How I Use ChatGPT to Analyze On-Chain Data — Whale Wallets, Exchange Flows, and Trading Signals

Why On-Chain Data Changes the Game

Most crypto traders stare at the same price charts and the same RSI readings as everyone else. On-chain data gives you something different: a view of what participants are actually doing with their coins, not just what the price is doing on an exchange orderbook.

When a whale moves 5,000 BTC from a cold wallet to Coinbase, that is not a chart pattern — it is an observable action with implications. When stablecoin supply on exchanges increases by $2 billion in a week, that is potential buying power sitting on the sidelines. When funding rates on perpetual futures go deeply negative while price is holding steady, that tension between derivatives positioning and spot demand tells a story.

The problem is that on-chain data is noisy, fragmented across multiple platforms, and presented in formats that require interpretation. This is where ChatGPT becomes genuinely useful. I do not use it to predict prices — I use it to structure my interpretation of data that I would otherwise spend hours piecing together manually.

In this tutorial, I will walk through my exact workflow: where I get on-chain data for free, the specific prompts I use with ChatGPT to analyze it, and a real example where an on-chain signal preceded a meaningful price move.

Free On-Chain Data Sources

Before you can analyze anything, you need data. Here are the tools I use regularly, all with free tiers:

A few notes on these tools. Glassnode’s free tier is limited but still covers the most important metrics — exchange netflow and basic supply data. Arkham is the best free tool for figuring out which wallet belongs to which entity. Dune requires some SQL knowledge, but you can use community dashboards without writing a single query. DefiLlama is completely free and has the best stablecoin tracking I have found. Coinglass is the go-to for derivatives data.

I do not pay for any of these. The free tiers give me enough data for a weekly on-chain analysis workflow. If you trade full-time and need real-time alerts, paid tiers start making sense, but I have not needed them yet.

The Core Workflow: Pull, Paste, Analyze

My on-chain analysis workflow follows a simple loop:

1. Pull data from one of the sources above (screenshot or copy numbers)
2. Paste into ChatGPT with a structured prompt
3. Get analysis that I then compare against my own read
4. Log the signal and check it against price action over the following days

The key insight I learned early: ChatGPT is much better at interpreting on-chain data when you give it context. Do not just paste numbers — tell it what the numbers are, what the baseline looks like, and what time period you are looking at.

Prompt 1: Exchange Inflow/Outflow Analysis

Exchange inflows (coins moving onto exchanges) historically correlate with selling pressure. Exchange outflows (coins moving off exchanges) suggest accumulation. This is one of the most reliable on-chain signals.

I pull this data from Glassnode’s free Bitcoin Exchange Netflow chart, then paste the numbers into ChatGPT:

You are an on-chain data analyst. I am going to give you Bitcoin exchange
netflow data. Positive values mean net inflows to exchanges (potential
selling pressure). Negative values mean net outflows (potential accumulation).

DATA (BTC exchange netflow, daily, last 14 days):
[Paste 14 data points here, e.g.:]
- Feb 1: -1,200 BTC
- Feb 2: -800 BTC
- Feb 3: +3,500 BTC
- Feb 4: +2,100 BTC
- Feb 5: -400 BTC
- Feb 6: -1,800 BTC
- Feb 7: -2,200 BTC
[... etc]

CONTEXT:
- Current BTC price: $[PRICE]
- 14-day price change: [X]%
- This month's average daily netflow: [AVG] BTC

ANALYZE:
1. What is the overall netflow trend over these 14 days? Net accumulation
   or net distribution?
2. Are there any single-day spikes that stand out? What might they indicate?
3. How does the recent netflow compare to the monthly average I provided?
4. If this netflow pattern continued, what would it suggest about near-term
   supply/demand dynamics?
5. What additional data would you want to see to increase confidence
   in any conclusion?

Be specific with numbers. Acknowledge uncertainty. Do not predict price.

What I look for in the output

The most useful part of ChatGPT’s response is usually section 2 (single-day spikes) and section 5 (what additional data it would want). Spike identification saves me from visually scanning charts, and the “what else would help” section often points me to a second data source I had not considered checking.

One pattern I have noticed: when there are 3 or more consecutive days of strong outflows (more than 2x the average daily netflow) and the price is flat or declining, that divergence has historically been interesting. It suggests someone is accumulating despite weak price action. ChatGPT does a decent job of flagging these divergences when you provide both netflow and price context.

Prompt 2: Whale Wallet Movement Tracking

Arkham Intelligence lets you track labeled wallets — exchanges, funds, known whales. When a wallet labeled as a major holder moves coins, I want to understand the context.

You are analyzing whale wallet movements for Bitcoin. Here is what I
observed on Arkham Intelligence today:

MOVEMENTS:
- Wallet A (labeled "[ENTITY NAME]"): Moved [AMOUNT] BTC from
  [cold wallet / exchange / unknown] to [destination]. Transaction hash:
  [optional]
- Wallet B (labeled "[ENTITY NAME]"): Moved [AMOUNT] BTC from
  [source] to [destination].
[Add as many movements as relevant]

CONTEXT:
- Current BTC price: $[PRICE]
- These wallets last moved funds on: [DATES if known]
- Market conditions: [brief description — trending up, range-bound, etc.]

INTERPRET:
1. For each movement, what is the most likely intent? (Selling, buying,
   rebalancing, moving to custody, moving to DeFi)
2. Is the destination an exchange (bearish signal), a cold wallet
   (accumulation signal), or something else?
3. Are any of these movements unusually large relative to the entity's
   known holdings?
4. Taken together, do these movements paint a coherent picture or are
   they mixed signals?
5. What is your confidence level in each interpretation? (High / Medium / Low)

Important: Wallet movements have many possible explanations. Do not
jump to conclusions. Present multiple hypotheses where appropriate.

A note on false signals

I added the “multiple hypotheses” instruction after getting burned. Early on, I saw a large transfer to Coinbase and ChatGPT confidently said “this is likely preparation to sell.” It turned out the entity was moving funds to Coinbase Prime for institutional custody, not to the exchange’s spot market. The transfer was actually neutral, not bearish.

Since then, I always ask for multiple hypotheses. A transfer to an exchange wallet could mean: selling, lending, providing liquidity, moving to custody, or even an internal rebalance. ChatGPT does better when you explicitly remind it that the obvious interpretation is not always correct.

Prompt 3: Funding Rates and Open Interest

Perpetual futures funding rates tell you whether the market is overleveraged long or short. I pull this data from Coinglass:

Analyze the following Bitcoin perpetual futures data:

FUNDING RATES (8-hour, last 7 days):
[Paste funding rate data from Coinglass, e.g.:]
- Binance: 0.01%, 0.01%, 0.015%, 0.02%, 0.035%, 0.04%, 0.05%, ...
- Bybit: 0.01%, 0.012%, 0.018%, 0.025%, 0.038%, 0.042%, 0.048%, ...
- OKX: 0.01%, 0.01%, 0.014%, 0.019%, 0.032%, 0.039%, 0.047%, ...

OPEN INTEREST:
- Total BTC futures OI: $[X] billion
- 7-day OI change: [+/- X]%
- 30-day OI change: [+/- X]%

CONTEXT:
- BTC spot price 7-day change: [X]%
- Recent liquidations (24h): $[X]M longs, $[X]M shorts

ANALYZE:
1. Are funding rates trending higher, lower, or stable? What does
   this tell you about market positioning?
2. Is there a divergence between funding rate direction and price direction?
   If so, what does this historically suggest?
3. Is open interest elevated relative to recent history? What does
   that imply about leverage in the system?
4. Based on the liquidation data, which side is getting squeezed?
5. If you had to describe the derivatives market positioning in one
   sentence, what would it be?
6. What funding rate level would you consider a warning sign for a
   potential squeeze? (either direction)

Stick to what the data shows. No price predictions.

Why I track funding rates

Funding rates above 0.05% per 8 hours (roughly 55% annualized) have historically preceded corrections. Not always, but often enough that I pay attention. Deeply negative funding rates during an uptrend suggest shorts are paying longs to stay open, which means there is a large short position that could get squeezed.

ChatGPT is particularly good at explaining these dynamics in plain English when you feed it the raw numbers. I have found that its explanation of funding rate mechanics is consistently accurate, even if its application to specific market conditions requires my judgment.

Prompt 4: Stablecoin Supply as a Leading Indicator

This is one of my favorite on-chain signals. Stablecoin supply on exchanges represents dry powder — money that is ready to buy but has not yet bought. I track this on DefiLlama’s stablecoin page:

Here is stablecoin supply data from DefiLlama:

TOTAL STABLECOIN MARKET CAP:
- Current: $[X] billion
- 30 days ago: $[X] billion
- 90 days ago: $[X] billion

STABLECOIN SUPPLY ON EXCHANGES:
- Current: $[X] billion
- 7 days ago: $[X] billion
- 30 days ago: $[X] billion

BREAKDOWN BY STABLECOIN:
- USDT: $[X]B total ($[X]B on exchanges)
- USDC: $[X]B total ($[X]B on exchanges)
- DAI: $[X]B total ($[X]B on exchanges)

CONTEXT:
- BTC price: $[PRICE]
- BTC 30-day change: [X]%
- Total crypto market cap: $[X]T

ANALYZE:
1. Is total stablecoin supply growing or shrinking? At what rate
   compared to the prior period?
2. Is the stablecoin supply on exchanges increasing or decreasing?
   What does this suggest about potential buying power?
3. Are there differences between USDT and USDC trends? (USDT growth
   often correlates with non-US demand, USDC with US institutional demand)
4. How does stablecoin supply growth compare to overall crypto market
   cap growth? Is there a divergence?
5. Based on this data, would you characterize the current environment
   as "dry powder accumulating" or "stablecoins being deployed"?

Provide specific numbers in your analysis. Qualify all conclusions
with confidence levels.

Why stablecoins matter

The logic is simple: before people buy crypto, they first convert fiat to stablecoins. A rising stablecoin supply on exchanges with flat or declining BTC prices can indicate that capital is positioning itself to enter. Conversely, declining stablecoin supply during a rally might mean the buying power is getting exhausted.

I have found this to be one of the more reliable leading indicators, with the caveat that the lead time is unpredictable — sometimes days, sometimes weeks, sometimes the signal fizzles entirely.

Complete Workflow Example: January 2026 BTC Accumulation Signal

Let me walk through a real analysis I did in late January 2026 that identified an accumulation signal before a price move.

The data I collected (January 22, 2026)

BTC was trading at approximately $42,800, down about 8% from its January high. Sentiment was cautious — Fear & Greed Index was at 38 (Fear).

Here is what I pulled from free sources:

Exchange netflows (Glassnode free tier): Five consecutive days of net outflows, averaging roughly -2,400 BTC per day. This was 3x the 30-day average outflow rate.

Whale wallets (Arkham): Two wallets labeled as known accumulation addresses received a combined 4,200 BTC over the previous week. Transfers came from OTC desks, not from exchange wallets.

Funding rates (Coinglass): Funding had gone negative on Binance and Bybit for the first time in six weeks. The market was paying shorts to stay open, meaning short interest was elevated.

Stablecoin supply (DefiLlama): USDT on exchanges had increased by $1.8 billion over the previous two weeks, while BTC price had declined.

The ChatGPT synthesis prompt

After gathering all four data points, I pasted them together into one synthesis prompt:

I have collected four on-chain data points for Bitcoin as of January 22, 2026.
BTC is at $42,800, down 8% from the January high. Fear & Greed: 38 (Fear).

DATA POINT 1 — EXCHANGE NETFLOWS:
5 consecutive days of net outflows, averaging -2,400 BTC/day.
30-day average is -800 BTC/day, so current outflows are 3x the norm.

DATA POINT 2 — WHALE WALLETS:
Two known accumulation wallets received 4,200 BTC total over the past week.
Transfers originated from OTC desks, not exchange wallets.

DATA POINT 3 — FUNDING RATES:
Funding turned negative on Binance (-0.01%) and Bybit (-0.008%)
for the first time in 6 weeks. Shorts are paying to stay open.

DATA POINT 4 — STABLECOIN SUPPLY ON EXCHANGES:
USDT on exchanges up $1.8B in the past 2 weeks while BTC price declined.

SYNTHESIZE:
1. Look at all four data points together. What story are they telling
   when combined?
2. Are there any contradictions between the signals?
3. Historically, when exchange outflows accelerate, whales accumulate,
   funding goes negative, and stablecoin supply rises — what has
   typically happened?
4. What is the strongest signal here and what is the weakest?
5. What could invalidate this reading? What would make all four
   signals misleading?
6. On a scale of 1-10, how confident should I be that this data
   points toward accumulation? Explain your rating.

Be honest about the limits of this analysis. I want to hear the bear
case as much as the bull case.

What ChatGPT returned (summarized)

ChatGPT identified the convergence of all four signals as an accumulation pattern. It rated confidence at 6 out of 10, noting that the whale wallet data was the strongest signal (specific, identifiable entities making large moves through OTC rather than exchanges) and the stablecoin data was the weakest (stablecoin supply can grow for reasons unrelated to crypto buying intent, such as remittance flows or DeFi yield farming).

It flagged two potential invalidation scenarios: (1) the whale wallets could be moving to OTC desks for selling, not buying — though Arkham’s labels suggested these were known accumulator addresses, not sellers; (2) the negative funding could deepen into a capitulation event rather than a short squeeze setup.

What actually happened

Over the following 12 days, BTC moved from $42,800 to approximately $47,500, a gain of about 11%. The negative funding rates did lead to a short squeeze on January 26, with roughly $340 million in short liquidations across major exchanges.

I want to be careful here: this is one example, and I am showing it because it worked. I have done the same type of analysis on other occasions where the signal did not lead to a significant move, or where the move took much longer than expected. Cherry-picking successes is the easiest way to mislead yourself. The point of this example is to show the workflow and the type of output you get, not to suggest that this works every time.

Limitations I Have Learned the Hard Way

Data lag. Free-tier data is often delayed by hours or even a full day on some Glassnode metrics. By the time you see a large exchange inflow on the free tier, the move may have already started. This is less of an issue for weekly analysis but matters a lot if you are trying to trade short-term signals.

Correlation is not causation. Exchange outflows often precede price increases, but not always. Sometimes coins move off exchanges for staking, for bridging to other chains, or for reasons that have nothing to do with trading intent. ChatGPT cannot tell the difference between these scenarios and neither can you with certainty.

Overfitting narratives. This is the biggest risk. When you paste on-chain data into ChatGPT and ask it to synthesize, it will almost always construct a coherent story. That is what language models do — they find patterns and create narratives. But the narrative might be wrong. I have caught ChatGPT building a convincing bearish case on one day and an equally convincing bullish case the next, using similar data. The model is agreeable by nature. Your job is to challenge its conclusions, not just accept them.

Single data points are meaningless. One day of large exchange inflows does not mean a crash is coming. One whale wallet transfer does not confirm a trend. I have learned to only pay attention to on-chain signals when multiple independent data points converge, and even then I maintain healthy skepticism.

ChatGPT does not have real-time data access. Every number I paste in is manually gathered from the sources listed above. If I make a mistake copying a number, ChatGPT will analyze the wrong data and give me confidently incorrect analysis. I double-check my inputs before pasting.

Free vs. Paid: Is It Worth Upgrading?

The honest answer: if you are doing this once or twice per week as part of a broader research routine, free tiers are enough. If you want to build a systematic, alert-driven workflow that notifies you of on-chain signals in real time, paid tools start making sense. I have not made the jump to paid tools yet because I do not trade frequently enough to justify the cost.

Building a Weekly On-Chain Analysis Habit

Here is the schedule I follow:

Sunday evening (20 minutes): Pull data from all four sources. Run the synthesis prompt. Log the overall on-chain read (accumulation / distribution / neutral) in my research spreadsheet.

Wednesday midweek check (10 minutes): Quick check on exchange netflows and funding rates only. If something changed dramatically since Sunday, update my thesis.

As needed: If Arkham shows unusual whale activity or Coinglass shows extreme funding rates, I run the relevant individual prompt regardless of the schedule.

This takes about 30 minutes per week total. I log every analysis and compare my on-chain read to what price actually did over the following 7 days. After two months of doing this, my hit rate on directional reads is roughly 60% — better than a coin flip but far from reliable enough to trade on alone.

That is why I treat on-chain analysis as one input among several. It informs my thesis alongside technical chart analysis and macro conditions — it does not replace them.

A Warning About Narrative Bias

The biggest danger of using AI for on-chain analysis is that it accelerates narrative construction. Without AI, you look at a Glassnode chart, form a rough impression, and maybe move on. With AI, you paste data, get a polished narrative, and feel more confident than the data warrants.

I counteract this by always asking ChatGPT to argue against its own conclusion. After any synthesis prompt, I follow up with:

Now argue the opposite case. Using the same data, construct the most
compelling counter-narrative. What would someone who disagrees with
your analysis say?

This forces a reality check. If ChatGPT can construct an equally convincing opposite case from the same data, the signal is probably weak and I should not act on it. If the counter-narrative feels forced and requires ignoring multiple data points, the original analysis might have some merit.

Next Steps

• Build a full daily routine — My daily AI trading research routine incorporates on-chain data alongside chart analysis and market briefings
• Complement with macro analysis — On-chain data tells you what participants are doing, but macro tells you why — see my AI-powered macro analysis workflow
• Get prompt templates for market analysis — Start with the ChatGPT trading analysis playbook for foundational prompts
• Understand AI trading tools — If you are new to AI-assisted trading, the AI Trading 101 guide covers the full landscape