It takes hard work to engage with all the people in your life. Especially now that social media has crammed an unmanageably large number of people into our lives.
A new app puts the burden of social media engagement into AI’s hands, so to speak, but using it may leave you feeling more disconnected than ever—and coming across that way too.
On Tuesday, Nilan Saha, Vancouver-based CTO of the wellness app Juna, tweeted a video demo of his latest project: Magic Reply. It’s a Chrome extension that uses artificial intelligence to instantly generate replies to posts on X/Twitter and LinkedIn. Whether the replies are indeed “humanlike,” as the landing page trumpets, is in the eye of the beholder.
Saha refers to the $10-a-month extension as a “growth tool,” the idea being that effortlessly replying to endless posts will balloon one’s following. As every bog-standard social media advice guide advises, a popular way to get others to engage with you is by first engaging with them. But many of those on X responding to the demo, which has more than 5 million views so far, seemed skeptical of whether this task can or should be outsourced to AI.
I decided to give Magic Reply a test run to see for myself.
The first reply I generated was to Saha’s announcement post. “Wow, that sounds cool,” I was urged to say. “Can’t wait to see what you’ve created!” Magic Reply either did not sense the attached demo video, or had no way of assessing its contents. The suggested reply seemed not to know anything about what Saha’s creation does, despite the video’s unmistakable depiction. While this reply made me sound a little slow on the uptake, it was worded vaguely enough to probably pass muster on a quick glance.
Most of the other replies on X took a similar tact—positive, enthusiastic echoes of affirmation. In response to a post highlighting a specific aspect of the friendship between Larry David and the late Richard Lewis on Curb Your Enthusiasm, Magic Reply offered: “That dynamic between Larry and Richard is gold, always cracks me up.” Replies like these are human-passing, but add nothing to the conversation, and often miss the point. Then again, a lot of artisanally human-crafted replies from strangers often do the same.
Sometimes, the echoes felt distinctly machinelike, though. An X friend was hyping an article about the Criterion Collection, and highlighted a passage in which director Michael Bay finally addressed Ben Affleck’s infamous commentary track for their film, Armageddon. The response seemed unsure of which part of the tweet was most relevant or why, let alone how I was supposed to feel about it. (“Sounds interesting, Michael Bay commenting on Ben Affleck’s ‘Armageddon’ track. Will check it out.”) If I had used this reply, my friend might have considered that I had been Body Snatcher’d, or at the very least, hacked.
The dissonance between post and reply—and between me and the AI’s idea of me—was even more pronounced on LinkedIn, where the content tends to have a confident, influencer-y bent. “Saying an intern can run your social media marketing efforts is like looking at a piece of modern art and saying ‘my kid could paint that,’” read a post in my feed. The way Magic Reply responded suggests it does not understand similes. “True, interns lack experience and nuance, not the best for social media,” went the recommended reply. It sounded like an intern-hating extraterrestrial, one incapable of understanding the irony of its describing someone else as lacking the “experience and nuance” that is “best for social media.” (It would later offer this paradoxical bon mot for another post: “Authenticity is key.”)
While many of these replies were destined for eye rolls, some seemed like they might end up getting me blocked. In response to a public-speaking coach’s lengthy advice for how to handle big presentations, Magic Reply mirrored some of the post’s language and then signed off with “Good luck on your presentations!” It was ill-equipped to grasp that presentations are a public-speaking coach’s bread and butter, and that a good one would be well past needing luck with them.
There’s a word for tone-deaf, soulless online dispatches that add nothing but suspicion, and that word is “spam.” Most people hate it, especially on social media, where they actively look for ways to mute the incessant Greek chorus proffering crypto and porn. Elon Musk even claimed, many times over, that part of his motivation for buying Twitter was to rid it of spambots, although those appear to not be going anywhere.
The whole point of engagement is to build affinity, but the only thing Magic Reply’s posts seem capable of building is awareness. Each of them might as well merely read, “Hey, don’t forget that I exist.” At least traditional spambots have the decency to be obvious plants. These replies instead approximate digital rapport by adopting a tone of fake familiarity, like those “hey girl” posts from human friends trying to rope you into multilevel marketing scams.
Even if Magic Reply got better at creating “authentic,” additive responses a few versions down the line, though, it would still be a disturbing and dishonest practice. Presenting as so interesting and affable on X that people want to hear more from you is a valuable skill. Carpet-bombing social feeds with AI replies is not just a poor cheat code for that, it’s an affront to basic social graces. It’s even more coldly impersonal than that software engineer who went viral earlier this week for wearing a Vision Pro to his own wedding.
We are wading further into Black Mirror territory here, outsourcing entire aspects of communication. Why not just get a bot to replace all social media activity—leave the engagement to AI, and disengage entirely? Our bots could make pithy observations about the day’s events and wish each other “good luck on the presentation!”
So much noise, with no one actually listening.