Expert insights on AI code detection and academic integrity
As AI coding assistants become ubiquitous, learn how institutions are adapting to detect AI-generated code and maintain educational standards.
Codequiry Editorial Team
·
Jan 5, 2026
Stay ahead with expert analysis and practical guides
General
17 min
Dr. Sarah Chen
·
1 day ago
A student submits Python. Their partner submits Java. A line diff reports 0% identical text, and the plagiarism checker stays quiet. Cross-language copying is a semantic clone problem, and it needs a different kind of comparison than the token matching most tools provide. Here is how the detection actually works, where it fails, and what to put in your syllabus before next term.
General
15 min
Dr. Sarah Chen
·
1 day ago
Dependency scanners read manifests. They cannot see the 300 lines someone pasted into a file with the header deleted. This piece walks through what actually constitutes an open source license violation, why SBOM tooling goes blind at exactly the wrong moment, and how provenance checks catch copied code before counsel does.
General
12 min
Dr. Sarah Chen
·
2 days ago
A mid-size CS department got 41 similarity flags from a single assignment and no written policy for what any of them meant. This is the calibration exercise they ran, the AI cluster that confused everyone, and the starter-file mistake that produced 61 false 100% matches.
General
9 min
Alex Petrov
·
3 days ago
A week-by-week account of the three-signal sweep one bootcamp runs at week 7 of every cohort: peer similarity, web matching, and AI detection in one batch. Includes the ignore-list mistake that cost us two evenings, what LLM-shaped student code actually looks like, and how to turn a flag into a conversation instead of a verdict.
General
10 min
Alex Petrov
·
5 days ago
387 files, three scores, one hiring round. Here's what a 12-week bootcamp learned after moving take-home review from three exhausted instructors to an automated pass that checks peer similarity, public web sources, and AI generation, including the false positives we cleared and the two assignment changes that mattered more than any detector.
General
12 min
David Kim
·
1 week ago
An AI detection score is a signal, not a verdict. This is the four-stage triage I borrowed from a fintech incident pipeline to decide which alerts deserve a conversation, which deserve a case file, and which deserve to be closed.
General
11 min
Rachel Foster
·
1 week ago
MOSS and JPlag compare Java to Java and Python to Python, which means a translated submission can score in single digits while the logic stays identical. This is how one lecturer, a TA, and a department chair handle ports, and what they've learned about the tooling that catches them.
General
15 min
Alex Petrov
·
1 week ago
Two Python submissions scored 4% against each other and in the 70s against a Java gist from 2017. Cross-language plagiarism is the fastest-growing blind spot in academic integrity because translation destroys the text while preserving everything that matters. Here's what survives a translation, what detectors actually see, and where the false positives come from.
General
11 min
Marcus Rodriguez
·
1 week ago
Ottenstein's 1976 detector hashed student Fortran token streams, and most of what we run today is a refined version of the same idea. This is the fifty-year arc from line diffs to winnowing, AST matching, web crawling, and statistical AI detection, plus the failure mode that still bites: a 0% similarity score that tells you nothing about authorship.
General
11 min
Marcus Rodriguez
·
1 week ago
Most statements of work say "original work" and never define it, which is how GPL code ends up in your settlement service. Here is the four-question intake review I run on every contractor deliverable, with the thresholds and tooling that hold up under scrutiny.
General
12 min
Rachel Foster
·
1 week ago
Renaming variables and swapping a for loop for a while loop defeats simple text matching, but it rarely defeats structural comparison. This report walks through the obfuscation ladder, the algorithms that climb it, and the published detection rates behind the claims, including the cases where every engine still misses.
General
10 min
James Okafor
·
2 weeks ago
Twenty-three of 412 submissions in a data structures course shared a Dijkstra implementation that differed by fewer than three tokens, and the peer similarity engine flagged all 23 as a single cluster. Nobody had copied anybody. Here is the mechanism behind convergent AI output, the identifiers and AST evidence that separated it from real collusion, and the three-pass workflow we used to keep the scores from contaminating each other.