Search Engine Journal published a contributor column Tuesday from Lemuel Park, a BrightEdge executive. The analysis is built entirely on his employer’s own data: more than 300 million monthly U.S. searches. The findings have not been independently verified. BrightEdge sells SEO software and generative engine optimization (GEO) tools, the category this kind of citation study is built to promote. Even with that limitation, the counts inside the sample are specific enough to change how a search team allocates social effort against Google’s AI Overviews.

Within that sample, Google’s AI Overviews cited Facebook as a source 19.5 million times. Instagram appeared about 877,000 times, and TikTok about 78,000. Those are counts inside Park’s analyzed set of 300 million searches, not a measure of every AI Overview Google generates anywhere. About one U.S. search in fifteen already draws on a Facebook or Instagram post somewhere inside its Google AI answer, per the same analysis.

The citations skew toward specificity, not scale. An Instagram post was Google’s chosen source for the query “mobile payment app,” a term Park estimates draws roughly 18.5 million searches a month. A local baseball team’s Facebook update served as the source for a query about watching the Brewers play the Reds. A TikTok clip was the cited source when someone searched a Yankees-Dodgers matchup. None of the three accounts won the citation on follower count.

Google’s AI Overviews route different question types to different platforms, per Park’s breakdown:

The split sharpens at the bottom of the purchase funnel. When Google’s AI cites Instagram near a buying decision, about 90 percent of those citations answer a buying question. The question is where to find the item, what it costs, or whether it is currently discounted. Facebook’s bottom-funnel citations skew the other way, appearing mostly after a purchase. About 23 percent of its bottom-funnel citations land there, more than twice Instagram’s post-purchase share, tied to a problem with an order, a return request, or a how-to fix.

Google and ChatGPT lean on that citation pool differently, and Park frames the split as directional since the underlying sample is still maturing. Roughly 11 to 14 percent of Google’s near-purchase social citations answer a “near me” or store-hours question. ChatGPT rarely touches location. It instead favors deals and pricing, representing roughly 20 to 24 percent of its social citations near a purchase. It often names one exact product: a specific GPU, a named tool brand, or a single sneaker drop.

Retailers, not manufacturers, capture almost all of that credit. Google’s AI names a big-box retailer or marketplace, not the company that made the product, in about 85 percent of Facebook and Instagram buying citations. The manufacturer draws only 3 to 4 percent of those mentions. Three-quarters of every brand cited in the sample appears just once. Park reads that as a long tail of product questions still open for a manufacturer willing to answer them directly.

This adds to a pattern our newsroom flagged on August 13. Search Engine Land ran Kevin Indig’s contributor analysis then, showing user-generated platforms holding a larger non-vendor citation share than publishers inside ChatGPT’s SaaS buyer-journey answers. The two data sets do not measure the same thing. Indig’s sample covers one engine and 35,000 ChatGPT citations built around software-buyer prompts. Park’s sample covers Google’s AI Overviews across 300 million general U.S. searches, and different platforms entirely: Facebook, Instagram, and TikTok rather than Reddit, Wikipedia, and LinkedIn. What connects them is direction, not magnitude. Both point at community and social platforms edging into a citation role publishers used to hold alone, without proving the two effects share a size or a cause.

In our own reading, the concrete move for the next quarter is not to post more on any single platform. It means opening the AI Overviews and ChatGPT answers for a market’s highest-volume buying queries, then identifying the exact post or thread that earned the citation. The next step is matching that specific, numbers-first format instead of chasing follower count.

Search Engine Journal published Lemuel Park’s contributor column, citing BrightEdge’s own analysis, on August 18, 2026.