AI Disclosure & Methodology

Why This Page Exists

Intelloro is an AI tools directory built using AI. To meet our obligations under the EU AI Act (Regulation 2024/1689) and to give every visitor a clear picture of how the platform works, this page documents:

• What AI systems we operate and what they do
• How we have classified each system under the AI Act risk tiers (Articles 6–9)
• What transparency obligations apply (Article 50) and how we meet them
• How we promote AI Literacy among users (Article 4)
• How we rank tools (the “methodology”) so you can interpret our scores
• How we handle errors and incidents

1. AI Systems We Operate

Search and AI features receive text you type. The Decision Engine and Semantic Search send your query to the AI providers named above (Google, OpenAI and, as a fallback, Groq), and the Stack Builder sends your goal, role and budget to Google (or Groq as a fallback). They do not receive your name, email address or account profile. The other systems in this table (Listing Builder, Source Checker, Score Generator) work only on public vendor information. See our Privacy Policy for what is stored and for how long.

2. EU AI Act Risk Classification (Articles 6–9)

We have reviewed each AI system above against the AI Act's risk tiers and Annex III high-risk categories. Our classification:

None of our systems are classified as high-risk under Annex III (biometric ID, critical infrastructure, education, employment, essential services, law enforcement, migration, or justice administration). If our use case ever expands into a high-risk domain, we will conduct a full conformity assessment per Articles 16 and 43 before deployment.

Prohibited practices (Art. 5): Intelloro does not engage in any AI Act-prohibited practices — no subliminal manipulation, no exploitation of vulnerabilities, no social scoring, no real-time remote biometric ID, no untargeted face-DB scraping, no emotion inference in workplace/education, no biometric categorisation by sensitive attributes.

3. Article 50 — Transparency Obligations

EU AI Act Article 50 requires providers and deployers of AI systems to inform users when they are interacting with AI or consuming AI-generated content. Our compliance:

• AI-Generated Editorial Content: AI-synthesized prose on detail pages — the description, tagline, “what makes it unique”, quick answer, best-for, pros, limitations, and “consider alternatives if” — is disclosed at the point of consumption: a small visible “AI” marker beside each field's heading plus an “AI-Assisted Data” summary on the page. The marker means the text is AI-generated; the underlying facts come from the vendor's own pages and go through our source check.
• AI-Generated Scores: Task Scores, Dimension Scores, and Trust Scores are labelled as AI-estimated, with links to this methodology page from the score badges.
• No Deepfakes / Synthetic Media: Intelloro does not host or generate deepfake images, audio, or video. The site does not produce synthetic media that could be mistaken for real persons.

4. Article 4 — AI Literacy

EU AI Act Article 4 requires providers and deployers to ensure a sufficient level of AI literacy among staff and other persons dealing with the operation and use of AI systems. Our measures:

• User education: this page itself is a public-facing AI Literacy resource. Score badges and tool detail pages link here so any user can understand how AI shapes what they see.
• Limitation disclosure: our AI methodology is published openly on this page so users understand AI-extracted data may contain errors.
• Right to verify: every listing links to the vendor's own website, so users can check any AI-generated claim against the source.

5. How We Rank Tools (Methodology)

Trust Score (0–100)

Calculated from five weighted pillars — Compliance & Verified, Operational, Reliability, Market Proof and Company — built from verifiable signals such as company verification, named customers, funding, status page, measured uptime, compliance certifications and data residency. A missing signal is not counted against an item, and an item with too little verifiable data gets no score. The score is recalculated whenever a tool is updated.

Task Scores (1–10 across 10 per-vertical task keys)

Each tool/agent is scored on every one of 60 category-specific rubrics that fits it (its own category's rubric plus one for each task it does) (e.g., chatbots, coding, image generation, design, video editing, voice agents, vector databases) Each rubric has 10 task keys specific to that vertical (e.g., chatbots rubric has “conversation_quality”, “knowledge_accuracy”, “context_retention”; image-generation rubric has “text_to_image”, “style_consistency”, “prompt_understanding”). Each task key is scored using a tier-based system: (Tier 1) countable evidence on the vendor website (e.g., “250,000+ templates” → high template-variety score); (Tier 2) AI-generated estimate based on the rubric's per-task anchor definitions. Third-party review-platform ratings are never fed into task scores.

Dimension Scores (1–10 across 6 dimensions for tools, 7 for agents)

Universal (tools + agents): Ease of Use, Customizability, Support Ecosystem, Integration Power. AI-specific (tools + agents): Compliance & Data Protection, Performance. Agent-only: Planning Quality. All tool dimensions are computed algorithmically from verified structured data; for agents, Planning Quality is LLM-assessed from public product content using its rubric anchors and kept within the range set by the agent's documented planning capability. Third-party review-platform ratings (Capterra, GetApp, G2, TrustRadius, Trustpilot) and ProductHunt are not used or publicly displayed anywhere on Intelloro; only npm and PyPI package-download counts are shown.

Search & Recommendation Ranking

Free-text search uses embedding similarity (cosine distance) on tool name + tagline + description. Personalized recommendations (Decision Engine) combine: (1) explicit user preferences from the profile, (2) Trust Score, (3) Task Score for the matching category, (4) the user's industry and team size. Sponsored listings are clearly labelled “Sponsored” and never displace organic ranking based on score.

6. Known Limitations & Errors

All AI systems make mistakes. We disclose the known limitations of ours:

• Tool descriptions may contain factual errors or outdated information — vendors may change features or pricing without our knowledge.
• Compliance flags (GDPR, SOC 2, HIPAA) reflect vendor self-reporting — we do not independently audit these claims.
• Scores can be wrong if the underlying source data is wrong (e.g., an outdated price on the vendor's page).
• The Decision Engine does not understand domain-specific nuance; recommendations should be a starting point, not a final purchase decision.
• Package-download counts (npm, PyPI) reflect a snapshot in time and may not be current.

If you find an error, please report it via our contact form or use the Report Content page if it is harmful or misleading. We correct verified inaccuracies directly.

7. Incident Handling & Reporting (Art. 73)

Should any of our AI systems produce a serious incident (e.g., systematic harm, large-scale misinformation, breach of fundamental rights), we will:

• Take the affected system offline within 24 hours
• Notify affected users without undue delay
• Report to the relevant national competent authority within 15 days (or 2 days for widespread incidents) if AI Act Art. 73 applies once enforcement begins
• Document the root cause, remediation, and lessons learned in our annual Transparency Report

8. Governance & Accountability

AI Lead / Operations Lead: Sultan Mahamud, Head of Operations, is accountable for AI governance at Intelloro. AI system changes are reviewed by the operations lead before deployment. We will move to a periodic (at minimum annual) re-review of risk-tier classifications as the platform scales.

Vendor models: we use foundation models from third parties (Anthropic, OpenAI, Google, and Groq as a fallback). We operate under each provider's published API terms and data-processing terms. Most inputs are public vendor information about AI tools. The exception is text you type into search and the Stack Builder, which is sent to Google, OpenAI or Groq as described in our Privacy Policy; we do not send your name, email address or account profile with it.

Updates: any material change to AI methodology will be reflected on this page.