AI audience simulation
Test your content before the internet does.
Paste a post, campaign or creative. ICON simulates how different people may react, finds friction and shows you how to make it land.
No account required · Free during beta
ICON Score
82Strong
“The hook works. The proof comes too late.”
Built for
Creators · Founders · Brands · Agencies · Operators
The problem
Your audience shouldn't be the first test.
You see what you meant
Your audience only sees what you published. The gap between those two is where good ideas quietly fail.
Feedback arrives too late
By the time the comments arrive, the post is already live and the first impression is already formed.
Generic AI rewrites aren't enough
Writing better isn't the same as understanding how people interpret it. Polish doesn't tell you who you lost.
Product
See the reaction before the reaction.
Audience simulation
See the spread, not a verdict.
ICON models a distribution of reactions across a simulated audience — the people who love it, the people who scroll past, and the ones quietly unconvinced.
Simulated reaction spread
- Strong positive28%
- Curious31%
- Neutral19%
- Skeptical17%
- Negative5%
Comment simulator
The comment section before it exists.
Ten to fourteen synthetic replies from different audience segments, including the ones that push back. Labelled by segment, never by fake name or face.
The last line is the whole post.
six months though. was that learning or scope creep
Do you have a writeup of the three versions? Would send to my team.
Friction scanner
Find where you lose people.
Weak openings, missing proof, buried payoffs and tone mismatches — each with why it matters and what to do about it. If nothing is wrong, ICON says so.
Your strongest sentence sits in the third paragraph, where the fewest people reach it.
Fix. Cut the middle beat and let the closer land second.
Rewrite engine
Three ways to make it land better.
Sharper, bolder and cleaner versions that keep your voice intact. Each one publishable as-is, with a note on exactly what changed and why.
Before
I spent six months building a feature users understood in six seconds. Not because they're impatient. Because six seconds is all any of us give anything now.
Sharper
I spent six months building a feature users understood in six seconds — and the thing I cut last was the thing I'd been most proud of.
A/B battle
Put two versions in the ring.
ICON picks a winner, argues its case, credits what the other version does better, and writes the hybrid it would actually publish.
ICON picks B
B opens on tension instead of context, so the first line does the work the second line was doing in A.
What A does better: it earns more trust with skeptics.
The lab
Run your first simulation.
No account, no card, no setup. Paste something you're about to publish and see it come back with a report.
What are you about to publish?
0 / 12000
ICON's AI provider isn't configured on this deployment.
Configuration needed
ICON's AI service isn't connected here.
Simulations run entirely server-side against a configured provider. Add an ANTHROPIC_API_KEY, OPENAI_API_KEY or OPENROUTER_API_KEY to this deployment and restart. ICON will not fabricate a report without one.
Everywhere you publish
One preflight for wherever you publish.
- X
- TikTok
- YouTube
- Newsletters
- Ads
- Web
Same audience intelligence. Different publishing context.
How it works
Three steps, no setup.
01
Add what you're publishing.
Paste text, drop a link or upload the creative.
02
Tell ICON who needs to care.
Platform, goal and audience. Twenty seconds, all optional beyond the basics.
03
See how it may land before going live.
Scores, reactions, friction, rewrites — in one report.
Example simulation
What a report actually looks like.
A real ICON output for a post about shipping. Included as a worked example — not live measured data.
I spent six months building a feature users understood in six seconds. Not because they're impatient. Because six seconds is all any of us give anything now. We rebuilt it three times. The version that worked was the one with the least in it.
The opening line is doing almost all the work, and it is doing it well. The drop-off risk is the middle: line two states something your audience already believes, which spends attention without adding anything.
Simulation confidence: medium
Three of thirteen simulated comments
Builder
The last line is the whole post. Least in it wins, every time.
Skeptic
six months though. was that learning or was that scope creep with a nice ending
Design lead
Rebuilding three times is the part nobody puts in the case study. Respect for saying it.
One friction point
Buried payoff
The strongest sentence — 'the version that worked was the one with the least in it' — sits at the very bottom.
One rewrite
I spent six months building a feature users understood in six seconds. The version that worked was the one with the least in it — and the thing I cut last was the thing I'd been most proud of.
Pricing
Pay for intelligence, not another subscription.
Free Beta
5 simulations / day
$0
- Full audience simulation report
- Synthetic comments and segments
- Friction and misinterpretation scans
- Three rewrites and five alternative hooks
- A/B comparison mode
No account. No card. No wallet.
Start freePay as you go
Priced per simulation
Coming soon
- Unlimited access based on usage
- Crypto-native settlement
- No subscription, no seat pricing
- Runs you buy don't expire
Founder
Built from years of understanding audiences.
ICON started with a simple belief: creators shouldn't have to publish blindly. AI can give people another perspective before an idea reaches the real audience.
Gia Macool
@GiaMMacoolKnow how it lands before you post.
Your audience will tell you eventually. ICON tells you first.
No account required · Free during beta