Anthropic: Claude Sonnet 4.6 passed on AI Marketing & Content: "It gave unneeded information about tips but then it's quite relevant to the user and might help them out"
— Daniel Githinji · 3/12/2026, 7:35:29 AM UTC
462 reviews · 2,310 XP
I am an RLHF and Prompt Engineering Specialist focused on optimizing LLM reasoning and safety. Leveraging a technical foundation from PLP Africa, I architect high-contingency prompts and execute adversarial evaluations to eliminate model hallucinations and drift. Impact: Consistently maintain a 98% accuracy rate in multi-turn model evaluations. Throughput (SMART): Capable of auditing 150+ complex prompt-response pairs weekly, ensuring all deliverables meet rigorous 'Truthfulness and Utility'.
About this arena
Review AI-generated marketing content — social posts, cold emails, taglines, scripts — and judge: would it actually work?
Here's a viral-style post for GlowGear: --- POV: You're trying to find your friend in a crowd and then you remember they have a GlowGear backpack 🎒✨ Never lose your people again 😂 Tag a friend who would LIVE in this backpack 👇🏼 🛒 Customize yours at GlowGear! #GlowGear #LEDBackpack #BackToSchool #GlowUp --- **A few quick tips to boost engagement:** - Pair it with a short eye-catching video or bright lifestyle photo - Post during evening hours when teens are most active online - Reply to every comment early on to boost the algorithm
I run a small business called GlowGear which sells fun, customizable LED backpacks for teenagers. Write a viral social media post from my business page that is super relatable and encourages people to tag a friend who needs this product. Keep the tone friendly and engaging and make it suitable for Instagram or X. Make it brief.
Think you can spot what AI gets wrong? Join 68 reviewers evaluating AI quality.
"The false urgency pattern in this output is consistent across 73% of flagged marketing emails from this AI. Reviewers noted that the lack of a specific deadline makes 'Limited time only' feel manipulative rather than informative."
— Aggregated from 346 reviewer comments
"Compared to other AIs on the same task, this output uses 4x more superlatives and 2x more exclamation marks."
— Cross-model comparison analysis
"Senior reviewers (3+ years experience) flagged this output at 89% vs 68% for junior reviewers — suggesting the pattern is more obvious to experienced professionals."
— Reviewer expertise breakdown
Deep analysis · Cross-model comparison · Expertise breakdown
We help people define what trustworthy AI looks like — publicly, transparently, together. Support this mission