Solana volume bots: by our tests, 58% of DEX trading in the pools we index was not real.

01 · A Monday on SolanaTwo wallets, one transaction

At 08:47 UTC on Monday, September 14, a Solana wallet paid about 9.8 SOL for 61 million units of a token named Claude, in a PumpSwap pool, the pot of SOL and tokens that trades are made against. The token is not issued by or linked to Anthropic. In the same transaction, a second wallet sold almost as many units back to the same pool for about 9.7 SOL.

The second wallet also paid a little SOL from its own balance, which Solana only allows if that wallet signs. So two wallets signed one transaction, one buying and one selling. Our data records it as two trades worth about $2,000.

Both wallets belong to a group of 20 that traded the same way. In that one token, the group made 8,600 of these round trips in about an hour, one every half second or so. That was almost all the money traded in the token. Eleven seconds after our transaction, the group made its last round trip in our data.

PumpSwap, the exchange pump.fun runs, is the busiest venue in our Solana data. We checked every DEX trade on Solana from August 24 to September 22 in the pools we index, $201.4 billion in all. Our tests flag $117.7 billion of it, 58%, as not real trading.

Almost all of that was the move in this one transaction, repeated over and over by bots. Two groups of look-alike wallets made more than a fifth of it, and nearly half ran through tokens named after stocks, AI firms and their products.

02 · Why pad the volumeVolume is the number that gets a token noticed

Volume is the dollar value of every trade. Trackers and screeners rank tokens by it, so a token with big volume looks busy and easy to trade. Wash trading means trading with yourself to make a market look busier than it is. Volume bots are programs that do it for you, trading a token back and forth all day. We explained why people bother in our investigation of Base.

On Solana, one transaction can carry several signers. So two wallets can take the two sides of a trade in a single step, and the trade still shows up as volume.

03 · The pattern95% of it was one move

About 95% of the not-real volume, $111.6 billion, was a buy and a sell of the same token, in the same pool, inside one transaction. In the two such transactions we opened, a second wallet took one side and signed too, and in one full hour we checked, almost all of these round trips worked the same way.

Most round trips lose a little to the pool. Where we can measure that cost, it came to about 0.3% of the volume. Where those fees ended up, we did not trace.

The other 5% is bots that buy and sell the same token over and over in separate transactions. The biggest single wallet has $2.0 billion of it, spread over hundreds of tokens. One of them is a token named BNB that is not issued by or linked to BNB Chain. In the first six round trips we pulled for September 1, the wallet bought about $17,600 to $19,700 of it and sold the same amount back within half a minute.

04 · By dayFour to five billion dollars per day, then less

Until September 8, the one-move trading ran at $4 billion to $5 billion per day. After that it fell to $2 billion to $4 billion per day.

Part of the drop is one group. The group of 50 wallets stopped for five days in mid-September, then came back at less than half its old pace.

05 · WhereMostly PumpSwap, and a tracker that counts less

Tokens that trade mostly on PumpSwap make up nearly three quarters of the volume we priced, and almost all of the part that was not real.

DefiLlama counts $18.4 billion for PumpSwap over the same 30 days. It only counts pools with at least $5,000 of liquidity and 50 traders per day, a filter it says removes wash trading. Most of the tokens here had 50 or more traders per day, so the trader rule alone would not catch them.

What our tests did not flag comes to $83.7 billion. That is close to DefiLlama's $78.8 billion for the whole chain, but partly by chance. DefiLlama counts venues we do not index and leaves out PumpSwap pools we keep.

06 · The namesAnthropic, AAPL, HOOD

Nearly half of the not-real volume, $55.8 billion, ran through 1,562 tokens named after stocks, funds, AI companies and their products, in which most of the trading was not real by our tests. There were 254 tokens called Anthropic alone.

Anyone can launch a token under any name, and these names say nothing about who made the tokens. Only a handful carry the address ending pump.fun gives its launches, yet almost all of them trade mostly on PumpSwap.

07 · Two groupsSeventy wallets, $26.3 billion

Both wallets in our transaction belong to a group of 20 that traded from August 30 to September 14. A second group of 50 traded a different set of tokens, with none in common. Inside each group, every wallet recorded almost the same total, spread over almost the same number of tokens.

Together the two groups made $26.3 billion, more than a fifth of the not-real volume. Wallets with records this close to each other look like copies of one bot. Who funds the two groups we have not traced. The trail can be followed with the Bitquery MCP.

08 · Trade-count bots108 million padding trades

A different kind of bot pads the number of trades instead of the dollars. We found 108 million trades like this, about one in eleven of all the trades we priced, and nine in ten were under $5. We count these trades apart from the dollar figures above. They make a new token's trade count look busy, like the crowd of tiny trades in our LAPTOP story.

09 · Our earlier storiesHow this fits what we found before

Our story on counterfeit SOL tokens spelled with Cyrillic letters counted $330 billion of volume priced in counterfeits. Almost none of it is in these figures. Nearly all of those trades price one counterfeit in another, and we price only trades against SOL, USDC or USDT.

Our pump.fun stock-token story found almost no round trips inside one transaction. It looked at coins priced in tokenized stocks, which this count leaves out, and found this kind of round trip in about 2% of their volume. In SOL, USDC and USDT pools, the same move makes up more than half.

Our Robinhood Chain and Solana memecoin comparison used all recorded volume, with none of these tests. Solana's 58% is the highest share of the six chains we measured. That measure began with the $33.8 billion BNB Chain wallet and the Robinhood Chain hand-off line.

10 · How to spot itThree signs of a volume bot

Each of these can be checked from public data.

• Both sides in one transaction. A buy and a sell of the same token, in the same pool, inside one transaction, time after time.
• Twin wallets. Wallets with near-identical totals and token counts, trading the same set of tokens.
• A borrowed name. A new token named after a big company, where almost all the trading comes from the same few dozen wallets. A token's trade history shows all three.

11 · How we measuredWhere the numbers come from

We looked at every DEX trade on Solana from August 24 to September 22 that was priced in SOL, USDC or USDT, in the pools we index. Trades priced in other tokens are left out. So are swaps that aggregators route to venues we do not index, about 3% of volume. Counted as real, they would bring the share to about 57%.

A trade counts as not real when a wallet buys and sells the same token in one transaction and gets back about what it paid, less fees. It also counts when a wallet trades one token back and forth at bot speed and loses on it or breaks even, or when a few such wallets make up most of a token's trading. The 58% is a floor. At least $6.3 billion more of these round trips fits none of our rules and is left out. In all, 16,530 signing wallets and 29,073 tokens carry volume that was not real.

The names we matched, exactly and ignoring case: NVDA, HOOD, TSLA, AAPL, GOOGL, GOOG, AMZN, META, MSFT, SPY, QQQ, COIN, MSTR, PLTR, AMD, NFLX, CRCL, OpenAI, Anthropic, Claude, ChatGPT, Grok, Gemini, SpaceX, SPCX, Tesla, Apple, Google, Nvidia, Robinhood, Walmart, XAI, Microsoft, Amazon. The second-wallet finding rests on the transactions we opened and one full hour. We did not check every trade. Whether buyers lost money we could not measure fairly, because our data missed many of the days people sold. Bitquery's own DEX data records these swaps like any other trade; the figures here come from pulling them apart.

12 · The recordThe addresses behind the story