By CanaryTape
You've seen them. A screenshot of a trading app, a green number with a lot of zeros, and a caption like "Let Grok bot trade for me for a week" or "Claude bot just 10x'd my account." Thousands of likes. Replies begging for the setup.
I spent years in the Marine Corps and a long time after that in freight. Both jobs teach you the same thing: paperwork can say anything. A bill of lading doesn't prove the truck is full. A screenshot doesn't prove the money is real.
Here's why so many of these posts are fake or misleading, and why the little Community Notes box almost never shows up to warn you.
Community Notes isn't a fact-check team. It's volunteers writing notes and other volunteers rating them. And it doesn't work by majority vote. X's own guide says notes "require agreement between contributors who have sometimes disagreed in their past ratings." [1] The open-source ranking algorithm looks for notes "liked by people who normally disagree." [2]
That's a smart guard against brigading. It also has side effects that scammers love.
Niche posts don't get enough raters. Every note starts as "Needs More Ratings" and stays there until it gets at least five ratings, and a note whose score lands in the middle stays there too. [2] To get past that, it has to win over raters on both sides of their usual split. A political post draws crowds. A crypto bot brag in a corner of finance X often doesn't. Low traffic among raters means notes sit in limbo.
A note needs a source. X's guide lists "cites high-quality sources" as one of the things that makes a note helpful. [3] Now try to source your way out of a P&L screenshot. There's no public record of a stranger's brokerage account. You can't link to proof that a trade didn't happen. "This looks fake" isn't a source. So even good notes struggle.
Delete, repost, repeat. A note is written on a post. When the post gets deleted and a fresh one goes up, nothing carries over by default. Notes on media can follow matching images to other posts, but only from contributors with a track record, and the match needs extra agreement on top of the normal approval. [4] Crop the screenshot, change the caption, and it starts over. A poster can move faster than a group of volunteers.
Engagement pays. X pays creators based on engagement with their posts from Premium users. [5] [6] Outrage, envy, and "how???" replies are all engagement. A fake win that gets 2,000 replies is a paycheck, whether it's true or not.
1. The screenshot. P&L screens are about the easiest thing on the internet to fake. Browser dev tools can edit any number on a web page in seconds. Demo and paper-trading accounts look just like real ones. Mock-up apps and templates exist for exactly this.
2. Only the winners. Even when a trade is real, you're seeing one ticket out of a hundred. Nobody posts the nine accounts they blew up. A single green day says nothing about a strategy.
3. The "AI" label. Slapping a famous AI name on it borrows credibility. There's no way for you to check that any bot made any trade. The name is marketing.
4. The funnel. This is the real product. The post sends you somewhere:
That last one is the worst. A wallet-connect drainer gets you to sign a transaction that hands over your tokens. One click and it's empty. No bot. No trades. Just a thief with a nice landing page.
This is the main point. Community Notes is a useful tool, but it's built to be careful and slow, and it only puts a note on a post when people who usually disagree agree that it should. [1] [2] Silence just means the system never reached a verdict. It doesn't mean the post passed inspection.
Treat an un-noted bot brag the way you'd treat a load with no paperwork. Assume nothing. Check it yourself.
If one of these posts pushes a Solana token, run it through CanaryTape first. It's free and independent, and it shows the on-chain warning signs: bundled launches, linked wallets, creator sells, liquidity, holder concentration, and mint authority.
CanaryTape never asks you to connect a wallet. If anything claiming to be us does, it isn't us.
Not financial advice. A clean score doesn't make a token safe.