In July 2026, the EA exam changes hands. Prometric steps aside, and PSI Services steps in — bringing with it a new testing method, a new question bank, and a level of uncertainty the EA community hasn't seen in the modern era of the exam. PSI calls their flagship delivery system Linear on the Fly Testing, or LOFT. The ambitions behind it are considerable. So are the unknowns facing every candidate who sits down in July.
This isn't a change in who checks your ID at the testing center. It's a change in the exam itself — the questions, the delivery method, and the entire knowledge base candidates have relied on for years. Three things are happening at once.
1. Every question has been rewritten from scratch.
PSI cannot use Prometric's question bank. Those questions are Prometric's intellectual property. So PSI — working with the IRS — has built an entirely new pool of test items from the ground up. Every question you will see in July was written by a new team of subject matter experts. None of them existed in their current form a year ago.
Consider what that means for every prep resource currently on the market. Every Gleim question. Every Surgent practice exam. Every FastForward Academy drill. All of it was built against the Prometric question pool — calibrated to the way Prometric framed tax scenarios, tested terminology, and constructed answer choices. That calibration is now obsolete.
2. Nobody has seen these questions.
Not you. Not your study group. Not the prep companies. Not even candidates who have taken the EA exam multiple times under Prometric. PSI's question bank has never been administered to a live candidate.
For years, a quiet ecosystem has existed alongside official study materials — forums, Reddit threads, informal networks where candidates who just walked out of a Prometric center shared what they remembered. Over time, that accumulated memory shaped how people studied. That ecosystem has nothing to work with yet. It will take the better part of the 2026-2027 testing window before the community rebuilds anything useful.
3. Nobody has any experience with this exam.
The first candidates to sit the PSI exam in July are, in the most literal sense, guinea pigs. No one has taken this version of the EA exam and reported back. There are no pass rate trends. No timing benchmarks. No one knows whether PSI's questions run harder or easier than Prometric's, or whether certain content areas land differently in practice than they appear on paper.
This is the real threat — not any single change, but the complete absence of institutional knowledge that candidates have always been able to lean on under Prometric.
Now for the lofty part.
PSI's LOFT system is worth understanding, because the name suggests something more dramatic than the reality.
Under the old system, every candidate received a pre-built exam drawn from a fixed set of forms. Under LOFT, PSI's algorithm assembles your specific exam in real time the moment you sit down, drawing from a large question pool and selecting your particular set according to a strict content blueprint. You still get the same number of questions as every other candidate. The same subject areas are covered in the same proportions. The difficulty is statistically equivalent across all versions. You just don't get the exact same questions as the person sitting next to you.
PSI calls their proprietary version Formcast™, and it has a long track record. They already use it for the FAA Airman Knowledge Test that pilots must pass for certification. They use it for UK Home Office immigration tests, for real estate licensing exams across the country, and for nursing certification through the American Association of Critical-Care Nurses. This is not an experiment. It is a proven delivery method deployed across high-stakes credentialing programs for over two decades.
One thing LOFT is not — and this is worth being clear about — is adaptive testing. Your exam does not get harder if you answer correctly, or easier if you struggle. The questions are selected before you begin, not in response to your performance. Think of it less as a smart exam and more as a secure one.
And it is effective on that front. When every candidate draws a different set of questions from a large pool, the informal question-sharing networks described above lose most of their value even as they rebuild. There is nothing consistent enough across candidates to harvest and pass along. The security benefit is structural, not incidental.
The irony is that LOFT — despite its prominent billing in PSI's announcement — may actually be the least of your concerns. The real exposure is punches one and two: a brand new question bank, written by people you've never heard of, that no living EA candidate has ever seen.
Put all three together and the picture is clear. You are being asked to sit a freshly written, never-administered exam with no community experience to draw on, using prep materials calibrated to a test that no longer exists.
That raises an obvious question: what kind of preparation actually holds up under these conditions?
We'll get into that next issue.