How ChatGPT Decides Which Pool Service to Recommend in the Coachella Valley

We asked ChatGPT a simple question: “Can you recommend a pool service in Palm Desert?” Then we asked it about every other city in the Coachella Valley. Then we dug into why it picked who it picked.

Here’s what we found.

ChatGPT doesn’t search Google.It pulls local business data from Bing. Specifically, Bing Places — Microsoft’s version of Google Business Profile. If your business isn’t claimed on Bing Places, you don’t exist in ChatGPT’s database. Full stop. Most local businesses have never even heard of Bing Places, let alone claimed their listing.

Reviews matter, but not the way you think. ChatGPT doesn’t just look at your Google reviews. It cross-references across Bing, Yelp, Facebook, and anywhere else it can find consensus. A business with 30 reviews on Google and zero everywhere else looks thinner than a business with 20 across three platforms. Consistency and recency beat raw volume.

Your website is being read — literally. When ChatGPT recommends a business, it pulls details from your website to explain why. Hours, services, pricing philosophy, service area — it’s all fair game. If your website is a one-page template with “Lorem ipsum” energy, there’s nothing for the AI to quote. The businesses with structured, current, content-rich sites get cited with links. The ones without get mentioned by name only, or not at all.

Schema markup is the cheat code nobody’s using. Schema is code on your website that tells the AI exactly what you are — your business type, services, location, reviews, FAQs. It’s like putting a label on the outside of a box instead of making the AI open it and guess. Businesses with proper schema get cited at dramatically higher rates. In the Coachella Valley pool service market, almost nobody has it.

The ranking isn’t random. It’s predictable. Bing Places presence, review consensus, website structure, schema markup, directory consistency — the businesses that check all five boxes are the ones ChatGPT recommends. The ones that check one or two get listed. The ones that check none are invisible.

The good news: this isn’t a mystery. The inputs are knowable, the gaps are measurable, and the fixes are straightforward. The hard part is doing all of it, keeping it current, and knowing when the rules change.

That’s what we do.