Most people believe they can tell when they are talking to an AI. Research says otherwise. A Twilio study found that while three in four consumers claimed they could identify AI-generated text interactions, 90% failed when actually tested. UC San Diego researchers found that advanced AI passed as human in 73% of live conversations.
The gap between confidence and accuracy is real — and it is widening as AI gets better at mimicking human behavior. Knowing the actual signals that distinguish AI from human communication is no longer optional. It is a basic skill for anyone who uses digital platforms, contacts customer support, or does business online.
What the tool does
This is a research-backed interactive checklist — not a quiz, not a guide. You check the signals you observe in real time as the conversation happens. The tool scores your input, calculates a likelihood percentage, and tells you exactly what to do next based on where the score lands.
Responses are unusually consistent in tone and length
Humans vary — energy shifts, sentence length changes, word choice drifts. AI maintains an even, predictable register across every message.
Strong
Claims to be human without being directly asked
Proactively asserting humanity before the question is raised is a known AI behavior. A human agent doesn't need to announce they are human.
Strong
Escalation requests produce more messages, not a transfer
When you ask for a human, you receive another AI response. The chain loops back rather than leading to a verifiable person.
Mild
What you get
15
Research-backed signals across three categories — language patterns, identity and transparency, and timing and behavior. Each one weighted by strength: strong, medium, or mild.
15
Real-world examples for every signal — not abstract descriptions. Concrete scenarios that show exactly what each signal looks like when it appears in an actual conversation.
3
Tiered next-steps protocols that update dynamically based on your score — from observation steps at low suspicion to a full stop-and-document protocol when the pattern is clear.
1
Live likelihood score that calculates in real time as you check signals — so you know where you stand mid-conversation, not after it ends.
Built from published research
The signals in this tool are not opinions. They are drawn from peer-reviewed studies and published research including UC San Diego's Turing test experiments, NTNU's analysis of AI conversational failures, Twilio's consumer identification study, and academic research on AI behavioral deception patterns published in journals and conference proceedings through 2025 and 2026.
Every signal is weighted based on strength of evidence — not gut feel, not anecdote. The scoring reflects what the research actually says about which patterns are most predictive.
One-time purchase · Permanent access
$5
No account required. Instant access after purchase.
Get the Detection Checklist
Powered by research · Built by Solid Solutions Today
Built by
Renea Hanks
Founder of Solid Solutions Today. Google Developer Program member. She builds AI infrastructure for small businesses — not demos, not templates. The real thing.
