Everything about the simulation
Starling simulates how AI agents discuss, persuade, and change their minds — just like real people. Unlike surveys that only show results, Starling shows you why people think that way, which arguments change minds, and where the risks are.
What you get
Results first, then the reasons behind them
Direction of reaction
See whether people react positively or negatively, and how opinions split — shown as clear percentages.
Arguments that change minds
Track which arguments actually shifted opinions. Know the difference between messages that work and ones that don't.
How opinions spread
See who speaks first, who follows, and where opinions split. Watch how public opinion forms in real time.
Hidden risks
Discover backlash points and controversies you didn't think of. The simulation finds them before your audience does.
Actionable recommendations
Get specific recommendations backed by simulation evidence — what to do, why, and what to expect.
How is this different from surveys
We show the process, not just the result
Surveys capture a snapshot
"Would you buy this?" → "Yes 60%, No 40%." You don't know why, or whether they'd change their mind after hearing others' opinions.
Starling shows the journey
Watch someone go from positive to negative after hearing a pricing argument. See who started that argument, and whether a counter-argument worked.
So you can actually act on it
Knowing "40% negative" isn't actionable. Knowing "price is the reason" means you can reframe your message around value.
How it works
AI agents discuss like real people
Unique people
Agents with different ages, jobs, personalities, and values are created — reflecting real-world diversity.
Read & decide
They read others' posts and decide whether to agree, argue, or ignore — based on their own personality.
Take action
Post, comment, like, follow, share — everything people do on social media.
Change & grow
When they encounter a compelling argument, they may change their mind. Every change is tracked with the reason.
Use cases
Questions Starling answers
Public Opinion
"What will people think about this?" — opinion distribution, key arguments, how opinions form.
Marketing Reaction
"How will people react to this product?" — purchase intent, criticism points, effective messages.
Policy Debate
"What debate will this policy trigger?" — pro/con arguments, resistance factors, persuasion potential.
Crisis Response
"How will public opinion shift after this event?" — information spread, sentiment changes, response strategies.
Community Analysis
"How will this community grow?" — engagement patterns, churn factors, growth drivers.
Any Topic
Simulate any topic that doesn't fit the categories above. Completely open-ended.
Validated Results
Proven with real-world cases
Starling's simulations have been validated against actual election results, product launches, and policy announcements. All simulations were run without polling data — using only facts and demographics.
Election result prediction
2025 Korean Presidential
2026 Wisconsin Supreme Court
2026 Thai General Election
Consumer sentiment prediction
Samsung Galaxy S25
Tesla Cybertruck
Tom Scott return video
Policy opinion prediction
Trump Tariff Policy
Korean Pension Reform
Pharma Tariff 100%
Crisis sentiment prediction
Hyundai Palisade Recall
Nestlé Formula Recall
vs Traditional Research
AI Simulation Features · Concept Testing
Everything you need for simulation
Goal-Driven Analysis
Enter your analysis goal and everything from agent composition to report sections is designed around it. Get precise answers to the questions you ask.
Agent Interactions
Follow, repost, comment, like. Agents interact like real social media users, simulating viral spread patterns and echo chambers.
Real-Time Monitoring
Watch agent thought processes, opinion changes, and network formation in real time as the simulation runs.
Dynamic Reports
3-5 analysis sections are auto-generated based on your topic and category. Includes insights, visualizations, and actionable recommendations.
Confidence Scoring
Pre and post-simulation confidence scores. Preview how accuracy changes with different agent and round counts to find optimal settings.
Follow-Up Q&A
Ask follow-up questions about the report. Get evidence-based answers citing specific agent statements.
Analysis Report
Here's what you'll get
Executive Summary
A 1-3 sentence summary directly answering your analysis goal, with key insights and confidence score.
Visualizations
Sentiment timelines, influencer rankings, opinion flows, network graphs, heatmaps — up to 10 chart types.
Recommendations & Risks
Prioritized action items with evidence and expected impact. Risk warnings with mitigation strategies.