New Breakthrough AI Detection. Our best yet for AI-written source code. Catches 90% of AI/GPT code with 1.3% false flags. 90% caught, 1.3% false flags. Read more Read more
New: MCP integration. Run plagiarism scans from Claude, Cursor or any AI assistant. Run scans from Claude or Cursor. Set it up
Hiring Take-home tests and code challenges

Catch AI-written answers in take-home coding tests.

A take-home test only tells you something if the candidate wrote it. Upload the submissions for a role and Codequiry scores each file for AI generation, checks it against GitHub and the web, and compares candidates with each other.

Check a batch free How screening works

Upload a folder of submissions, or send each one through the API.

90%
of AI-written code caught in held-out testing (model v3e)
1.3%
of human-written code falsely flagged, so honest candidates are not rejected on a score
3
checks on every submission: AI, web and GitHub, candidate to candidate
65+
languages, from Python and Go to SQL
The take-home problem

A polished solution no longer means a strong engineer.

An AI assistant can finish a typical take-home in minutes, and solutions to popular challenges circulate on GitHub. Hiring teams end up advancing candidates on work they did not write, and spend onsite hours finding out.

Answers written by a model

Each file gets a probability that it was AI-generated, with the passages that drove the score highlighted.

Solutions lifted from public repos

Submissions are checked against GitHub, Stack Overflow and the web, and every match links to its source.

One solution, several candidates

Every candidate for a role is compared with every other, which is how a leaked answer shows up.

What a hiring check covers

Screen the code before you schedule the onsite.

One upload per role. The results table sorts candidates by what needs a second look, so reviewers spend time where it matters.

AI-generated answers

Per-file AI probability, flagged regions and an estimate of the model family, for code produced with ChatGPT, Claude, Copilot and similar tools.

  • Flagged passages, not a bare score
  • Model family estimate
  • Included in every check

Public solution matches

Popular challenges have public answers. Matches against GitHub, Stack Overflow and tutorial sites link straight to the page the code came from.

  • Linked sources
  • Renamed variables still match
  • Works across languages

Candidate to candidate

Shared or leaked solutions surface as near-identical pairs across a role or across hiring rounds.

API for your pipeline

Send each submission as it arrives and read the result when it finishes, from any language.

Reports to share

Export results for the hiring committee, with the evidence behind each flag.

How screening works

From a folder of submissions to a shortlist.

Collect submissions

Download them from your assessment tool or repository, one ZIP per candidate.

Upload per role

One check per role compares every candidate with every other one.

Review what is flagged

Sort by AI score, web matches and candidate pairs. Open the evidence.

Ask about it live

Use the flagged passages as interview questions. Real authors can explain their code.

Fair to candidates

A flag starts a conversation. It does not end a candidacy.

No detector is right every time, so a high AI score should earn a follow-up question, not a rejection email. The report shows why each file was flagged, so the next interviewer can ask the candidate to walk through it.

AI score on solution.pyHigh
Match to a public GitHub repoNone
Match to another candidateNone
Suggested next stepAsk to explain
Example: a high AI score with no source match. Ask the candidate to extend the solution live; the person who wrote it can.
Questions

AI in coding interviews, answered.

Can you detect ChatGPT in a take-home coding test?

Yes, when the answer is submitted as code files. Each file is scored for AI generation and the passages that drove the score are highlighted. Codequiry checks the code a candidate submits; it does not watch a live interview or a candidate's screen.

How accurate is it?

In held-out testing, model v3e caught 90% of AI-written code and falsely flagged 1.3% of human-written code. Use a flag as a reason for a follow-up question, not as grounds to reject on its own.

Does it catch solutions copied from GitHub?

Yes. Every submission is compared with public repositories, Stack Overflow and the web, and each match links to its source. Renamed variables and reordered functions still match.

Can it tell if two candidates shared an answer?

Yes. All candidates in a check are compared with each other, so a solution that spread between applicants shows up as a near-identical pair.

Can we automate it?

Yes. The REST API creates a check, uploads each submission and returns the results, so it can run inside your hiring pipeline. See the AI-generated code detection API.

Hire the engineer, not the assistant.

Screen your next batch of take-home submissions for AI-written answers, copied solutions and shared code before anyone books an onsite.

Check a batch free See pricing
Keep reading

More on AI-written code

Product names such as ChatGPT, Claude, GitHub Copilot, Cursor and Gemini are trademarks of their respective owners. Codequiry is an independent service and is not affiliated with, endorsed by, or sponsored by any of them.