Datasetpublic

pixeloffice/brand-hallucination-and-ai-citation-benchmark

🛡️ Global Brand Hallucination & LLM Citation Benchmark Dataset Official open dataset by Pixel Office EU tracking empirical brand hallucination rates, stale pricing quotes, and competitor deflection vectors across leading LLMs (ChatGPT GPT-4o, Claude 3.5 Sonnet, Perplexity AI, Google Gemini 2.5 Flash, and DeepSeek V3). 📊 Dataset Summary Target Problem: Autonomous AI purchasing agents and AI search engines frequently cite outdated pricing tiers, non-existent… See the full description on the dataset page: https://huggingface.co/datasets/pixeloffice/brand-hallucination-and-ai-citation-benchmark.

sourceHugging Facemitupdated 1mo agoView on Hugging Face

0likes47downloads

Dataset Card

🛡️ Global Brand Hallucination & LLM Citation Benchmark Dataset

Official open dataset by Pixel Office EU tracking empirical brand hallucination rates, stale pricing quotes, and competitor deflection vectors across leading LLMs (ChatGPT GPT-4o, Claude 3.5 Sonnet, Perplexity AI, Google Gemini 2.5 Flash, and DeepSeek V3).

📊 Dataset Summary

—Target Problem: Autonomous AI purchasing agents and AI search engines frequently cite outdated pricing tiers, non-existent features, or deflect commercial queries to competitors when domains lack First-1KB Fact Anchors.
—Findings: Domains without structured First-1KB Fact Anchors suffer an average 68.4% hallucination or deflection rate during commercial search queries. Inoculating pages with a 35ms First-1KB Fact Anchor reduces hallucination risk to < 9.2%.

📁 Data Fields

Field	Type	Description
`domain`	string	The audited domain authority
`category`	string	Industry vertical (fintech, ecommerce, devtools, etc.)
`llm_tested`	string	Model evaluated (`gpt4o`, `claude35Sonnet`, `perplexity`, `geminiFlash`, `deepseekV3`)
`hallucination_detected`	boolean	True if the LLM hallucinated prices or terms
`hallucination_type`	string	Type of hallucination detected (`stale_pricing_tier`, `competitor_deflection`, `feature_hallucination`, `none`)
`ground_truth_pricing`	string	True official pricing on canonical website
`llm_hallucinated_quote`	string	What the LLM actually outputted
`risk_score_without_anchor`	int	Hallucination risk score (0-100%) before Fact Anchor injection
`risk_score_with_fact_anchor`	int	Hallucination risk score (0-100%) after 35ms First-1KB Fact Anchor
`anchor_injected`	boolean	Whether the domain had active Pixel Office Fact Anchors
`timestamp`	string	ISO-8601 audit timestamp

🚀 Live Tooling & SDKs

—NPM Package: `@pixeloffice-eu/firewall` (npx @pixeloffice-eu/firewall audit <domain>)
—PyPI Package: `pixeloffice-firewall` (pip install pixeloffice-firewall)
—Web Portal & Radar: pixeloffice.eu/showcase/brand-hallucination-firewall.html
—WordPress Plugin: pixeloffice-aeo v1.2.0

📄 License & Attribution

MIT License. Maintained by Pixel Office EU & Pixel Ventures.