Direct answer

Prove what my AI did

To prove what my AI did, you do not argue from the chat. You point to the receipt that was fixed when each call ran, and you let the checker recompute it: a receipt exists for the step, it names the tool and the run, its fingerprint still matches the saved result, its fields agree with the claim, and its place in the chain is intact. Run the work through Zambo, the verifiable work layer for AI, and every call leaves that receipt behind. Anyone can then check it free, with no account and no screenshots.

Check a real receipt first: open zambo.dev/verify, run a receipt through the five checks on this page, then run one call of your own at the live demo and check the receipt it hands you. The full method lives on verify AI work.

Updated October 20, 2026.

The situation

It usually starts with a simple question you cannot answer. A manager asks what the AI actually did on the project yesterday. A client asks which parts of the deliverable the AI produced and which parts you did. Or you ask it yourself, a week later, because the summary in the chat says the research was done, the spreadsheet was updated, and the draft was sent, and you would like to know whether any of that is true before you repeat it to someone else.

The AI claims done. That claim is fluent, confident, and produced by the same system that did the work, with the same blind spots. When a step silently failed, the summary often reports it as finished anyway. When a result was never observed, it can still be described as a fact. So the searcher in this situation, whether that is you, a client, or a reviewer, needs something that is not the AI describing itself: a record made while the work ran, by the system that ran it, in a form a stranger can recompute.

Why screenshots and chat logs fail as evidence

Screenshots are the evidence most people reach for first, and they prove the least. A screenshot of a finished dashboard, a sent email, or a completed checklist shows a state, not the work that produced it. It does not show which tool ran, what it returned, whether the state came from the run in question or from a manual fix afterward, or whether three failed attempts came before the one tidy image.

A screenshot shows what a screen looked like at a moment someone chose. It can be cropped, staged, retaken after an edit, or taken from a different run entirely, and nothing inside the image lets a checker tell the difference. A chat log is weaker still: it is a transcript the operator controls, it can be edited or regenerated before it is handed over, and it records what the AI said, which is the claim, not the work. Logs kept by the tool itself are better, but they are usually private, mutable by whoever runs the system, and written for debugging rather than for a stranger to check.

That is why the proof standard on this page is a receipt, not a picture. A verifiable receipt is fixed at execution time and carries a fingerprint over its saved result, so editing the result after the fact breaks the check. The checker needs nothing from you: not your login, not your session, and not your cooperation. Proof that requires the prover to be trusted is not proof, it is a request for trust with better formatting.

The check, step by step

First, set the work up so there is something to check. Connect your AI client to the Zambo MCP at https://zambo.dev/mcp: one URL, no key, and it works with the clients and frameworks people already use, including ChatGPT, Claude, Claude Code, Cursor, Gemini, Grok, Codex, Copilot, VS Code, Windsurf, Cline, Zed, n8n, CrewAI, and LangChain. Zambo is the verifiable work layer for AI: 153 tools in total, 35 on the default surface, and every call made through it leaves a verifiable receipt as it runs, fixed at execution time instead of written up afterward. Then work through these steps:

Split the claim into steps. Write down each thing the AI says it did, one claim per line: the search it ran, the file it wrote, the price it pulled, the message it sent. Each line will either resolve to a receipt or it will not, and that split is the whole method.
Run the work through the MCP. Each tool call saves its result and mints its receipt at the moment it runs. Work done outside the record leaves no receipt, so route the steps that matter through Zambo from the start.
Collect the receipts. Keep the receipt for each step with the deliverable, or gather the whole job into one handover link that carries the goal, the steps, the outputs, and the receipts together.
Open the free verifier. Go to zambo.dev/verify, enter the receipt, and run the check. It costs nothing and needs no signup, so the person checking never has to take the word of the person or the AI being checked.
Apply the five checks below to each step, then read the saved result itself instead of the summary written about it.
Decide from the record. Steps whose receipts pass all five checks are proven as recorded. Steps with no receipt, a broken fingerprint, or a saved result that does not match the claim are unproven, and unproven steps get redone, repriced, or disclosed as unchecked, never repeated as fact.

The five checks

Existence: a receipt exists for the call that supposedly did this part of the work. If there is no receipt, this step is unverified, whatever the summary says. Existence is the first gate when you prove what an AI did, because a check cannot start from a story.

Identity: the receipt names the tool that ran, when it ran, and which run it belongs to. That stops a receipt from a different job, a different day, or a different tool being offered as evidence for this claim, and it ties each claim to one specific call.

Fingerprint: the digest recorded over the saved result recomputes from the result itself. If anyone changed the saved result after the fact, even by one character, the recomputed digest will not match the recorded one, and the check fails in the open instead of passing quietly.

Self-consistency: the parts of the receipt agree with each other and with the claim: the tool on the receipt is the tool the claim needs, the time falls inside the run window, and the saved result is the result the claim describes. A receipt can be intact and still be the wrong receipt for this claim.

Chain: the receipts for a multi-step job link in sequence, each one pointing back to the step before it. A missing step shows as a gap in the chain and a reordered step shows as a break, so the sequence itself is checkable, not just the individual calls taken one at a time.

The whole loop is the same one described on the family anchor page, verify AI work, and the receipt underneath it is defined on what is a verifiable execution receipt. For a finished run read as one ordered record, see the AI agent audit trail. Those three pages and this one describe one method from four angles: run the work so it leaves receipts, then check the receipts instead of the story.

Ways people mean this

Prove what my AI did through an API: the same receipts are checkable programmatically, so a script can pull each receipt for a run, recompute its fingerprint, and report which claimed steps passed and which have no receipt underneath them, without a human clicking through a dashboard.

Prove what my AI did with open source checking: the receipt format underneath is AER-1, an open IETF Internet-Draft, so the check does not depend on a proprietary report from the same company that sold you the work. Any checker that implements the draft can recompute a receipt, which is what makes the proof portable.

Prove what my AI did for teams: a team does not need five people reading five chat threads. The receipts for a job, or one handover link carrying them, give every teammate the same record to check, so a reviewer in another timezone reaches the same verdict from the same evidence.

Prove what my AI did for a small business: an owner who delegates research, bookkeeping prep, or customer follow-up to an AI can check a day of that work in minutes before paying for it or acting on it, instead of either trusting the summary blindly or redoing the work by hand.

Prove what my AI did without screenshots: skip the screenshot folder entirely. Send the receipt links or the handover link. The checker opens the verifier, runs the five checks, and reads the saved results. If a step only exists as a screenshot, treat that step as the unproven one and say so.

Prove AI did the work: this shorter phrasing is the same problem: someone is asking whether AI, rather than a person, produced a result, and whether that result is the one on record. The receipt answers both parts, because it identifies the tool call that produced the saved result and fixes that result at execution time.

Prove AI agent actions: an agent in n8n, CrewAI, or LangChain takes actions in sequence: searching, writing, calling tools, sending. Each action that runs through Zambo leaves its own receipt, and the chain check shows the actions happened in the order claimed, with no step quietly missing.

Prove AI output: output is proven by comparison, not by polish. Open the receipt for the call that produced the output, read its saved result, and compare it with the output you were handed. If the handed-over version was edited, regenerated, or summarized on the way, the comparison shows exactly where it differs.

What the check proves, and what it costs

Once a job runs this way, the conversation changes. You are no longer asking a client, a manager, or yourself to believe a summary. You are handing over a record that any of them can check in the same way you did, and the five checks either pass or they do not. That is also what makes the record reusable later: in a payment dispute, a review, or a handover to the next person, the same receipts carry the same proof, because they never depended on your account or your memory.

What it proves is narrow and useful: the saved result on the receipt was not changed after the fact, the call it records is the call the claim points to, and a multi-step job ran in the order its chain shows. That is the floor every argument about AI work currently lacks. Run the loop once on a real call first: the live demo runs one call and hands you its receipt, free, in about ten seconds.

The Day Pass is the fix for a day of unverifiable work

Every unverifiable day has the same shape: the work is done, the summary sounds fine, and there is nothing underneath it to check. The fix is not a better summary, it is recording the next day properly. A Day Pass turns on a full day of recorded work: route the day through the Zambo MCP, let each call leave its receipt, and by evening the question on this page answers itself, because the proof was captured while the work ran instead of reconstructed after someone asked.

For $0.99, a Day Pass covers 24 hours of recorded work through the Zambo MCP at https://zambo.dev/mcp. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders. Public handovers are free to send and free to open. Humans pay by PayPal or card through PayPal; agents pay USDC on Base through x402. Checking any receipt at zambo.dev/verify stays free either way.

Honest limits

Three edges, stated plainly. First, A verifiable receipt proves the saved result was not changed after the fact; it never proves the result was right, the work was well done, or the price was fair. That judgment stays with the checker, now made against a record that was actually checked. The record also only covers steps that ran through connected tools: work done entirely outside it has no receipt, and an honest record lists that work as unverified instead of smoothing it over. AER-1, the receipt format underneath, is an IETF Internet-Draft (draft-zambo-aer1-24), not a finalized standard. Second, proof of what the AI did is not proof that the AI should have done it: a receipt can faithfully record a call that was the wrong call to make. Third, older work that ran before you connected Zambo cannot be given receipts after the fact; the honest answer for that work is a manual outcome check, and the fix applies from the next job forward.

Frequently asked questions

How can I prove what my AI did?

Split what the AI claims into individual steps, then find the receipt for the call behind each step. Check that a receipt exists, that it names the right tool and run, that its fingerprint recomputes over the saved result, that its fields agree with the claim, and that its place in the chain is intact. Steps that pass are proven as recorded. Steps with no receipt are unproven, and the honest move is to redo them, reprice them, or disclose them as unchecked.

Can I prove what my AI did without screenshots?

Yes, and the receipt is stronger than the screenshot it replaces. A screenshot shows a state someone chose to capture and can be cropped, staged, or taken from a different run. A verifiable receipt was fixed when the call ran and carries a fingerprint over its saved result, so a checker at zambo.dev/verify can recompute it and see whether the result changed afterward. Send receipt links or one handover link instead of an image folder.

How do I prove AI agent actions to a client or a manager?

Send the record, not a walkthrough. A handover link carries the goal, the steps in order, the outputs, and the receipt for each step, and the client opens it in a browser with no account. They can check any receipt free at zambo.dev/verify and read the saved result behind each claimed action. Actions with no receipt stay listed as unchecked, which is what makes the checked ones believable.

What does a receipt prove about AI output, and what does it not prove?

It proves the output you were handed matches the saved result that was fixed when the call ran, unchanged since. It does not prove the output was correct, complete, or worth what was charged for it. Correctness is a judgment the checker makes against the record: the numbers, the code, the sources. A checked record makes that judgment possible on evidence. An unchecked summary asks the checker to skip the judgment and trust the AI instead.

How much does it cost to prove what my AI did for a day?

Checking receipts is free at zambo.dev/verify, with no signup, and the live demo is free. For $0.99, a Day Pass covers 24 hours of recorded work through the Zambo MCP at https://zambo.dev/mcp, which runs 153 tools in total with 35 on the default surface, so a full day of AI work leaves a full day of receipts. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders.