the sentence reads on the page. And the certification gives one specific commercial chatbot as its worked example — we have replaced the brand with a bracketed description, which is our standing practice for product names. The category the sentence points at is AI text generators, whichever one you had in mind.

Three sentences earlier in the same block, APTA also has you certify:

"I certify that I have read and will abide by all program-specific instructions for my designated physical therapist programs."

Hold onto that one. It matters at the end.

Why you have probably never seen it

The page applicants actually open while writing is the PTCAS essay instructions in the Liaison help center. That page carries the required prompt, confirms the 4,500-character limit, and lists three requirements. Its entire authorship guidance is a single line:

"Use your own words: Plagiarizing any part of your essay is a violation of the code of conduct and may subject you to sanctions."

Retrieved August 10, 2026. Nothing on that page mentions artificial intelligence, generative tools, or text generators.

There is one more piece on the APTA side, and it is easy to scroll past. Just above the certification block, under Additional Applicant Expectations, APTA lists what applicants are expected to do. One item is a single flat line, retrieved August 10, 2026:

"Compose a personal essay without assistance from others."

No exception, no mention of feedback, no qualification of any kind. So APTA states the authorship expectation twice on one page in two different registers, and the two statements are not identical in scope. That is worth reconciling rather than picking whichever you prefer, and the next section does it.

So the pages split cleanly. The page you read while drafting says "use your own words," which nearly everyone reads as an ordinary plagiarism warning. The language that actually governs authorship lives on a different website, in a section about applicant responsibilities rather than about essays — once as a bullet in a long list, once as a term you accept at submit. If you drafted in June and first met the AI sentence in September at submit, nothing about your process was careless — the binding language was simply not where you were looking.

Our PTCAS essay guide for 2026-2027 works from that same essay page, because that is where the prompt and the limit live. The certification is a separate document with separate terms, and you are bound by both.

Clause 1: one named exception, and a sentence with none

"...no assistance from other individuals beyond providing feedback on my drafts."

This is the half applicants most often have backwards. Read it in two moves. The default is strict — "composed in its entirety solely by me." Then the certification adds exactly one named exception, and it is the word feedback.

Now set that beside the expectations bullet quoted above: "Compose a personal essay without assistance from others," and nothing after it. The two sentences are published by the same organisation on the same page, and they are not identical in scope. One states a general expectation with no exception attached; the other states the expectation and then names one. The certification is the text you actually attest to at submit, so it is the operative document — but nothing on the page tells you the bullet has been superseded, and the safest reading is the stricter one. If you are deciding how much help to accept, decide against the bullet and you cannot fall foul of either.

Under the certification's exception, a reader reacting to your draft is inside the line: a friend can tell you they got lost halfway down the first page, that they finished without understanding why you chose physical therapy over nursing, or that a line you were proud of landed flat on them. What neither sentence carves out is that reader rewriting the paragraph for you, supplying replacement sentences, or handing back a marked-up file whose new phrasing you paste into the field.

The practical test is simple: whose sentences end up in the box. If someone's comment changes what you decide to say and you then write it yourself, the words are yours. If someone's version of a sentence survives into the submitted text, they are not.

Worth noticing that this is a narrower permission than several neighbouring application services grant, and it is narrow in a direction most people do not expect. Applicants worry about the AI half of the rule and assume the human half is permissive. Here it is the human half that draws the tighter line.

Clause 2: "or other online sources" catches more than chatbots

The second sentence has two objects, and readers routinely see only the first:

• "any artificial intelligence (AI) text generators"
• "or other online sources for the content of my personal statement"

That second half is not decoration. It reaches content lifted from anywhere online — an example essay from a forum thread, a template's ready-made opening move, a sentence from a program's mission page repurposed as your own conviction. If the content came from an online source rather than from you, this clause names it directly.

Pause here if you started from a template, because old PTCAS templates carry a second problem on top of the sourcing one: most of them answer a prompt PTCAS no longer asks. Reading examples is fine and useful; the discipline is to read for structure, close the tab, and then draft. Our PTCAS essay examples for the current prompt are labelled constructed teaching examples for precisely this reason — they are invented compositions built to be analysed, not source text to adapt.

How this compares with CASPA

Applicants who apply to both DPT and PA programs are working under two different rules at once. Neither is simply "stricter"; they are strict along different axes.

The sharpest difference is the grammar question. CASPA's 2026-27 language explicitly permits AI tools for "non-substantive changes such as correction of spelling and grammar." PTCAS has no equivalent carve-out in either direction: its sentence bans AI text generators "for the content of" the statement and stops there, leaving the status of a mechanical spell-check unaddressed rather than resolved.

Our clause-by-clause read of the CASPA certification covers the substantive-AI axis in depth; PTCAS is the tighter of the two on human editing and the more explicit on non-AI online sources. If you are applying through more than one service, the workable approach is to hold a single strictest-reading standard across all of them rather than tracking a different line per portal. How the medical application services compare on AI rules lays out the wider set.

What the consequences clause actually says

Immediately after the AI sentence, in the same block:

"I understand that withholding information requested on the PTCAS application, or giving false information, may be grounds for denial of admission to a program participating in PTCAS or may be grounds for expulsion from the institution after I have been admitted."

Read it for scope rather than for tone. It is not bounded by the admissions cycle — "expulsion from the institution after I have been admitted" reaches forward past matriculation. And its trigger is "giving false information," which, once you have certified that your statement is original and unassisted beyond feedback, includes the certification itself. That statement is a signed assertion of fact about work you have already done, not a promise about how you will behave later.

That is what the sentence says. How any individual program would handle any individual case is entirely theirs to decide, and nothing here is a prediction about that.

What PT-specific research found

Physical therapy is one of a small number of fields with published, peer-reviewed research on AI-generated admissions essays in its own applicant pool — most verticals have none.

Hollman JH, Cloud-Biebl BA, Krause DA, Calley DQ. "Detecting Artificial Intelligence-Generated Personal Statements in Professional Physical Therapist Education Program Applications: A Lexical Analysis." Physical Therapy 104(4), April 2024, article pzae006 (PubMed 38243411).

The team collected 152 personal statements from DPT program applicants plus 20 machine-generated statements written to the same standardized prompt, then ran recurrence quantification analysis — a family of measures for how much a text repeats its own patterns. The machine-written statements scored higher on recurrence, determinism, mean line, and entropy. The strongest single discriminator was a determinism rate of 13.04%: statements above that rate were roughly eight times more likely to be machine-generated, at 70% sensitivity and 91.4% specificity.

Those two percentages cut in both directions, and it is worth sitting with each.

• 70% sensitivity means about three in every ten machine-written statements fell below the line and read as human.
• 91.4% specificity means about one in every twelve genuinely human statements landed above it.

The authors' own conclusion is the honest one: the method can distinguish AI-generated from human-generated text, but "it is not absolute," and they argue that admissions committees and the bodies that publish admissions guidance should reconsider how much weight the personal statement can carry.

Two things follow for you as a writer. First, no detection figure is proof about any single essay in either direction — a flag is a signal that invites a question, never a finding. Second, look at what the measures actually track: self-repetition and predictable pattern structure. The study did not test templated human writing, so this is inference rather than a result, but writing assembled from a template plausibly produces some of the same signature without any tool involved. The move away from that pattern is the same move that makes an essay worth reading — the specific clinic, the specific patient population, the specific constraint you actually ran into.

Program level: absence is not permission

Of the five DPT program admissions pages we checked (August 2026) — USC, Pittsburgh, Northwestern, Delaware, and Emory — none published a named AI or authorship policy for applicants.

That is a finding about five pages on one date. It is not a statement about DPT programs generally, and it is emphatically not permission. Two reasons it settles nothing:

1. You separately certify that you "have read and will abide by all program-specific instructions." A program can publish instructions inside a supplemental application, a portal page, or a program handbook that never appears on its public admissions page, and your certification reaches those.
2. The PTCAS certification binds you regardless of what any program does or does not publish. Silence at the program level does not loosen a rule set at the platform level.

If a program you are applying to has published something, the likeliest places are the supplemental questions and the program handbook rather than the admissions landing page. Our AI policy hub tracks published institutional policies across the verticals we have coded so far; physical therapy is not yet among them, and in the meantime this certification is the rule that applies to every PTCAS applicant.

A checklist for the drafting window

1. Read the certification before you draft, not at submit. It is on APTA's application preparation page under "PTCAS Certification and Authorization Statement," linked above. Read "Additional Applicant Expectations" on the same page while you are there; the authorship line sits in that list too, without the exception.
2. Draft in a plain editor with AI features switched off. Predictive completion and inline rewrite suggestions are easy to accept without registering that you did.
3. Ask readers for reactions, not repairs. "Where did you lose the thread?" produces feedback. "Can you fix this paragraph?" produces someone else's sentences.
4. Type every sentence yourself, including ones that came out of a conversation with a reader. Rewriting an idea in your own words is exactly the distinction the clause turns on.
5. Keep example essays in a separate session from your draft. Read them, close them, then write from a blank page.
6. Keep your dated drafts. A trail of your own revisions is the simplest record of your process if anyone ever asks about it.
7. Check each program's own instructions separately. You certified to those as well.

What a GradPilot review is and is not

GradPilot reads a PTCAS essay you have already written against the public PTCAS essay rubric and returns commentary: which parts of the rubric the draft is answering, which it leaves untouched, and where a claim is asserted rather than evidenced. It does not write your essay and it does not rewrite it. No replacement prose comes back for you to paste in. The rubric page is public if you want to see the standard before you send anything, and the PTCAS essay review applies it to the current prompt.

That is a description of behaviour, and we are going to leave two things named rather than answered. The drafts-feedback exception is written about assistance from "other individuals" — an automated reader is not an individual, so that exception is not written about tools of this kind. And APTA has published no reading of how the second sentence, the one about AI text generators and other online sources, applies to a tool that returns critique rather than content. Those are open questions on the page as it stands, and we are not going to settle either one on your behalf.

The certification is quoted in full above. Read it yourself and decide what you are comfortable certifying — that judgment is yours, and it is the one the signature is asking for.