I Read Five Public FE Pass Stories That Used AI. AI Was Not the Hero.
In the most useful stories, AI organized, explained, challenged, or diagnosed. The candidates still did the problems.
What I was looking for
"I used AI and passed" is an attractive headline. It is also almost useless unless we ask what the AI did, what the candidate did, and what else changed.
I reviewed five public FE stories in which candidates reported passing and using ChatGPT. They reveal different jobs for AI: planning, explaining, checking, and organizing. This is not a representative sample. Success posts have selection bias, the inputs are self-reported, and there is no comparison group.
The stories do not establish that AI caused anyone to pass. They do show a pattern that I think is worth taking seriously: AI was most credible when it removed friction around learning, not when it replaced learning.
Case 1: A failed diagnostic became a 40-day plan
One FE Civil candidate reported passing on a third attempt after giving ChatGPT available resources and a prior diagnostic to build a 40-day plan. The candidate reported roughly 15 hours per week for five weeks, 700 to 800 practice problems, and multiple resources including NCEES practice products. The candidate also stressed attempting problems before watching explanations.
The responsible interpretation is not "upload a report and pass in 40 days." It is that AI helped convert scattered evidence and resources into an executable calendar. Consistent practice supplied the learning.
Case 2: Eighteen years out of college, with AI as an organizer
Another candidate reported passing FE Other Disciplines on a first attempt 18 years after graduation. ChatGPT organized material, identified gaps, and created focused plans alongside PrepFE, NCEES practice exams, and TI-36X Pro practice.
Years out of school is context, not a diagnosis. AI can organize current observations; it cannot infer mastery from graduation year.
Case 3: Forty-eight hours, but a very unusual starting point
A Mechanical FE candidate reported passing after two days of preparation, with four to five hours of practice each day. When stuck, the candidate asked ChatGPT for the solution approach instead of searching for a separate video.
This is the easiest story to misuse. The candidate was still in school and explicitly said that recent coursework helped. That starting point is radically different from someone returning after ten years or recovering from several unsuccessful attempts.
The transferable part is the just-in-time loop: attempt, identify the obstacle, get a focused explanation, and return to solving. AI reduced search friction; recent coursework supplied the unusual starting point.
Case 4: A full-time worker used AI to challenge imperfect material
An FE Electrical candidate described 3.5 months of preparation while working full time, using a free course, NCEES practice material, a commercial book, correction notes, calculator practice, and timed work.
The unusual AI use was verification after noticing errors in free instructional examples. I would add another step: challenge the source with AI, then challenge the AI with an authoritative reference or reviewed solution. Two unverified systems agreeing is not proof. The candidate's correction notes were the durable asset.
Case 5: A ninth attempt shows why explanation is not attribution
One FE Civil candidate reported passing on a ninth attempt after 18 years out of school. Preparation eventually included a formal course, videos, several problem sources, and ChatGPT for explanations and support.
The candidate emphasized greater attention to earlier-session subjects and recurring time pressure. It is powerful evidence of persistence but weak evidence for one resource: too many elements changed.
The candidate also used AI to estimate a prior score. That is a useful warning: NCEES reports results as pass or fail, does not publish the passing score, and uses scaled scoring. Any unofficial percentage reconstructed from a diagnostic report should be treated as an estimate, not a fact.
The common workflow hiding inside the stories
Across these accounts, I see four defensible roles for AI.
1. Planner
AI can turn an exam date, availability, resources, and weak-area evidence into a schedule. Humans still set rest, coverage, and revision rules.
2. Explainer
AI can respond when a learner is stuck. Ask it to identify the first incorrect assumption without revealing the complete solution.
3. Auditor
AI can flag a possible inconsistency, but cannot certify its correction. Check technical conclusions against the current handbook, a reviewed solution, or a qualified expert.
4. Learning historian
AI can classify repeated errors by concept, equation, units, calculator, lookup, or time. Structured history beats an unorganized chat transcript.
What these stories do not prove
We do not know how many AI-assisted candidates failed and did not post. The stories do not compare similar users, measure retention, or establish chatbot accuracy on engineering work.
Model evaluations document confident errors. A 2025 undergraduate physics trial found strong gains from a structured AI tutor using expert solutions and sequential scaffolding. It was neither unrestricted chat nor an FE pass-rate trial.
That is why the product principle I take from these five stories is conservative: let AI shorten the distance between an honest attempt and useful feedback. Do not let it erase the attempt.
Primary sources
Build the plan from your own baseline
A diagnostic cannot predict your result, but it can show which topic is the best next use of your study time.
Take the free diagnostic