Credit the AI tools you used
MLH's rules guide says submissions should credit publicly available frameworks and state how LLMs or AI tools were used. That means model names, generated code, prompt-heavy workflows, and AI-assisted assets should not be hidden.
Crediting AI does not make the project weaker. It makes the build auditable. Judges can separate your product decisions, integration work, data handling, and demo craft from model output.
Do not submit a reskin
MLH also says projects should not be reskins of existing AI tools. A chatbot with a new landing page and no original workflow is the exact pattern this rule is trying to catch.
If you use an AI API, add your own task flow, data source, interface, evaluation, or automation. The question is simple: what did your team build that did not already exist before the event?
Timestamps are part of the build
AI ATL 2025 required a Devpost submission, a GitHub repo created between the official start and end of hacking, and a demo video roughly 5 minutes long. That GitHub timing rule turns your repo history into evidence.
Create the repo after kickoff if the rules require it. Commit steadily. Keep setup notes. A single final upload may be honest, but it looks harder to verify than a visible build trail.
Cheating checks look for old work
MLH's cheating-check guide says cheating is more common for digital events and recommends checking all winning projects. It tells organizers to inspect GitHub start times, one-big-commit patterns, old project references, team-member contributions, video timestamps, and Devpost history.
Assume winners will be audited. If you reused code, say what was pre-existing and what changed during the event. If the rules ban prior work, do not bring it in.
Make the submission self-defending
HackAI 2025 required a Devpost explanation, Google Form track selection, GitHub repo, and demo video of 2 minutes max. Devpost's judging criteria guide says baseline requirements such as eligibility, required tools, accessible URLs, and demo limits can decide whether a submission is valid.
Your final checklist should include AI credit, repo timing, public links, demo length, track selection, and any sponsor API requirement. If a judge has to ask whether the build was allowed, you already made the submission too vague.
< read by a human · updated as things change >
browse hackathons