People assume the hard part of Spotguide is the map. It isn't. The hard part is the few seconds between "someone sent us a video" and "we know exactly which restaurant this is." Here's what happens in that gap.
Step one: read the post, then watch it
The caption is the easy part. Spotguide reads it, along with who posted it, both the handle and the display name. A reel from "Eric Taliercio" with #njdeli in the caption is a strong clue the place is Taliercio's, in New Jersey. If the post has a location tag, that's read too.
But the name of a restaurant is often never written down. It's shown. So for Reels, Spotguide watches the video. It listens to what's said and reads the text on screen: the menu board, the price on the receipt, the sign over the door. A long or slow video gets the time it needs rather than being given up on.
It also understands that an account like @joes_pizza.nyc probably means "Joe's Pizza in New York," and that the account posting might be the restaurant.
Step two: find the real place
The name gets matched against Google's map of the world, biased toward wherever the post says the place is. Hashtags and handles that name a city win over your home city, so a deli in New Jersey doesn't get matched to a lookalike down the street from you.
Step three: check its own work
Every match gets a confidence score, and Spotguide is honest about it.
A clear winner saves automatically. A shaky match, say three same-named locations or a story that names the place but not the city, goes to your Review queue, where you pick from the candidates with one tap.
When nothing on the map even resembles the name in the video, which happens with pop-ups and brand-new spots, Spotguide says "we couldn't find it" rather than guessing the closest-sounding business. Before giving up, it tries the same trick our team uses by hand: search the web for the post's caption, find the original post, and read the restaurant's real name off it. Only a confident, verified answer counts.
Step four: remember why you saved it
This is the part that makes the save useful later. The dishes mentioned, the vibe words, and a one-line summary of the post are kept with the place. If the same restaurant gets sent twice, Spotguide says "already on your map" and folds the new video into what it knows. Three friends' reels about the same taco spot make it three times more understood, not three duplicates.
Roundups and galleries
A "5 best taco spots" video saves all five. A swipe-through gallery with a different restaurant in every picture is read photo by photo. A real 19-photo "best bites in San Diego" guide produced six saved spots from one link.
Why we're strict about it
A wrong pin is worse than no pin. You'd show up at the wrong place, or stop trusting the map. So Spotguide would rather ask you a question than guess. Every correction someone makes teaches it, and the next person who sends that same video gets the right answer first time.
If you want to see it work, send a Reel to @spotguideapp. The reply usually arrives before you've finished reading this sentence.