It turns out that in other categories, a LOT of people search for things but very few buy them. About 1.5 million luxury cars are sold in the US in any given year but you could bet the number of searchers would be in tens to hundreds of millions.
So the Google data works for some businesses which are in the Goldilocks zone when it comes to search behaviour (ie. It's a product that just the people who wamt to buy will actively research before buying) but for many other businesses - especially the ones with giant marketing budgets, search is just hygiene cost, not a business driver - ie. spend enough so your competitors don't appear as the first results in search ranking.
At the scale of many businesses (especially the ones in the first paragraph), broad demographic segments (and yes TV!) turns out to be a very efficient way to build a brand and yes you can also selectively augment with other data based marketing.
For other businesses (think niche cosmetics brand) paid word of mouth (aka influencer marketing) and pushing the paid word of mouth content to more people who are likely to be have the same interests (eg. Follow fashion bloggers) turns out to be hugely efficient and relevant and that's something Facebook excels at.
Contrived example: I search for "buying new SUV" so clearly you should show me new car ads and car buying articles and reviews. But if I am a mid-30s married person, you might also try to sell me car seats for the babies I demographically am likely to have or want to have soon. OTOH, if am an older 60s retired widow, you may want to advertise warranty and car maintenance plans for my peace of mind about taking care of the car.
Actually it's a little more nuanced than that, because I could be mid-30s and married but choosing to be childfree, or I could be a 65 year old widow and former car mechanic who is more than capable of taking care of my own vehicle. So collecting the demographic data is about building a gradient of probabilities, things that are ranked in likelihood of value/relevance, and using that information to inform the demand side of advertisers deciding how much to bid for my eyeballs.
Facebook finds out what you think to share, Google finds out what you're doing and generally the path you took to get there.
Google News has always uncannily followed by current job's tertiary areas, for example.
The power is in combining that data, like one of the grandparent comments mentioned. They will know if you buy a SUV (since the dealer uploads the data as an offline conversion) and that you're browsing baby clothes on Walmart/Target. Now you're a "new or expecting parent".
Some of these interest categories are not that explicit but the lookalike audiences are. That's a whole other post :)
- Google are the experts/masters of knowing "what you want", expressed via your search intents
- Facebook are the experts/masters of knowing "who you are" (including demographics, as well as family/friends/connections, interests/hobbies/personality, etc.)
Google can guess at the same type of "who" information that Facebook owns (e.g. gender inference), and vice versa Facebook can guess at the type of info that Google owns (e.g. search/purchase intent), but both have distinct relative advantages over the other, which is why there is still competition in this space (the GP comment was questioning, why hasn't Google just "won", why is FB even in the picture, etc.).