I think it's useful at this point to ask why various directories had value during their heyday. The "Yellow Pages" had value, simply because there wasn't as comprehensive a localized listing otherwise. Yelp had value at first, because the reviews and ratings combined with the location search were super useful. Unfortunately, too many people realized how valuable this was and the gaming of the system on all sides started, greatly reducing the usefulness of the Yelp directory.

Google might be able to create a such directory based solely on traffic and AI identification of the type of business? Could AI identify long lines outside of restaurants? Maybe Google should buy Yelp, which would solve the "pay Google up front" conundrum. Whatever IP Yelp has which killed the "near" searches on Google would also cease to be an impediment to Google.

I'm not sure if this is exactly it or if you're letting Yelp off the hook. Yelp was perfectly happy to try to profit from the gaming. That didn't just damage its usefulness. It damaged trust; and trust is a lot harder to get back.

Not letting them off the hook. Yelp contributed to this through their actions.

It damaged trust; and trust is a lot harder to get back.

That was a part of what I was thinking about. One advantage of the Yellow Pages: less opportunity to lose trust. I think displaying AI interpreted traffic has a key advantage over ratings: It would cost a whole lot more money to get crowds to show up day after day with smartphones, than it would cost to get people to astroturf ratings.

Google already gives you the hours that restaurants are busiest, so they basically have this info

I suspect that AI applied here would be highly useful.

Busyness wouldn't just be used as a measure of busyness. It could be used as a proxy for other qualities. Another example: There are restaurants that have ample waiting areas and others that have none. Those in the waiting area wouldn't necessarily mean the same thing to a customer's experience than the same people seated at a table. Likewise for people standing outside vs. waiting inside.

At a very basic level, throughput (Little's Law) should also be taken into account.

My point is there would be value in such proxy measures, if some AI were used for interpretation.

Google already does plenty of stuff like this, and has for years (including historical and live busyness). Hell, I haven't used Yelp in years, because their app is generally less usable, isn't integrated with Maps (for directions etc), and cripples mobile web by not letting you see most of the content. Though it's possible that the data for G Maps is only comparably rich because of my location, and that Yelp has a data advantage in many more markets.

I'm well aware. There's a cousin comment about using such data with contextual information and AI to use traffic data as a proxy for attributes like popularity and quality, instead of providing the raw data and letting users make their own inferences. This would enable a more Yelp-like UX, and might be easier for lots of users to consume.