AI Search Got 7 of 7 Dental Programs Wrong — Our Audit

We verified AI search summaries against seven dental programs' own admissions pages. Every one contained an error that would cost an applicant money or a cycle.

We wrote seven CAAPID program guides and checked every fact against the school's own page. All seven AI search summaries we started from were materially wrong \u2014 here is each error, what it would have cost, and how to check anything yourself. return ( <> What We Did, and Why Between July and August 2026 we wrote seven guides to CAAPID advanced standing programs — Michigan, Pacific Dugoni, Harvard, Washington, UNC, UT San Antonio and VCU. For each one we started the way most applicants do: we searched, and we read the AI-generated summary at the top of the results. Then, because we were publishing, we did the part most applicants skip. We opened the school's own admissions page and checked every number against it. In all seven, the summary we started from contained at least one material error — not a nuance or a rounding difference, but something that would change what an applicant did, spent, or submitted. What we mean by

'material'. We are not counting stylistic differences or slightly stale phone numbers. A material error is one that would cause a qualified applicant to take a wrong action: order the wrong credential evaluation, sit the wrong test, target the wrong score, apply through the wrong system, or miss a deadline. The Seven Errors Each row is documented in our research notes for that guide, with the school page we checked it against. Program. What the AI summary said. What the school's own page says. UNC Adams (ASPID). "We recommend receiving a course by course evaluated by World Education Services (WES)". UNC requires ECE. Its FAQ answers "No" to WES outright.. Michigan (ITDP). TOEFL "minimum overall score of at least 94. No waivers will be granted.". Michigan uses the January 2026 scale. An applicant testing now needs a 5, not a 94.. VCU (IDP). TOEFL "at least 90 with a minimum of 20 in each section"; NBDE Parts I and II as the only board

route. VCU publishes 4.5 on the new scale, accepts the iNBDE, and offers a TOEFL waiver the summary omitted entirely.. Harvard (ASID). A verified AADSAS application, DAT scores, and a $95 fee. ASID uses CAAPID, requires INBDE/NBDE rather than the DAT, and charges $150. The summary had described Harvard's domestic DMD track instead.. Washington (IDDS). "NBDE Part 1 is required (earned within the past 5 years) and NBDE Part 2 is required". UW accepts the INBDE and publishes no five-year recency rule. The rule appears to have been invented.. Pacific Dugoni (IDS). No TOEFL requirement mentioned, and no minimum given. Dugoni has one of the strictest rules we found: scores from before 21 January are rejected outright.. Nova Southeastern (IDP). Bench test "involves class 2 amalgam and full gold crown prep"; deadlines of June 15 and June 1. NSU says the test "may include" an amalgam and a full metal crown preparation — never "class 2", never

"gold". The June 15/June 1 pair comes from a 2024-25 catalog that is two cycles stale.. The Errors Fall Into Four Kinds What struck us was not that summaries were wrong, but how consistently they were wrong in the same four ways. Stale data presented as current. The most common failure by a distance. TOEFL changed its scoring in January 2026, and summaries kept quoting the old 0–120 scale as though nothing had happened. An applicant aiming for "94" on a scale that now tops out around 6 is not slightly off — they are preparing for a test that no longer exists in that form. NSU's deadlines came from a catalog two cycles old. Program conflation. Almost every dental school runs two separate tracks: a four-year domestic DDS or DMD through AADSAS, and a shorter international advanced standing program through CAAPID. They have different applications, different tests, different fees and different deadlines. The Harvard summary blended them so

smoothly that nothing in the text signalled a problem — it read as a confident, coherent answer to a question we had not asked. Invented specifics. Washington's "earned within the past 5 years" is the clearest case. The rule is precise, plausible, and the kind of thing a dental school genuinely might publish. It simply is not on UW's page. Silent omission. Harder to catch than a wrong fact, because nothing looks wrong. The Dugoni summary did not say "Dugoni has no TOEFL requirement" — it just never raised the subject, and a reader would reasonably conclude there was nothing to know. Dugoni's rule is one of the strictest in the pool. The omissions are the dangerous ones. A wrong number invites a double-check because it looks like a fact. A missing requirement invites nothing at all. If you take one habit from this piece, make it this: when you finish reading a summary, ask what it didn't mention — waivers, score expiry dates,

supplemental fees, which evaluator is accepted — and go looking for those specifically. What These Errors Cost The UNC one is the example we keep coming back to. A credential evaluation is not a form you fill in — it costs a few hundred dollars, takes weeks, and has to be sent from the agency directly to CAAPID. An applicant who read that summary, ordered a WES evaluation, and waited would discover the problem when UNC's FAQ told them "No", by which point the money is spent and the calendar may have run out. The Harvard error costs the whole application. Someone acting on it would prepare for the DAT, register with AADSAS, and budget $95 — and none of those things is part of the program they were trying to apply to. These are not exotic failure modes. They are the ordinary result of reading a fluent paragraph about the wrong program. We Got Things Wrong Too It would be convenient to end there. But when we audited our own Nova

Southeastern page in August, we found four claims that no source supported — including that NSU ran "the longest CAAPID program at 39 months", when the University of Florida's is listed at 45. We had repeated NSU's own marketing line, that it is "Florida's only standalone IDP", as though it were a fact. It is not; ADEA lists a University of Florida program. We had also published a list of Kira-using programs described as "verified" that was missing UT San Antonio, which states its Kira usage on its own admissions page. All of those are now corrected, and each of our program guides carries a research file recording what we confirmed, what the school does not publish, and what we had to correct. We mention our own errors because a piece arguing for verification is worth very little if the author exempts themselves from it. How to Check Anything Yourself, in About Five Minutes None of this requires special skill. It requires going one

click further than the summary. Find the program's own page, not a page about it. The domain should be the university's. If you are reading a test-prep company, an aggregator or a forum, you are reading a copy, and copies go stale silently. Check you are on the right track. Look for the words CAAPID or AADSAS, "advanced standing" or "four-year". This one check would have caught the Harvard error on its own. Ask what year it is talking about. A page that names no cycle may not have been touched in two. Cross-check against ADEA. The ADEA CAAPID Program Finder lists seats, length, deadlines and fees for participating programs. Where it disagrees with the school, you have found something worth a phone call — NSU's page currently contradicts itself on its own deadline, and no source resolves it. Phone the admissions office for anything that costs money or has a date on it. Ask for the answer by email. A written reply from the office beats

every other source on this list. Our own program guides — including UNC , Michigan , Harvard and VCU — are built this way, and each says plainly when a school publishes nothing rather than filling the gap. The full program comparison collects what we verified across all of them. Practise the interview, once your facts are straight. Answer real dental school interview questions out loud and get scored on clarity, structure and reasoning.. Try it free →. Sources, So You Can Check Us Every claim in the table above was verified against the programme's own admissions pages between 20 July and 15 August 2026. Those pages are below. If one of them now says something different from what we report, the school has changed it since we looked — and we would rather you trusted the school than us. UNC Adams (ASPID) — dentistry.unc.edu Michigan (ITDP) — dent.umich.edu VCU (IDP) — dentistry.vcu.edu Harvard (ASID) — www.hsdm.harvard.edu Washington

(IDDS) — dental.washington.edu Pacific Dugoni (IDS) — www.pacific.edu Nova Southeastern (IDP) — dental.nova.edu Seats, programme lengths and deadlines were cross-checked against the ADEA CAAPID Program Finder . Our full working notes for each programme, including what the school does not publish, are kept alongside the guides themselves. Citing this page. Journalists and advisers are welcome to use these findings. A workable citation: DentPrep Editorial Team, "We Checked AI Search Against 7 Dental Programs' Own Pages. All 7 Were Wrong." DentPrep, 15 August 2026. https://dentprep.me/blog/ai-search-dental-school-errors/ If you want the underlying per-programme notes, or want us to walk through the methodology, get in touch and we will send them. The Limits of This Seven programs is seven programs. This is a record of what we hit while doing research we were doing anyway, not a controlled study, and we did not sample randomly or test the

same query repeatedly. Summaries are regenerated constantly, so the specific wording we saw may not be what you see today, and some of it may since have improved. What we would defend is narrower and, we think, more useful: across seven consecutive programs in a domain where the underlying facts are published, unambiguous and free to check, the summary was wrong every time in a way that mattered. Dental admissions is not an obscure corner of the internet. If the error rate is this high here, the reasonable assumption is that it is not lower somewhere the facts are harder to find. Verify anything that costs money or has a deadline attached. That was always good advice. It is now load-bearing. ); }; export default AiSearchDentalAdmissionsErrors; How often is AI search wrong about dental school requirements? In our audit, seven out of seven programs had at least one material error in the AI-generated summary we started from. That is a

record of what we encountered while researching seven consecutive CAAPID guides between July and August 2026, not a controlled study \u2014 but every one of those facts was published, free and unambiguous on the school's own website, so the errors were avoidable. What kinds of mistakes does AI search make about admissions? Four recurring kinds. Stale data presented as current, such as quoting the pre-January-2026 TOEFL scale. Program conflation, such as describing a school's four-year domestic DMD when asked about its international advanced standing track. Invented specifics, such as a five-year board score expiry rule that the school does not publish. And silent omission, such as never mentioning a TOEFL requirement that exists and is strict. Does UNC accept a WES credential evaluation for advanced standing? No. UNC's Adams School of Dentistry requires an ECE course-by-course evaluation, and its FAQ answers "No" to WES directly. An AI

summary we encountered recommended WES. Acting on that would have cost several hundred dollars and possibly the application cycle. How do I verify a dental school's requirements myself? Open the program's own page on the university's domain rather than an aggregator. Confirm you are reading the right track \u2014 look for CAAPID and 'advanced standing' versus AADSAS and 'four-year'. Check whether the page names an admissions cycle. Cross-check seats, deadlines and fees against the ADEA CAAPID Program Finder. Then phone the admissions office about anything that costs money or has a date attached, and ask for the answer in writing.