"How the hell is AI citing Facebook?"
Facebook is the most-cited social network in Google's AI answers. Here's why, and what that means for your town's groups.
Last week's issue on how AI decides which companies to recommend got the best response I've had in months. Same question came back over and over: how is AI citing Facebook?
So I went digging. Here's what I found.
The number
BrightEdge ran an analysis across roughly 300 million US monthly searches.
Facebook appeared as a source in 19.5 million Google AI Overviews. Instagram, 877,000. TikTok, 78,000.
That's 22x Instagram and 250x TikTok. About one in every fifteen US searches now puts Meta content inside a Google AI answer.
The platform everybody wrote off is the most-cited social network in AI search.
Why: one word
Public.
AI can only cite what it can read. Google's crawler doesn't have a Facebook account. It can't see your friends-only posts. It can't get into a private group. It can't watch a Reel and understand what you said out loud.
What it can read is everything Facebook serves to a logged-out visitor. That's a mountain of content.
Readable:
- Public business Pages — posts, photos, About section, hours, Recommendations
- Public Groups — every post, every comment
- Anything posted with the globe icon instead of Friends
- Text. Always text.
Invisible:
- Private and members-only Groups
- Friends-only or restricted posts
- Anything behind a login
- Video with no text
That last one is the one people get backwards. A Reel with "check this out" underneath is a blank page to a machine. The caption is the part that gets read.
Facebook's lane is your town
Each platform gets cited for different things. Instagram for culture and shopping. TikTok for trends. Reddit for firsthand experience. YouTube for how-to.
Facebook gets cited for local and community questions.
Which means the most valuable page on Facebook for your business probably isn't yours. It's your town's group.
Somebody posts: "Cleaning out my mom's house in Norwalk this weekend, anyone know a good junk removal company?" Eleven neighbors reply. Two say they used you and you showed up on time. One says avoid the other guys, they no-showed twice.
If that group is public, that thread is a permanent, crawlable document about who your town trusts. It sits there for years. When someone three towns over asks an AI who's the best junk removal company near them, that thread is exactly what the machine reads before answering.
That's not a social media post. It's a review that got written down where a computer could find it.
Go check whether your local groups are public. Five-minute job, and it tells you whether years of neighbors recommending you are working for you or sitting in a locked room.
Your Page is the corroboration
The Page isn't the star. It's what the AI checks to confirm you're real after it finds your name in a group thread.
Recommendations. Facebook's version of reviews. Public, text, on a domain the AI already trusts. If you've got 200 Google reviews and 6 Facebook Recommendations, you're leaving proof on the table.
About section. Full service list, service area spelled out by town name, phone, hours, website link. This is where your NAP consistency gets checked. Sloppy here and you look like a different business than the one on your website.
Captions in words. "Cleared a 3-bedroom estate cleanout in Trumbull today, full house, garage and shed, two loads" is a sentence a machine can read. A photo with a fire emoji is not.
Answer questions publicly. Someone asks what something costs, answer with the actual answer. Public, text, specific. Those become citable too.
The double dip
On June 15, Meta launched AI Mode inside Facebook search. Global, both app stores, no phased rollout.
Ask it a question and it writes an answer pulled from public posts, Groups, comments, Reels, and Marketplace listings. Not a list of links.
So the public stuff you put on Facebook now pays twice. Once when Google's AI cites it. Again when somebody searches inside Facebook and Meta's AI answers using it.
Same work, two machines.
One thing to know going in: Meta uses public content to train its AI. Public posts, comments, photos, videos. Not private messages. You can opt out if you want, but opting out is also opting out of the thing that makes any of this work.
The honest part
I'm not going to oversell this.
When Google's AI cites Facebook for a buying question, it leans heavily toward big retailers and marketplaces. Small local brands are the minority of citations. Building out your Facebook presence is not going to flood your phone next Tuesday.
This is proof, not lead generation. It's one more place your name shows up when a machine goes looking to confirm you're a real, trusted, local business. Same job your reviews do. Same job your BBB listing does. A piece of the evidence file, not the whole file.
And the evidence file is the point. In August, Reddit's share of ChatGPT citations went from 3.8% to 0.5% in about four days. An 86% drop. Best current theory is Reddit blocked the crawlers itself. Doesn't matter which — anyone who'd bet their visibility on one platform got switched off overnight.
Spread out. Facebook is a piece that costs almost nothing and that your competitors already quit.
Punch list
- Check whether your local community groups are public. Town groups, buy/sell groups, neighborhood groups.
- Join two and be a neighbor, not a billboard. Answer questions. When somebody asks for a recommendation and a customer tags you, that's the whole ballgame.
- Ask your next 10 happy customers for a Facebook Recommendation instead of only a Google review.
- Fix your About section. Full service list, every town spelled out, correct phone and hours, website link.
- Write captions like sentences. Describe the job in plain words. Where, what, how much work.
- Check your audience selector. If your posts are set to Friends, you're publishing into a locked room.
An hour a week. No money.
The operators who show up in AI answers next year will be the ones whose names are all over their own town, in writing, in public, where the machines can read it. That's being known in your market. The computers just started paying attention.
That's it for this week. Ttyl ✌️
— Justin