Plenty of local businesses have quietly written Facebook off. Google reviews win the customer's eye, the thinking goes, so why keep the Facebook Page tidy? The answer got more interesting once shoppers started asking ChatGPT and Copilot for recommendations instead of scrolling a results page. Nobody outside these AI companies knows their exact recipe, so treat everything here as informed reading of how the systems behave, rather than confirmed internal fact. With that caveat set, Facebook reviews still earn their keep, mostly for human trust and possibly for a bit more.
What Facebook Reviews Still Do for Humans
A real person deciding between two plumbers often checks Facebook before calling. The Page recommendations, the recent comments, the way a business replies to a complaint, all of it shapes the call. Facebook dropped its old star scale years ago and now shows recommendations, which is a yes-or-no from each customer plus their written comment. For a local business that gets discovered through Facebook and Instagram ads, a thin or empty recommendations tab can quietly cost you a booking that Google never touched. That alone justifies keeping the Page active.
The AI Question, Handled Honestly
The newer question is whether AI assistants read Facebook reviews when they decide which business to name. Here the honest answer is that it depends on the assistant, and much of it is inference rather than confirmed fact. What we can reason about is which underlying sources each assistant appears to sit on top of, because that shapes what it can and cannot see.
Google's own AI answers lean on Google's index and Business Profile data, so Google reviews are the obvious signal there. Microsoft Copilot draws on Bing. Based on how Bing indexes local business listings, and on reports from businesses watching where their mentions surface, Facebook Pages that Bing has crawled can feed into what Copilot sees. This appears to happen because Bing catalogs public local listings broadly, and a well-maintained Facebook Page is a public listing. State that as observed behavior, though, since the exact weighting is not something anyone outside Microsoft can confirm.
ChatGPT and Perplexity-style assistants behave differently again. When they browse the live web, they can in principle reach a public Facebook Page. Whether a given answer actually pulled from one is hard to verify, and Perplexity at least shows its citations openly, so you can sometimes see for yourself which sources fed a recommendation. Some research over the years has suggested Facebook ranks among the more prominent review sources that Bing surfaces for local businesses. Treat any such finding as a directional signal rather than a hard number, and do not assume it holds identically today.
What Each Assistant Appears to Read
| Assistant | Review data it appears to read | How confident that claim is |
|---|---|---|
| Google AI answers | Google reviews and Business Profile data | High. Google surfaces its own review data directly. |
| Microsoft Copilot | Bing-indexed listings, which can include Facebook Pages | Medium. Reported and plausible, based on how Bing indexes local listings, but not confirmed by Microsoft. |
| ChatGPT (browsing) | Whatever public pages it fetches for the query, potentially a Facebook Page | Low to medium. Possible when browsing, hard to verify per answer. |
| Perplexity-style assistants | Cited public sources, sometimes review platforms and Pages | Medium. Citations are visible, so you can check case by case. |
Read that table as a map of educated guesses, not a spec sheet. The confidence column is the point. The one row anyone can state plainly is the Google row.
Why Google Reviews Are Not the Whole Story
A business with a strong Google profile and nothing else is still betting everything on one source. If an assistant happens to lean on Bing for a query, and your Facebook Page is a ghost town, you have handed it less to work with. Spreading a consistent review presence across Google and Facebook, with matching name, address, and phone details, gives any assistant a cleaner and more agreeing picture to draw from. Consistency across sources appears to matter to these systems, which weigh evidence from several angles rather than trusting one listing alone.
Do Facebook Recommendations Count the Same as Reviews?
They function as Facebook's version of reviews, so for practical purposes yes. A recommendation is a customer publicly vouching for you with a written comment attached. To a human reading the Page, and to any system reading that public Page, a stack of recent recommendations reads as social proof in the same way a run of positive Google reviews does. The format differs, the signal is similar.
How to Check What the Assistants Say About You
You do not have to guess at any of this. Open ChatGPT, Copilot, and Perplexity, and type in a question shaped the way a real customer would ask it, for instance a plumber recommendation near a specific town or a family dentist in a specific area. Note which businesses get named, and on Perplexity, note which sources it cites underneath the answer. Ask the same question a handful of times in one sitting, since these tools can give different answers from one run to the next, so a single reply tells you very little by itself. Where a competitor keeps coming up, the citation list underneath the answer usually points at the exact directories and roundups worth keeping active. If your own Facebook Page ever shows up as one of those sources, that is direct evidence that keeping it healthy paid off.
The Practical Move for a Local Business
Keep your Facebook Page claimed, keep the recommendations tab switched on, and keep a steady trickle of honest recommendations coming in, especially if any of your customers find you through Meta ads. Do the same on Google, which remains the source you can state with confidence that AI answers read. You do not need to chase every assistant individually. A consistent, active review presence on the platforms your customers already use is the version of this you can actually control, and it happens to be the version that gives the AI systems the most to go on.