There was a time when producing something that looked authoritative took effort. A real-looking news article, a believable screenshot, a video of a known person saying something: each took skill, tools, and time. That friction did quiet work for us. It meant most of what we saw had at least passed through some hands.
That friction is gone. A convincing fake of almost anything is now a few seconds of effort. The result is not that everything is fake. It is that looking real no longer tells you anything. The weight has shifted from "does this look legitimate" to "can I confirm this from somewhere I already trust."
That is a learnable skill. Here is the method.
Why your instincts are now miscalibrated
For most of your life, polish was a reasonable proxy for legitimacy. Clean design, fluent writing, a real-sounding quote, footage that matched the story. Your brain learned to relax when those signals were present.
Every one of those signals can now be manufactured for free. So the instinct that says "this looks professional, it is probably real" is actively working against you. The fakes are specifically good at the surface, because the surface is the cheap part.
In the AI era, how real something looks tells you almost nothing about whether it is true. Polish is cheap now. Stop using it as evidence. The only reliable signal is whether you can trace the claim back to a source you already had reason to trust.
The method: trace, don't trust
When something matters enough to act on, especially anything touching money, accounts, or a decision you cannot easily undo, run this. It takes well under a minute once it is a habit.
- Find the original source, not the messenger. A screenshot of an announcement is not the announcement. A video clip is not the press release. Go to the actual origin: the project's own verified site, the official account, the primary document. If a claim cannot be traced to a real source, that absence is your answer.
- Confirm it in a second, independent place. One source can be wrong or fake. Two independent, credible sources reporting the same thing is a different level of confidence. Independent matters: ten accounts copying one rumor is still one source.
- Check the door, not just the page. For anything asking you to log in, send funds, or connect a wallet, the web address is the real identity, and the page is just paint. Type the address yourself or use a bookmark you saved earlier. Never reach a sensitive destination by clicking a link you were handed.
- Ask what the source wants. Every message has a motive. Someone urging you to buy now, move funds, or act before a deadline has an interest in your speed. That interest is information. Slow down in exact proportion to how hard you are being pushed to hurry.
The shape of this is simple: go upstream to the origin, confirm it twice, reach sensitive places through your own door, and notice who profits from your haste.
Screenshots of messages, balances, endorsements, and news are among the easiest things to fabricate, and they spread fastest because they feel like proof. Treat any screenshot as a claim to be verified at the source, never as the verification itself.
Using AI to check AI, carefully
AI assistants can help you verify, but only if you understand their limits. A chatbot can confidently state something false, present a real-sounding citation that does not exist, or be out of date. It is a research assistant, not a witness.
Useful ways to use one:
- As a lead generator. Ask it where an official announcement would live, or what the real web address of a service is, then go confirm that yourself at the source.
- As a skeptic. Paste a suspicious message and ask what techniques it uses and what it is trying to get you to do. Models are good at naming the pattern, which can break the spell of a convincing one.
- As a translator of jargon. Ask it to explain a confusing term in plain language so you can evaluate a claim on its merits.
What you do not do is treat its answer as the final fact. The model points you toward sources. The sources, checked at their origin, are what you trust. We go deeper on the risks of trusting model output blindly in the prompt-injection lesson.
Make it a reflex
You will not run a four-step audit on every meme you scroll past, and you should not. The method is for the moments that matter: a message about your money, an account warning, an investment that wants a decision today, a video that would change how you act if it were true.
For those moments, the reflex is short enough to remember under pressure: trace it to the source, confirm it twice, and reach anything sensitive through your own door. Practice it on low-stakes things so it is automatic when the stakes are high.
Verification is the core literacy of this era. When something matters, trace the claim to its original source, confirm it in a second independent place, and reach any login or payment through an address you typed yourself. Treat AI as a tool that points you toward sources, never as the source itself.