On-chain, increasingly bots. A crypto VC report in April 2026 estimated that automation and agents drive more than 19% of on-chain activity, and that bots generate over 76% of stablecoin transfer volume. But most of those are simple, narrow bots, not thinking AI agents, and that changes what an edge looks like for a human.
Who is actually on the other side of my trade?
Often a program, and usually a simple one. Most automated trading is rule-based software doing one job very fast, not an AI agent making judgement calls.
The main types you are likely to meet:
| Type | What it does | When it touches you |
|---|---|---|
| Market-making bot | Posts buy and sell prices all day and earns the gap | Often the counterparty when you buy or sell instantly |
| Arbitrage bot | Closes price gaps between venues within seconds | Keeps prices aligned, and takes the easy profit before you see it |
| Front-running bot | Sees a pending trade and trades just before and after it | When your order is large or your price limit is loose |
| Liquidation bot | Closes borrowers’ positions when collateral falls too low | In sharp drops, adding forced selling |
| AI agent | Pursues a goal and decides its own steps | Still a small share, mostly testing |
The April 2026 report backs that last row. An investment associate at the firm behind it said most of the 19% is narrow bots doing work such as capturing trading profits from transaction ordering and routing stablecoins. Truly agentic activity, the associate said, is a minority share.
Read the headline numbers carefully. The 76% figure covers stablecoin transfers, not trades: money moving between wallets and venues, much of it routine. The 19% figure covers on-chain activity broadly. Neither tells you the bot share of trades on any one exchange, and we found no reliable public figure for that.
So the honest answer has two parts. You are very likely trading against machines. You are much less likely to be trading against an AI that is outthinking you.
Is this new?
No. Machines have dominated the fastest parts of markets for years, and crypto inherited the pattern. The clearest public record comes from traditional markets.
In the 2010 Flash Crash, high-frequency traders accounted for about 49% of futures volume in one critical 14-second window, according to the joint SEC and CFTC staff report. They traded over 27,000 contracts while buying only about 200 net. Much of that was machines passing the same contracts between themselves.
What is new is the kind of machine. The Bank of England’s Financial Policy Committee said in April 2025 that autonomous AI trading strategies are a future possibility, with significant hurdles still to clear. Rule-based bots are the present. AI agents are the arriving wave.
Does it matter if a bot is on the other side?
It matters in three specific places: races, predictability and volume. Outside those, a bot is just a counterparty.
Races. You cannot win a speed contest against software. If a trade depends on being first, after news, a listing or a new token launch, assume a bot got there before you.
Predictability. Bots profit from orders they can see coming. A January 2026 academic study measured this on a large decentralised exchange. Sandwich attacks, where a bot buys just ahead of a visible order and sells just after it, cut a transparent trading strategy’s profits by 27.6% in one experiment.
Volume. When bots generate most of the activity, raw volume tells you less about how many people actually care about an asset. That is our inference from the figures above, not a measured result. Treat high volume as a question, not proof.
Where it matters less is time. A bot capturing a fraction of a cent on your trade does not decide whether your view over months is right.
Where does a human still have an edge?
Humans still win on judgement in messy, changing situations and on the freedom not to trade. The current evidence supports both.
In an open-ended trading contest reported in April 2026, with each account starting at $10,000, the top human traders beat the top AI agents by more than five times. An investment associate at the firm behind the report said agents cannot yet react when markets change and conditions are unclear.
That points to where a person’s edge now sits:
- Time horizon. Bots fight over seconds. You can hold a view over weeks.
- Ambiguity. Reading a messy situation is still hard for software.
- Patience. No rule forces you to trade. You can wait for a price you like.
- Context. You can weigh things no data feed carries, such as who is behind a project.
Speed has moved permanently to machines. Judgement has not, at least for now.
What could go wrong?
The real risk is not that a bot outsmarts you. It is that your own order settings hand bots an easy profit, or that machine-made activity misleads you. Here is how that happens.
Loose slippage settings. Slippage tolerance is how far from the quoted price you will accept a fill. Set it wide and a front-running bot can take the whole margin from you, automatically.
Chasing machine momentum. A sudden burst of buying may be bots trading with bots. Jumping in late means buying from the machines that started it.
Racing at launch. New tokens attract automated buyers in the first seconds. A human clicking “buy” is almost always last in that queue.
Everyone’s bot does the same thing. As more traders use similar agents, their behaviour converges. Our guide on what happens when thousands of agents run the same strategy explains why that can turn a dip into a cascade.
Copying a visible strategy. If you follow a public strategy, bots can follow it too, and trade ahead of you. Our guide on copying someone else’s AI trading strategy covers that trade-off.
The trust takeaway
Assume a machine is on the other side of fast trades, and set your orders so it cannot profit from your carelessness. Your edge is not speed. It is patience, judgement and refusing to trade on someone else’s clock.
Before your next trade, check:
- Is your slippage tolerance tight? Use the smallest setting that still lets the trade go through.
- Can you use a limit order? Name your price instead of accepting the market’s.
- Is this a race? If being first is the whole idea, a bot will be first.
- Is the volume real interest? Look at how many separate wallets are trading, not just the total.
- What does your “AI” tool actually do? Ask whether it is a rule-based bot, and whether it can withdraw funds.
One way to keep decisions on your own clock is to follow the market and your holdings in one place. TrustSwap’s The Crypto App is a crypto news and portfolio tracking app: it helps you see what is moving and what you hold before you act, rather than reacting in the seconds where bots have the advantage.