With the automated bidder maximising ad prices for shareholder returns it can become zero sum equivalent. Every gain is matched by a loss elsewhere.
This changes if ads are creating new demand, or if consumers are getting value they otherwise wouldn't have found. But in many markets that won't be the case.
This means is often
It's not a simple "both sides" issue - the scenario has multiple sides. You find a good keyword, someone finds you, you both benefit. Great!
Then a third party finds and uses your strategy, and bids higher in the ad auction. This reduces profits. As more parties join in, the profits decrease.
The thesis is that those profits don't simply go to zero, but can actually result in a net loss. In the meanwhile, those higher bids end up as profits for the auction house.
I'm picking on the gambling analogy. It is wrong.
I agree with the general thesis of the article.