TL;DR
- BrightEdge co-founder and CTO Lemuel Park published an analysis of 300 million-plus US monthly searches in Search Engine Journal on 18 August 2026.
- Google's AI Overviews cited Facebook 19.5 million times, Instagram roughly 877,000 and TikTok around 78,000 in that US sample.
- On buying questions citing Facebook or Instagram, Google named a big retailer or marketplace about 85% of the time; manufacturers got 3 to 4% of brand mentions and about 75% of cited brands appeared once.
- The sample is US only and the methodology is BrightEdge's own, unverified independently, with no equivalent Thai dataset.
BrightEdge co-founder and CTO Lemuel Park published an analysis of more than 300 million US monthly searches in Search Engine Journal on 18 August 2026, mapping which social platforms Google's AI Overviews pull from. Facebook was cited in 19.5 million AI Overviews, Instagram in roughly 877,000 and TikTok in around 78,000.
The label on that data matters as much as the numbers. This is a vendor's analysis of United States searches, written by BrightEdge's own co-founder and CTO and published in Search Engine Journal. The methodology is BrightEdge's own and has not been independently verified, and there is no equivalent Thai dataset. Every figure below is a US figure. None of them is global and none of them is Thai.
What the BrightEdge AI Overviews analysis counted
The BrightEdge AI Overviews analysis, published on 18 August 2026, covered more than 300 million US monthly searches and recorded which social platforms appeared as cited sources inside Google's AI Overviews. These are citation counts, meaning the number of AI answers in which a platform showed up as a source. They are not clicks, sessions or referral visits, and Park attached no traffic figure to them.
The counts, as reported by BrightEdge on 18 August 2026:
| Platform | AI Overviews citing it (US, monthly sample) |
|---|---|
| 19.5 million | |
| roughly 877,000 | |
| TikTok | around 78,000 |
From the same sample Park derived a second figure: roughly one in 15 US searches places Facebook or Instagram content inside a Google AI answer.
Google split the platforms by question type
The BrightEdge analysis published on 18 August 2026 found that Google's AI Overviews do not treat social platforms as interchangeable. Each one was pulled for a different class of question, which is the more useful half of the study for anyone deciding where to spend effort.
The intent split BrightEdge reported, platform by platform:
| Platform | Question type Google's AI Overviews pulled it for |
|---|---|
| timely, local and community questions | |
| culture and shopping | |
| TikTok | trends and how-to |
| firsthand experience and troubleshooting | |
| YouTube | step-by-step instruction |
BrightEdge also compared Google with ChatGPT on the same ground. Google leans on location and stock signals, with roughly 11 to 14% of near-purchase social citations being "near me" or "is it open" prompts. ChatGPT barely touches local at all and skews instead to deals and pricing, at roughly 20 to 24% each.
Follower count was not the selector
Two examples in the 18 August 2026 BrightEdge analysis show that audience size did not decide the citation. Google used a single Instagram post to answer "mobile payment app", a query BrightEdge estimates at 18.5 million monthly searches. It used a baseball team's Facebook post to answer "where to watch Brewers versus Reds".
Analysis, not a BrightEdge claim: what those two posts have in common is not reach. Both stated the literal answer, in text, on a public URL, with a date on it. The citation went to whoever had already written the answer down in a form a machine could lift.
At the buying moment, Google's AI names a retailer about 85% of the time
The most commercially awkward finding in the 18 August 2026 BrightEdge analysis sits at the bottom of the funnel. When Google's AI cites Facebook or Instagram on a buying question, it names a big retailer or marketplace about 85% of the time. Manufacturers receive 3 to 4% of brand mentions. About 75% of cited brands appear exactly once.
The intent split sharpens at the same point. Around 90% of bottom-funnel Instagram citations are purchase questions of the where to buy, price and on sale kind. Facebook turns up mostly after the sale, in about 23% of its bottom-funnel citations, more than twice Instagram's rate.
Analysis: taken together those three numbers describe a leak more than an opportunity. Social content supplies the evidence the AI answer rests on, and the answer then hands the shopper to a retailer. A brand can be the subject of a citation without being its destination. The 75% figure says something separate and easy to miss. If three quarters of cited brands appear exactly once, the cited-brand list is long and shallow, so a single citation is the ordinary outcome rather than a position anyone occupies.
Why a public social post works as answer content
This section is mechanism, not a BrightEdge finding. There is nothing exotic about a Facebook post ending up inside a Google AI answer. A public post is crawlable text on a public URL, it names an entity, and it carries a timestamp. That is the same raw material a web page supplies, minus the website. For questions whose answer changes week to week, such as opening hours, stock, a fixture or a price, the post frequently has a fresher date on it than the business's own site does.
The same mechanism explains the intent split without any special machinery. Facebook carries local and timely material because that is what people post there. YouTube carries step-by-step instruction because that is what is on it. The platform is not being ranked for its brand; it is being read for the kind of sentence it happens to contain.
One in 15 US searches is a derived ratio
Label the one in 15 US searches figure correctly before repeating it anywhere. BrightEdge presented it as a ratio derived from its own sample of more than 300 million US monthly searches, not as a separately measured quantity. It merges Facebook and Instagram citations into one number, it is US only, and it depends entirely on how BrightEdge defined a citation and which queries entered the sample. As a summary of that sample it is reasonable. As a general statement about search it is not, and it should not be quoted as one.
What a brand controls here, and what it does not
Analysis. The controllable part is narrow and concrete: the specificity and factual completeness of a brand's own public posts. If a page's posts state the opening hours, the price, whether an item is in stock, which branch has it and what the product actually does, in plain text rather than baked into a graphic, then that text exists as an answerable statement. If the same information lives only inside an image, only in the comments, or only in a story that expires, it does not exist in a form anything can quote.
The uncontrollable part is everything downstream of that. Nothing in the BrightEdge analysis describes a way to request, submit or opt into a citation, and nothing in it explains how Google chooses between two posts that both answer the question. Treating post quality as a lever is defensible. Treating it as a guarantee is not, and any promise built on these figures is a promise about somebody else's system.
What the BrightEdge analysis did not say
- It did not describe a ranking mechanism. The counts say which platforms were cited, not why one post was selected over another.
- It did not offer an opt-in, a submission route or any markup for getting a post cited.
- It did not report traffic. A citation count is not a visit, and no click-through number was attached to the 19.5 million.
- It did not include a Thai or wider APAC breakdown. The sample is United States searches.
- It did not claim the pattern is stable over time. It is one reading of one sample at one point in 2026.
What this means for Thai marketers
Nothing in the 18 August 2026 BrightEdge analysis measured Thailand, and that belongs in the first sentence rather than a footnote. What follows is reasoning about the mechanism, not a finding from the study.
Thailand is a Facebook and LINE first market, and local business information here genuinely lives in Facebook posts and comments rather than on a website. Plenty of Thai small and mid-sized businesses publish their hours, their prices, their new stock and their branch details on a Facebook page and nowhere else. If the mechanism behind the US pattern is that Google pulls the freshest public text answering a local or timely question, then a market where that text sits mostly on Facebook is a market where the pattern is plausibly stronger. Plausibly is the operative word. Nobody has measured it here, and this post is not going to pretend otherwise.
The practical read is to stop treating a Facebook page as a broadcast channel and start treating public posts about hours, stock, price and common product questions as answer content, written as Thai text rather than as a graphic. That is the same discipline behind generative engine optimisation, applied to a surface most brands hand to the social team and never look at again.
The second read concerns marketplaces. If the US bottom-funnel pattern carries over at all, the assumption that a Shopee or Lazada listing will be the thing Google cites deserves checking rather than assuming, because in the US sample the citation named a big retailer far more often than the manufacturer. For brands whose demand generation runs mostly through paid Facebook activity, the organic page is doing quiet work that no campaign report will show.
FAQ: Google AI Overviews and social citations
Does this mean Facebook posts now rank in Google?
No, and the analysis did not claim that. BrightEdge counted how often a platform appeared as a cited source inside an AI Overview, which is a different thing from a blue-link ranking, and it did not describe any ranking mechanism behind the selection.
Is this happening in Thailand too?
The source did not say. The sample was United States searches only, BrightEdge published no Thai or APAC breakdown, and no equivalent Thai dataset exists, so any statement about Thai AI Overviews here is reasoning about the mechanism rather than measurement.
Do I need a large follower count to get cited?
Not on the evidence given. BrightEdge reported that Google used a single Instagram post to answer "mobile payment app" and a baseball team's Facebook post to answer a where-to-watch question, and it stated that follower count is not the selector.
Can I submit my posts or opt in to being cited?
No route of that kind was described in the source. The analysis reported what Google cited, not how to be cited, and it named no submission process, markup or setting that would make a post eligible.
Do I actually have to do anything?
Nothing is required, and no deadline exists. The defensible action is small: make sure the answers people ask a business for, such as hours, price, stock and branch, appear as plain public text on the brand's own social pages instead of only inside images or expiring stories.
If you want a review of which public content a brand actually has available for AI answers, and where the gaps sit between the website, the Facebook page and the marketplace listing, Relevant Audience covers that work under GEO and AI search.
