The Buying Moment: How ChatGPT and Google Cite Facebook and Instagram at the Bottom of the Funnel

Our previous analysis showed where Facebook and Instagram surface in AI search overall. This one isolates the prompts that matter most to revenue: transactional and post-purchase questions. The platforms take on completely different jobs, the engines want different evidence from them, and the brands named in the answers are overwhelmingly retailers.

Over the past several weeks we have mapped how AI engines use Facebook and Instagram as sources. The picture has been largely upper funnel: people, news, culture, and everyday how-to. This analysis asks the question that follows: when the user is ready to buy, or already owns the product, where do these platforms play? We isolated the prompts where Facebook or Instagram is cited in transactional and post-purchase contexts, split by engine, and examined what each platform was being cited for and which brands appeared in the answers.

Start With the Split: The Platforms Have Different Jobs

At the bottom of the funnel, the two Meta properties stop looking interchangeable. Instagram is cited almost entirely in the buying moment. Roughly 90% of its lower-funnel citations on Google are transactional: where to buy, what it costs, whether it is on sale, whether a product restocked. Facebook carries the post-purchase load. About 23% of its lower-funnel citations come after the sale, more than double Instagram's share, and the pattern holds directionally on ChatGPT.

The content behind the citations reinforces the split. Within post-purchase prompts, Facebook over-indexes on troubleshooting, with app, account, and device problems representing roughly 15% of its post-purchase citations, about double Instagram's rate. Instagram over-indexes on how-to and usage content at roughly 11% of its post-purchase citations, versus about 4% for Facebook. This is consistent with what lives on each platform: community threads and owner groups on Facebook, tutorial and demonstration content on Instagram.

What We Analyzed

We isolated the prompts where Facebook or Instagram is cited, separately for each engine, and filtered to transactional and post-purchase intent only. On that set, we classified each prompt by the job it asks the platform to do, and we extracted and categorized the brands mentioned in the answers. Every comparison is reported as a proportion within an engine, never as a raw count, because prompt coverage is still maturing and the two engines are measured at different scales.

Data Collected

Key Finding

When Google cites Facebook or Instagram in a lower-funnel answer, about 85% of the time a major retailer or marketplace is named in that same answer. These citations are not trivia surfacing. They are part of answers that recommend where to spend. And the brands receiving those mentions are overwhelmingly sellers rather than makers: product brands account for roughly 3 to 4% of the brand mentions in these answers. The user prompts about a product. The engine cites social content as evidence. The answer names a retailer.

Google Uses Social for Local. ChatGPT Uses It for Deals.

The engines want different things from the same platforms. On Google, roughly 11 to 14% of transactional citations are near-me and store-hours prompts, and where-to-buy and availability questions make up roughly 30% more. Google appears to treat these platforms partly as local and availability signals.

ChatGPT barely touches local, at about 1% of its transactional citations. It concentrates instead on deals and pricing, each at roughly 20 to 24% of its transactional citations, about double Google's rate. The prompts themselves differ in kind: fully formed conversational questions, frequently at the level of a specific product, such as a specific GPU model, tool brand, or sneaker release. ChatGPT figures in this analysis are drawn from a smaller pool and should be read as directional.

The Long Tail Is Wide Open

Brand mentions in these answers are heavily concentrated at the top and heavily fragmented everywhere else. The most-mentioned brands are nearly all mass retailers and marketplaces, but roughly three quarters of the unique brands in the dataset appear exactly once. A product brand is unlikely to displace a mass retailer on a broad deals query. It can plausibly own the answer for questions about its own products, which is exactly the shape of question ChatGPT users are asking.

What Marketers Need to Know

These citations end in buying recommendations. About 85% of Google's lower-funnel answers that cite these platforms also name a major retailer. Social content is functioning as evidence inside purchase guidance, whether or not it was built for that.

On Facebook, get the basics machine-readable. Location pages, hours, and contact information should be complete and current on every store page. Promotions should state the product, price, and dates in the post text rather than only in the creative.

Show up where your owners are. Facebook carries the post-purchase load. When someone asks why a product is not working, the engines are citing a Facebook thread. Publish how-to and troubleshooting content with the product named in plain text, and maintain a presence in the owner groups where your products are discussed.

On Instagram, caption for the transaction. Instagram's lower-funnel citations are almost entirely purchase moments. Name the exact product, where it is sold, and the price when there is an offer. This applies to influencer briefs as well: a tag alone is unlikely to be cited, while a caption with the product name and where to get it can be.

Own your own long tail. Roughly three quarters of brands cited appear once. Availability and pricing content for your own products, published in text the engines can read, is an uncontested opportunity for most product brands.

Technical Methodology

Key Takeaways