In future public prediction markets, large-scale participants may no longer be "crowds," but AI agents.

But first, let's understand why today they look like sports betting pools.
Why are public prediction markets increasingly resembling sports pools?
It's not that platforms or CEOs aren't working hard; it's that the structure is inherently like this.
(A recent long English article explains this chain thoroughly, with a link at the end.)
Over the past year, about 65% of trading volume on Kalshi and Polymarket has been sports-related.
Technology markets account for only about 1%.
Why can't many "interesting" markets emerge?
Among four types of participants:
"Savings-oriented funds" can't enter—prediction markets are zero-sum for traders and can't generate value.
"Hedgers" are also rare—people want to hedge event outcomes, not the events themselves.
In the end, only gamblers keep the market alive, with smart money following gamblers into short-term sports pools.
People wanting to open new markets also have to do some math:
Launching a new pool requires subsidizing liquidity, which is very expensive.
Even more complicated: the odds are usually only given as a probability,
for example, "There is a 67% chance this event will happen."
Companies want a full analysis: why, based on what, and what are the risks.
So many institutions prefer to hire consultants or research teams directly,
rather than spend money on a dedicated public market.
The article suggests that AI agents might be a patch:
Low cost, capable of scanning niche small pools, and able to explain their reasoning process.
The key is not "AI helping you place bets."
It's about filling the missing participant type in public prediction markets.
The diagram illustrates the entire logical chain.
Original text:

The entry point for prediction market research: