Threads gets a Grok clone, and the @meta.ai reply is the citation surface
When the AI's answer is a public post, the post itself becomes the artifact other models will eventually read
Photo by Szabo Viktor on Unsplash
Meta is testing a Threads feature where you can mention Meta AI in a post and get a reply from the @meta.ai account, currently in beta in Malaysia, Saudi Arabia, Mexico, Argentina, and Singapore per TechCrunch. The examples in the article are things like "Why are people talking about the World Cup this month?" and "How are the Knicks doing in the playoffs?" Read past the Grok comparison and there's a quieter mechanic worth thinking about.
the reply is a post
It's a Threads post, with a URL, indexable like any other Threads post, quotable, screenshot-able, and (this is the part) potentially scraped into the next training run or surfaced via real-time retrieval by other models that pull from social.
Grok works the same way on X and you can already see the downstream effect: "Is this real?" replies from Grok get cited in news write-ups and embedded in other social posts. The AI's answer becomes a primary source for humans, and eventually for other AIs.
what this means if you want to be cited
If Meta AI mentions your brand inside a public reply on Threads, that reply is a citation event with a much longer tail than a normal ChatGPT answer, because it's permanent and public. If it gets your brand wrong, that's also permanent and public, sitting on a URL that ranks.
So the question isn't "how do I get Meta AI to mention me" (you can't game an invocation that hasn't even rolled out broadly yet). The question is what Meta AI says when someone in your category invokes it. The examples in the article are all trend and recommendation queries. "Whose Met Gala looks are trending right now?" is a brand-naming question. So is anything about playoffs, restaurants, products, or breaking stories.
one thing to do this week
When the feature rolls out wider, you'll know whether Meta AI's default framing of your category names you, names a competitor, or names nobody, and you'll have evidence to escalate with if it's wrong.
I'm genuinely unsure how much weight to put on this versus the dozens of other retrieval surfaces popping up each quarter, and it's possible Meta quietly kills it the way they've killed plenty of Threads experiments, but the structural thing (AI answers becoming permanent indexed social posts) is the pattern to watch even if this specific product doesn't survive.
the safeguards line
TechCrunch notes Meta AI "notably has stronger safeguards in place than Grok, though it remains to be seen whether it will be prone to similar issues.
