authentiq-ai

Authentiq combines World ID and machine learning to ensure every review is from a real human and meets quality standards — making fake and spam reviews impossible.

Comment

Inspiration

In today’s digital world, trust in online content is rapidly eroding. Reviews on platforms like e-commerce sites, marketplaces, and gig apps are increasingly manipulated by bots and AI-generated spam. As large language models improve, fake reviews are often indistinguishable from genuine ones.

We realized that most existing solutions focus only on detecting bad content, rather than addressing the root problem:

$$ \text{Who is generating the content in the first place?} $$

This led us to combine proof-of-personhood with machine learning to ensure content is both human and high-quality.

What it does

Authentiq is a trust infrastructure layer for online reviews that combines:

  • World ID (Proof of Personhood) — ensures every reviewer is a unique human using zero-knowledge proofs
  • Machine Learning Scoring — evaluates each review for usefulness, sentiment, and spam likelihood in real time

Together, this creates a dual-layer system:

  • Identity Trust — Is this a real human?
  • Content Trust — Is this a meaningful review?

We don’t just detect fake reviews — we make them impossible by ensuring every submission comes from a verified human and is evaluated by AI.

How we built it

We built Authentiq as a full-stack web application:

  • Frontend: Next.js + React
  • Backend: API routes for verification and data handling
  • Identity Layer: World ID integration (QR-based verification + RP signing)
  • Machine Learning: Review scoring pipeline using trained models
  • Data Layer: Lightweight JSON-based storage for rapid prototyping

World ID Integration

  • Generated signed rp_context via backend
  • Displayed QR-based verification using IDKit
  • Verified proofs through the World Developer API
  • Enforced one-human-per-review using nullifier-based logic

Machine Learning Pipeline

  • Analyzed review text for sentiment and usefulness
  • Detected spam or low-quality content
  • Produced real-time scores for each review

Challenges we ran into

World ID Integration

  • Handling RP signing and verification flows
  • Managing HTTPS and ngrok constraints
  • Debugging action mismatches and verification errors

Combining Identity + ML

  • Designing a seamless UX between verification and scoring
  • Ensuring users understand both trust layers

Time Constraints

  • Balancing rapid iteration with building a robust end-to-end system

Accomplishments that we're proud of

  • Successfully integrated full World ID verification flow end-to-end
  • Built a real-time machine learning scoring system
  • Created a unified system that enforces both identity and content trust
  • Designed a clean, intuitive UI for a complex trust pipeline

$$ \textbf{One human. One review. Verified and scored.} $$

What we learned

  • Trust is multi-dimensional — identity and content both matter
  • Proof-of-personhood becomes powerful when paired with application logic
  • AI systems are stronger when grounded in verified human input
  • Building secure, real-world integrations requires careful coordination across the stack

What's next for authentiq-ai

We plan to expand Authentiq into a scalable trust platform by:

  • Adding persistent storage and user reputation systems
  • Improving ML models with larger datasets and fine-tuning
  • Supporting additional use cases like voting systems and marketplaces
  • Integrating advanced fraud detection and analytics

$$ \text{Reactive Moderation} \rightarrow \textbf{Proactive Trust Systems} $$

Share this project:

Updates