A well-written question with the wrong answer keyed is worse than no question at all — it teaches something false, confidently, to someone who can't yet tell.
That sentence is the reason this product is built the way it is. You can check a question for grammar, for plausible distractors, for whether the options are the same length — and a beautifully constructed item with a wrong answer will pass every one of those checks. The only defence is knowing where the answer came from.
The method
-
Nothing gets written without a source
Every question begins from a document: a standard in the ACA Code of Ethics, a criterion set in the DSM-5-TR, a task on NBCC's published content outline, the Standards for Educational and Psychological Testing, or a named counseling model. If a claim can't be grounded in one, the question doesn't get written.
-
The source is recorded with the question, not remembered
Each item stores what grounds its answer as a field, next to the question itself. The build refuses to ship an item that doesn't have one. That's a mechanical check, not a promise — it can't be skipped on a tired afternoon.
-
Numbers get checked against the source, automatically
Diagnostic thresholds are where confident errors live — two weeks, six months, three days. Every duration in a DSM-sourced question is checked by script against a table of criteria transcribed from the published text. Mismatches fail the build.
-
Verification uses the actual edition
Criteria changed between DSM-IV-TR and DSM-5-TR in ways that matter: the ADHD onset age, the acute stress disorder floor, the bulimia frequency, the substance use severity bands. Genuine text from the wrong edition will confidently key dozens of questions wrong, so edition is checked before anything is verified against it.
-
Human review goes where the machine can't help
Rapport is built by a Licensed Professional Counselor. The automated checks above catch a wrong number or a missing source, but they cannot catch a fluent, plausible, wrongly-keyed judgment question — so those are the ones read line by line against their source and signed off by hand. Approval is bound to the text of the item, so editing a question returns it to the queue rather than inheriting the old sign-off.
-
Mistakes get recorded, not quietly fixed
There's a log of the errors this project has made and the rule each one produced. A process that has never caught itself in an error isn't careful; it just isn't looking.
What you see as a result
Under every rationale there's a line naming what the answer rests on — DSM-5-TR, ACA Code of Ethics (2014), motivational interviewing. It's on the question itself, not hidden behind a tap, because a trust signal nobody sees isn't one.
All 601 carry one. Of those, 310 cite a published document — the DSM-5-TR, the ACA Code, NBCC's outline, the testing Standards. The remaining 291 name the model and the person who developed it. That's a real distinction and we'd rather draw it than blur it: "every question names what it rests on" is true and you can check it. "Expert-verified" would be vaguer, and weaker.
What isn't finished
An item-by-item clinical countersign across the whole bank is not complete. Every question has a recorded source and passes the automated checks; the hand review described above has covered the highest-risk items rather than all 601. We would rather state that plainly than let "reviewed by a licensed counselor" do more work than it has earned. When the full pass is done, this page will say so — and until then it says this.
The case bank carries the source line too. All 393 NCMHCE case questions name what they rest on, drawn from 130 distinct sources. Both banks are covered, so there is no corner of the product where this claim quietly does not hold.
What we don't claim
Nothing here reproduces an actual exam item — every question is written from published sources. We don't publish a pass rate, because we don't have one, and nobody else in this category who publishes a chart has one either. We don't predict whether you'll pass: NBCC sets the passing score and it varies by form.