Is Facebook Killing Car Groups with AI Chatbots?

When we last spoke with ChatGPT about classic cars, it recommended we “join online classic car forums and communities to connect with other enthusiasts, share your experiences, and learn from others.” That is sage advice from an AI chatbot. Or at least it used to be.

Those of us who use Facebook Groups to connect with other enthusiasts are likely noticing a disturbing new trend: the unprompted implementation of artificial intelligence from Facebook’s parent company, Meta. AI technology has benefits we may never see, but Meta might kill many a Facebook group with their new AI chatbot technology. You know, just like how they killed the car forum.

Let’s start with the chatbot itself, as its implementations are tone-deaf for places like Hagerty’s Radwood Facebook group. RadRetro is the chatbot’s name, but it doesn’t fully understand that this group is about retro cars from the 1980s and 1990s. Instead of looking like something from Miami Vice or a movie like E.T., the RadRetro avatar looks like it’s from The Lord of the Rings.

But perhaps I buried the lede of this story, because Meta AI’s chatbot technology isn’t the problem. Instead, it’s the rollout of this technology, crammed down the collective throats of Facebook Group administrators without their consent.

When Facebook decides that AI is needed for a group, they inform the administrator. It comes via pop-up as shown above, and the only way to enter your group is to push the “Continue” button. Don’t want AI? Too bad.

While the above prompt suggests Meta AI can be customized “to fit your community,” I have yet to find these adjustable parameters. In my time as a social media/reputation manager, Facebook has been nothing if not obtuse for back-end access. They instead focus on the user experience for the general public. So in case you have never run your own group before, here’s a behind-the-scenes look at what Meta AI is doing to Facebook Groups, as told by my experience running the Brown Car Appreciation Society (BCAS).

Perhaps a little backstory is needed: the BCAS was created in February 2012 for the sole purpose of posting pictures of brown vehicles. Joining as a chatbot should be easy, given there’s over a decade of pictures and conversations it can use as research material. Heck, we’ve even used AI to create images of brown cars, including one that turned Ricardo Montalban, our official spokesperson (yes, really), and his Chrysler Cordoba into a freakish, chocolate-toned, Malaise Era centaur.

The BCAS’ Centaur was fun, but the polar opposite happened when Meta AI’s chatbot joined the group. It’s called “Brownie,” and manifested as a Hamadryad with bark for its skin. It’s a bizarre avatar for a group about brown cars, and Brownie’s contributions have been similarly out of step.

Brownie never spent any time reading the room. Nor did it read the rules, as “posts without brown vehicles might be deleted.” I now visit the BCAS on a daily basis to delete the engagement bait that Brownie contributes to the group. (Brownie doesn’t post every day, but I still need to check.)

This AI-generated slop is a far cry from what we saw when Alex Nuñez created the BCAS. You might remember that name from his work at Autoblog.com. I texted Nuñez these screenshots of Brownie’s deep thoughts on brown cars. Here’s his feedback:

“We created the BCAS as a fun inside joke between myself, Sajeev, and the late, great Davey G. Johnson. Thing is, it resonated with people and was a lot of fun. Now, thanks to Meta thinking that AI makes everything better, it’s turning the whole thing into an actual joke. And that’s not fun at all.”

Nuñez is right when he mentioned how BCAS’ content resonated with others, because Jalopnik didn’t get the idea for the brown/diesel/manual/wagon trope by themselves. But no matter how we may feel about Meta AI, the Facebook platform is still a free service. In exchange for hours of free entertainment, they ask for our personal data and the ability to serve advertisements to our eyeballs.

My feelings are just that, and let’s remember that Hagerty is a data-driven organization. So I poked into BCAS engagement statistics after the implementation of the Meta AI chatbot. The metrics are all over the map, thanks in part to our small size, infrequent postings, and lack of engagement in the comments section.

Unless you are a day trader, the sheer volume of spikes presented on these graphs is a bad sign. This is the result of posting infrequently (the spike up) and the dead spots between where member engagement disappears (the fast retreat down). The BCAS needs a regular cadence of posts and interactions to get accurate data, but the additional “content” created by Brownie isn’t helping. It may be successful in creating more spikes of engagement, but the comments and reactions are almost exclusively laughing and angry emojis, and some members report these postings to me for deletion. Brownie is seen in the same light as Facebook spammers and scammers, and I have a bad feeling that people will leave the group if this keeps up.

Put another way, things aren’t great now, but AI will make things worse if I don’t step in as the group’s administrator and perform a course correction. I wonder if the Meta AI implementation is solely a pain point for participants in groups that demand a level of decorum, which, of course, isn’t mandatory on Facebook. Perhaps there’s a “garbage in, garbage out” methodology at play to grow Facebook at all costs, and as quickly as possible. This feels like part of their modus operandi, as Facebook makes it clear they like to move fast and break things.

But now Facebook is breaking stuff created by their user base, and that’s not very nice. After over a decade of providing content that keeps people entertained on the Facebook platform, the Meta AI chatbot feels like a slap in the face to early adopters like yours truly. It would be nice if Facebook allowed feedback, but that seems impossible for a mere administrator. While beta testing has existed since the dawn of commercially available software, this AI chatbot experiment could be the start of a downward spiral for Facebook groups. Something needs to change with Meta AI, and it needs to happen quickly.