We added the "share verdict" feature to verifAI because users kept asking for it. The most common workflow on the product is: paste reel, read verdict, want to do something about it. Posting a link to the verdict in the reel's comments is the obvious thing to do.
What surprised us is how big a difference small changes make to whether anyone reads it. After watching hundreds of these shares unfold in the wild and reading a fair amount of the research on persuasion, here is the short version of what we wish we had known earlier.
What doesn't work
Three things that feel right and almost never persuade anyone:
1. "This is wrong" with a link. This is the default. The problem is that on a phone, a bare link looks like a banner ad. Most readers scroll past it without even registering what they skipped. The link isn't clickable in Instagram or TikTok comments either, which compounds the issue.
2. Sarcasm or condescension. "Lol this is so wrong, please read a book." Even when the underlying point is correct, the framing puts the audience in a defensive posture before they have read the evidence. Nobody updates their beliefs in response to being insulted.
3. Dumping the whole verdict. Pasting the full text of a fact-check into a comment makes it look like a wall of text from someone with an agenda. Comment platforms are designed for short interactions; a long block reads as off-platform.
What works better
A few small changes shift the ratio meaningfully:
Lead with the specific claim, not the verdict
Quote the exact sentence from the reel that you're checking. This anchors the comment to the original content and makes it clear you actually watched it. Something like:
The reel says "X causes Y in 80% of cases". That number comes from a 2019 study with 27 participants — here's the actual paper: [link].
The reader can see the specific thing being challenged before they decide whether to engage. They are also being asked to engage with a specific number, not with the political framing around it.
Cite one source, not five
Counter-intuitively, more sources lower the apparent credibility of a comment, because it looks like you're padding. One well-chosen primary source (the original study, the official press release, the actual law) outperforms a list of secondary aggregators every time. If you have a verifAI verdict link, that is fine as the second link — but the primary citation should be primary.
Match the register of the platform
A comment on Instagram should sound like an Instagram comment, not like a Reddit post. Short, conversational, no headers, no bullet lists. The same content rewritten for TikTok should be shorter still — the median TikTok comment people actually read is one sentence.
Post within the first 6 hours
Reels and TikToks have a sharp engagement curve. By 24 hours, the comments section has stratified into camps and your message is unlikely to be the one that changes anyone's mind — it will get lost. Posting early, when commenters are still arriving rather than reinforcing, gets disproportionate visibility.
Don't argue back
If your comment provokes a reply, replying again rarely helps. The first comment is for the silent reader who is browsing the comments and is undecided. The replies become a back-and-forth that everyone scrolls past. Say your piece once, leave the link, move on.
What the research actually says
If you want to go deeper, two findings from the persuasion literature that translate well to social comments:
- Inoculation effects are real. Forewarning a reader that a video uses a specific rhetorical technique ("notice how the host front-loads three personal anecdotes before the claim — that's a classic credibility-borrowing move") tends to make them more resistant to the same technique elsewhere, even when they don't change their mind on the current claim. This is more powerful than rebutting individual claims one by one.
- The "backfire effect" is much weaker than it was sold. The original studies that suggested correcting misinformation entrenched it have largely failed to replicate. People do, on average, update toward the truth when given evidence — they just update slowly, in private, and not in the way that makes you feel rewarded for posting. You probably will not see the update. That doesn't mean it didn't happen.
So the practical version: post once, with one good source, in a tone that doesn't insult, in the first few hours, and accept that you won't see most of the effect.
What we built into verifAI for this
The verdict pages on verifAInow.es are deliberately designed to look acceptable when pasted into a comment:
- The URL is short, the OG card is clean, and the headline is the claim itself — not "verifAI verdict #123".
- Each verdict has its own permalink, so you're linking to the specific claim, not a wall of them.
- The page itself is readable without an account. You don't want your link to dead-end in a sign-up wall.
These choices come straight out of watching how people actually use the share feature in the wild and noticing where the friction was. There's still a long way to go — we'd like to expose Open Graph images per verdict and improve the embed appearance on each platform — but the basic shape works.
The honest part
Posting fact-checks in comments will not, on its own, fix the misinformation problem. Most of the people who see your comment have already made up their mind. The ones you do reach are the ones who were undecided and looking — and there are more of those than the comments section makes it feel like.
So the goal is not to win the comments. The goal is to leave a clean, short, credible link for the next person who scrolls past, hesitates, and clicks. If you've done that, you've done the work. Whether or not the original poster responds is not actually the point.
You can paste a reel into verifAInow.es any time and the verdict page is yours to share. If you find yourself wishing the share flow worked differently — different copy, different open-graph card, different timing — let me know at fernandoruedaoliva@gmail.com. The "share verdict" UI was the last thing we built and is the part most ready to be rethought.