Moonlit uses email header checks to verify AI research answers

Moonlit says occasional checks of signed email headers can help verify AI-generated research answers without giving the company direct access to a participant’s inbox. The approach is meant to make delegated survey work more credible for both people doing the tasks and the companies paying for research.

Why it matters: - Moonlit is trying to make AI-assisted research contributions easier to trust without requiring full inbox access. - The approach gives companies a way to ask what supports an answer while still letting participants use their own AI assistants. - Accepted answers earn rewards only after verification, which ties payment to source review.

What happened: - Moonlit, a service from Vulture Labs, uses occasional checks of signed email headers as one input for verifying answers prepared by participants’ assistants. - The company says participants do not give Moonlit a direct inbox connection. - Participants review the proposed research answer and reward before deciding whether to submit it. - Research buyers receive aggregated results, and groups smaller than five are withheld. - Teams can discuss source requirements at research. - Participants can read the privacy explanation at privacy.

The details: - Many commercial emails use DKIM, a digital signature that lets a receiving system check signed parts of a message against the sender domain’s public key. - Where an assistant can retrieve suitable headers, Moonlit can use them to check aspects of a message’s origin and its association with the participant. - Moonlit discards the headers after the check. - A headers-only check does not prove the amount extracted from a message body, the completion of a purchase or the truth of a person’s opinion. - Moonlit combines source details with other review instead of treating a signature as a guarantee of every answer. - Work availability and eligibility determine earnings.

Between the lines: - The model is designed to sit between automation and verification: participants delegate drafting to AI, but Moonlit still wants a paper trail of sorts for source review. - The limit on smaller groups suggests Moonlit is also trying to reduce the risk of exposing individual responses in aggregated research output. - The verification method is useful, but it is intentionally partial. That keeps privacy boundaries intact while avoiding overclaiming what a signature can prove.

What's next: - Moonlit says companies can request a study through its research page. - People who already use an AI assistant can connect it through the setup guide. - People new to AI can use the same guide to choose a supported assistant and set it up for paid surveys. - Rewards will continue to depend on accepted answers, study availability and eligibility.

The bottom line: - Moonlit is pitching a middle ground: enough verification to support trust, but not enough access to turn participant inboxes over to the platform.