Direct answer

Show proof of AI work

To show proof of AI work, you send evidence the other side can recheck without trusting you: the receipt for each step of the job, or one handover link carrying them all. The client, manager, or teammate opens the link, runs any receipt through the free verifier, and sees the five checks pass: existence, identity, fingerprint, self-consistency, and chain. Screenshots ask for belief. A verifiable receipt, fixed when the work ran through Zambo, asks for a recompute, and that difference is what makes it proof you can show.

Show yourself first: run the live demo, take its receipt to zambo.dev/verify, and watch the check a client would run. Then read verify AI work for the full method behind what you are showing.

Updated October 20, 2026.

The situation

The moment arrives in a meeting, an email thread, or a payment dispute. A client says: show me the AI actually did this work. A manager says: show me proof before this goes to the board. A teammate inheriting your project says: show me what was already done so I do not redo it. In each case the AI claims the work is done, you believe it, and belief is not transferable. What the other person needs is not your confidence, it is evidence they can test with their own hands.

Most people walk into that moment holding the wrong artifacts: a slideshow of screenshots, a chat export, a summary document written for the occasion. Each one was assembled by the person asking to be believed, which is precisely the flaw the other side can feel even when they cannot name it. Showing proof is a design problem: the evidence has to be fixed before the dispute, in a form whose checking does not route through you.

Why screenshots and chat logs fail as evidence

Screenshots are the weakest thing you can show, and experienced clients know it. They can be staged after manual fixes, cropped to hide failures, or captured from a test run. Chat logs are a transcript you control. Even a live screen share only shows the present state, and proves nothing about which calls produced it or whether the outputs on display are the outputs the system returned at the time. Every one of these asks the viewer to extend trust to the presenter, which is the thing proof was supposed to replace.

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.

A receipt presentation works differently. You show the link, and then you stop talking while the other side checks. The verifier is public, the check is free, the fingerprint either recomputes or it does not, and your credibility is no longer a load-bearing part of the evidence. That is also why honest gaps help rather than hurt: a step shown as unchecked, next to steps that pass, reads as a record. A flawless screenshot deck reads as a performance.

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. Decide what you are showing before you show it: the two or three steps the decision turns on, with the full chain one click behind them. Proof shown well is focused: the payment step, the delivery step, the test step, each with its receipt ready to open.
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. Let the checker drive. Answer questions by opening the saved result behind the receipt, not by narrating over it. If a step has no receipt, say so first: showing the gap yourself is what makes the passing steps believable.

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 show proof of AI work, 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

Show proof of AI work through an API: for a technical audience, the same receipts can be presented as data: a check script recomputes every fingerprint in the run and prints the verdicts live. The audience watches the record pass its own checks rather than watching you describe them, which lands differently in a room full of engineers.

Show proof of AI work for teams: a team shares one record instead of five retellings. The handover link goes in the project channel, every teammate checks the same receipts, and the next standup argues about the work itself, because what happened is no longer up for competing memories.

Show proof of AI agent actions: present actions in chain order: the search, the write, the send, each with its receipt. A client who can see the action sequence, and recompute any link in it, stops asking whether the agent did anything and starts discussing what it did, which is the productive argument.

Show proof of AI generated work: generated work, drafts, analyses, summaries, is shown with the receipt for the call that generated it. The viewer compares the shown output against the saved result and sees they match, or sees exactly where a human edited afterward, which is honest information either way.

Show proof of AI output: output proof is a comparison performed in public: the output in the deliverable on one side, the saved result on the receipt on the other, and the fingerprint check confirming the saved result is the one fixed at execution time. Nothing in that presentation depends on your say-so.

Show proof of tool calls: when the dispute is about a specific call, a price pulled, a record changed, a message sent, show that call receipt alone. Its identity names the tool and run, its fingerprint covers the saved result, and the argument narrows to one checkable fact instead of the whole project history.

Show proof of what my AI did: the retrospective question gets the retrospective artifact: the receipts in chain order are the showing. Walk the chain, step by step, and what your AI did is no longer a reconstruction from chat scrollback; it is the record, presented in the order it happened.

What the check proves, and what it costs

Proof that can be shown this way travels. The client forwards the link to their finance person, who checks the same receipts and reaches the same verdict. The manager puts the link in the board pack. The teammate opens it next month and trusts it exactly as much as you did, because none of the trust was ever placed in you. That portability is the real product here: evidence whose strength does not decay with distance from the person who produced the work.

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

People rarely know in advance which day of AI work they will be asked to prove. The practical answer is to make the days that matter provable in advance: for $0.99, a Day Pass records a full day of work through the Zambo MCP, every call leaving its receipt, so when the meeting, the invoice, or the dispute arrives, you are not assembling evidence under pressure. You are sending a link to evidence that already exists. Proving a day costs a dollar. Reconstructing one costs a week and convinces nobody.

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, showing proof shows the record, not your whole process: tools outside the record and human work between calls will not appear, and claiming the record shows more than it does will cost you the credibility the record just earned. Third, proof settles facts, not relationships: a client can accept every receipt and still dislike the work, and the receipt will not argue that point for you.

Frequently asked questions

How do I show proof of AI work to a client?

Send a handover link for the job, or the receipt links for the steps that matter. The client opens the link in a browser with no account, sees the goal, the steps, the outputs, and a receipt per step, and can check any receipt free at zambo.dev/verify. Let them run the check themselves rather than presenting screenshots of it. Steps with no receipt should be listed as unchecked by you first, before the client finds them.

What is the strongest proof of AI agent actions I can show?

The receipts for the actions, in chain order, checked live in front of the other side. Each receipt names the tool that ran and fixes its saved result with a fingerprint, so the checker can recompute any action and read what it actually returned. A chain with no gaps shows the sequence is complete. That combination, per-action receipts plus an intact chain plus a public free check, is stronger than any summary, log export, or screen recording.

Can I show proof of AI work to a whole team from one link?

Yes. One handover link carries the goal, the steps in order, the outputs, and the receipt for each step, and everyone with the link sees the same record and can check the same receipts free at zambo.dev/verify. That sameness is the value for a team: no screenshots forwarded out of context, no competing versions of what the AI did, and a new teammate can recheck the record months later and reach the same verdict.

Does proof of AI output prove the output was right?

No. Proof of AI output proves the output you are showing is the saved result that was fixed when the call ran, unchanged since. Whether that output is right, good, or worth paying for is a judgment the viewer makes against the proven record. That separation is a strength in the room: you are not asking anyone to take correctness on faith, only to agree on what the output actually was, which the receipt settles.

How much does it cost to show proof of a day of AI work?

Showing and checking are free: anyone can open a handover link and run any receipt through zambo.dev/verify with no signup, and the live demo is free. Recording the day costs $0.99 for a Day Pass covering 24 hours through the Zambo MCP at https://zambo.dev/mcp, 153 tools in total with 35 on the default surface. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders.