5 Lessons Learned from Growing an AI Facebook Page

A Bit About the Page and Concept

Before diving into lessons learned, I should take a moment to explain the account that was created, why we used AI, and the overall intention.

I will be the first to admit that I find AI pages to be, on average, terrible. But I also recognize that Facebook doesn't seem to care about “quality” content anymore and that the algorithm favors simply whatever gets engagement (aka the new definition of “quality content”).

So I wondered, could there be a use of AI content where I could check these boxes, perhaps join the social bonus program, and make a bit of money while using AI in my favor? How do you do that on the more “ethical” side of things? Can you do that on the ethical side of things? All these and more were things I wanted to find out.

So we created an account that we thought followed the line of ethics by reimagining art that should be in the public domain in styles of other artists that should also be in the public domain. Above the board? We hoped. A neat idea we had never seen before? We thought so. Overlapping the topics that everyone may be interested in with viral potential, just what Facebook is loving these days? You know it.

So Remix the Masters was born, with the goal of creating “engaging, quality content” with AI.

On one hand, the idea took off rather quickly, and we started hitting good growth milestones with only a meager advertising budget (6,000+ followers in 90 days!). But, on the other hand, we also learned a lot about growing a new account on Facebook at the same time. So in this one, let's dive into some lessons learned from the experience here.

• Note: We get that AI content is generally very polarizing. We fully agree with most negative sentiments against AI content and never use it to create content for brands that rely on our personal expertise. This post is not going to dive into the merits or drawbacks of AI content on Facebook too much, and you're more than welcome to judge this test however you see fit. To repeat, we get it.

Accounts Reached From Ads Were… Questionable

Now, although the content on this Facebook page was 100% AI-generated, I did not use any shortcuts to grow the presence of the page- only native ads on Facebook and Instagram, and no buying of fake followers/engagement. Everything for growth was using Meta-provided tools, as intended.

Here, I noticed a few things of significance.

First, Facebook's AI-generated text for A/B testing of ad copy is consistently terrible. It feels very flashy, oh-my-god-wow hype style, and not at all interesting, but if their data suggests that works, well, I was willing to give it a go for this account (but definitely not for my real pages). See an example below:

Does it work? Well, we got a lot of engagement on the boosts. But also, we got a lot of engagement on boosts with our own copy, too. Personally, I just don't like the AI-generated copy solely because it is a hype machine to the max.

Second, when geotargeting broad swaths of the world (namely, the USA, Canada, and the EU as a whole), it felt like the ads optimized around the lowest cost-per-engagement countries within the target set, and almost always optimized around Eastern Europe.

That said, when we only tested ads in the USA and Canada, performance was also not nearly as good- giving further credibility to this lowest-common-denominator targeting. I've seen this happen many times over the last decade on Meta ads, and I'm not sure whether or not this is a good thing. I would like my ad to reach people across that entire geo set, not just the cheapest one, in normal cases at least.

So if you're targeting large areas with ads, you may want to be careful when defining your dataset.

Finally, engagement from these ads was, for the most part, very bot-like. I am not sure whether or not the people reached are real or not, but when you get a surge of “Amen!” comments on a post of a work of art, in multiple languages, and only very loosely having a religious sub-context at all (and also, may we remind you, being completely fake/AI-made), it makes you wonder- especially when we did not ask for a “can I get an Amen!” or anything like that.

I don't have enough data to make a definitive conclusion, but it just felt, shall we say, weird. (When I say a surge of these comments, I mean hundreds across several posts, not dozens.)

Still, in three months and only a few hundred dollars in ad spend, we grew the page to over 6,000 fans with moderate organic engagement and some semi-viral shares within that. So, while it is still pretty easy to use Facebook advertising to grow a page and generate engagement, we have to wonder if there is a fine line between good engagement and engagement-at-all-costs, with the former being much, much harder to achieve and the latter being far too easy.

Many Users Can't Tell the Difference Between AI and Real

Perhaps the most disturbing thing that we noted was that many, many users simply couldn't tell that images were AI. That, or they just don't care.

The most common comments that we could identify as “real” that actually discussed the posts were things like “beautiful painting” or something of that nature, even though, you know, we have our profile marked as being AI-generated. While it is likely that people do recognize that it is, naturally, fake, some comments really didn't seem to act like they could tell- which is very frightening.

In the nearly 200,000 views generated, across all ads and organic performance in the first month, we had only one, yes, just one person call us out for using AI content in a way they didn't agree with. Their comment was equally silly, but was the only time we really thought someone gave what could be considered a rational thought. (In the next two months, and a few million more impressions later, we saw no such comments.)

So, to continue the commentary here, that was depressing.

Facebook Spam is Abysmal

Now, I know what you're thinking, isn't an AI account spam too? We're not going to argue this one, but one thing we did notice after running a new account is just how bad Facebook spam, particularly messenger spam, has become.

We all know the scam by now, where accounts send you fake messages saying your account is in violation of some nonsense rule and is about to be closed unless you sign over all of your account details to an obviously fake 3rd party website.

We get these all the time on our other pages, but it wasn't until starting this AI account that we realized just how bad it could be. In the first month alone, we received over 100 messages attempting to gain access to our account, all unfiltered and delivered by Facebook just fine.

For me, as someone who has been running Pages on Facebook for the better part of 20 years, this is par for the course. But for someone who may be new in the content creation space, this level of spam is daunting and, honestly, is completely unacceptable. Meta really needs to get its act together as the onus shouldn't be on the user to filter this crap out. That's just poor service (but, then again, users are the product, so we often get no support even at the best of times).

The ironic thing here, however, is that Instagram does not have this issue nearly as much. While we get similar fake messages on Instagram left and right, they clearly have better filters set up to catch these messages and automatically move them to hidden folders. So where we would say that Facebook's filtering is good 0% of the time, we'd argue Instagram is actually pretty good at catching 90%+ of spam.

The irony of being owned by the same company!

Organic Growth Milestones

Whenever I start a new Facebook page, I always wonder how large an account has to be before you can start seeing good, organic growth from shares from real people. Advertisements can be good to get the initial audience, but at some point, I also want to stop spending money and let the account grow organically from there.

Can your page could blow up with 100 followers? I won't say it is impossible, but it is just unlikely. Do accounts that have thousands or tens of thousands of followers have a leg up here? Absolutely.

A decade ago, when Facebook had less algorithmically controlled feeds and people simply saw content they chose to follow, it was fairly predictable. If you could use advertising and external means to build a Page to 2,500 to 5,000 real and engaged followers, the stage would be set for a good piece of content having the chance to go viral. This happened to me on my local blog, partly because whenever someone shared their friends also live in the city and liked the content, and the rest is history.

Now, that logic may not hold as much water, as the recommended feeds decide winners for everyone (and the industry is still trying to figure out what it takes to get into that). There is some logic that things that are engaged with are more likely to get put into the recommended feeds, but it isn't as black and white as it was a while back.

So, in this one, we can only offer some initial observations.

First, organic engagement on new posts was slow (borderline non-existent) until we got to somewhere between 1,000 to 2,000 followers. This may be a sign that the people who followed us from advertising campaigns just didn't care that much, or it is a sign that is just how the algorithm is anymore with the current, abysmal organic reach for non-recommended accounts. While I lean to the latter, honestly it is hard to say.

Second, after about 2,000 followers, we started to see some decent movement organically, with more comments and shares without ads; however, our best performance was still by throwing $5 on any given image as it would blow up pretty quickly from there.

Third, Facebook groups are an underutilized resource for sharing content and getting free, organic growth. There are groups for seemingly everything, AI art included, and we piggybacked off of some very large groups that allowed shares to help amplify our reach. This is something we think others may not be tapping into as much as they should, so find some groups to share your content into if there is a fit. Their group gets more content, you get access to extra reach on day one- a win-win if you can find a match!

Finally, our ultimate goal here is to see if we can get into a social bonus campaign, and as of now at 6,000 followers, we have not received any invitations. We will update if we ever get into such a program.

Facebook Performed Better Than Instagram

This is just a quick observation, but we noticed that Facebook was far and away the better performer for this kind of content in all cases (organic, paid ad, etc.).

Now, we can speculate as to the many reasons why this may be the case, ranging from just a different user type on Facebook, the native share features letting posts blow up (although Instagram may be getting that soon), groups amplifying reach, AI generated copy in bosted content being more “engaging”, the invite button helping get follows after engagements, or fake accounts possibly plaguing the network.

These all are possible explanations for why the Facebook account grew the way that it did.

But, when it is all said and done, we took the same content, with the same copy (pre-ad, at least), boosted to the same audiences, for the same amount of money, on two different networks, and Facebook outperformed Instagram by about 10:1 in virtually all respects.

Could Instagram really just favor Reels that much more than images? Most likely. Is the kind of audience just not on that network? Unlikely. Or is it something more nefarious about how the algorithms work and/or the quality of users on such networks? I'm not saying we ascribe to the Dead Internet Theory, but if there is a social network that may be bot-driven more than others, we would very much nominate Facebook for the category. The accounts we reached with very real ad targeting were questionable at best.

Still, for these mega apps with huge membership numbers and very similar advertising methods, it was a night and day difference between the two. Next up, we're going to test some Reels in a similar style and see what happens as we expect Instagram to perform far better in that regard.

Facebook Should Care More About AI

This one is not a lesson learned for us, but perhaps more a lesson learned for Facebook (and most social networks) in particular.

When I started this AI account, I took a dive into the requirements for disclosures for AI-created posts. I was pretty shocked to learn that most Meta products don't actually have much in the form of requirements for AI-created images, despite many public articles saying the requirement is more or less across the board. The main distinction is “photorealistic video”, “realistic-sounding audio”, and the like.

The policy in the Instagram Help Center went on to say “Meta does not require you to label images that have been created or altered with AI. Images will still receive a label if Meta's systems detect they were AI-generated.”

So, were any of my posts flagged as created with AI even though our account bio literally states “AI created”? Absolutely not. Not even once. Not even when we boosted them. Never.

Here we have an account that is 100% AI-created, disclosed in the bio, and the crawling apps don't bother with it. Not only that, Meta ads accepted every single boost without adding disclosures either.

Well, that's depressing.

But, I suppose in a way, the policy also makes a bit of sense. The distinction of “photorealistic” and other similar terms here implies that the biggest concern that Meta has about AI is deepfakes of real people, and everything else just goes into the engagementbait machine that social media has become. Still, the bar feels very, very low here, and yet we still don't have great requirements.

Overall, we learned a lot about where Meta is heading with regards to content creation in the era of AI, and most of it is pretty depressing. Although we are viewing this page as more of a dead-end product simply to test some theories we had about Facebook et. al., we will be continuing it just to keep the what-if going- because if we can get a new account on a social bonus program and it only takes 5 minutes a day of work? Well, why not?

What trends are you seeing in running social media in the age of AI? Comment below to share!