Confidently Wrong AI – The World's Most Useless Ad Algorithm

#418challenge #ai #webdev #javascript

April Fools Challenge Submission ☕️🤡

This is a submission for the DEV April Fools Challenge

What I Built

I looked at how targeted ads work and thought — what if it was just completely, confidently wrong every time?

Modern ads try to understand everything about us.

So I built an AI that understands absolutely nothing.

Confidently Wrong AI is a parody advertisement engine that generates completely irrelevant ads with extremely high confidence and extremely low accuracy.

You enter an interest like:

coding

fitness

travel

money

The AI confidently recommends something completely unrelated:

Invisible book
USB powered rock
Banana phone
Luxury bottled air
Guide to nothing

It then explains its incorrect reasoning with full confidence.

Because sometimes technology should make things worse.

Demo

Live Project:
👉 https://worst-ad-ever.netlify.app/

Example behavior:

Input:
coding

Output:
Invisible Book

Relevance:
3%

Confidence:
97%

Perfectly wrong.

Code

GitHub Repository:
👉 https://github.com/Shsarvagya19/confidently-wrong-ai.git

Tech stack:

HTML
CSS
JavaScript

No frameworks.
No machine learning.
Just artificial ignorance.

How I Built It

The project is built using simple front-end technologies:

HTML for structure

CSS for retro internet inspired UI

JavaScript for generating random irrelevant ads

The logic intentionally produces mismatched recommendations with high confidence scores to parody modern AI recommendation systems.

Key idea:
Create an AI that appears confident but is completely incorrect.

The retro UI design helps emphasize the humorous concept and makes the experience feel intentionally outdated.

Prize Category

Community Favorite

This project focuses on humor, relatability, and interactive experience.

It parodies modern AI systems that often feel overly confident in their predictions.

The goal was to create something simple, memorable, and fun to interact with.

Thanks for reviewing my useless AI.

Top comments (11)

Archit Mittal • Apr 21

This is delightful. 'USB powered rock' genuinely made me laugh out loud. There's an accidentally interesting test pattern here though — parody ad engines like this are actually useful for testing downstream moderation and content filter logic, because they generate plausibly-shaped-but-semantically-nonsense output. Are you seeded with fixed combos or letting the LLM freestyle every ad?

Sarvagya Pandya • Apr 23

Glad you enjoyed the “USB powered rock” 😄
It’s actually all seeded randomness right now — no LLM involved. I’m deliberately mixing structured ad patterns with completely unrelated concepts to get that “plausible but meaningless” effect.
But I really like your point — hadn’t thought of it as a moderation test harness, that’s a pretty interesting direction 👀