Turn a research report into a campaign, with the required disclosures about how the research was done

A campaign package of message angles from a research report, with the required methods disclosures.

Choose one of three approaches. The steps for that option appear below.

Choose an option above. On a phone, swipe the tabs left or right to see all options →

ChatGPT: builds the table and drafts the package

ChatGPT builds the finding table and drafts the package in one conversation. You check both stages.

  1. You

    Clear the inputs

  2. ChatGPT

    Build finding table

  3. You

    Check the table

  4. ChatGPT

    Draft the package

  5. You

    Check and decide

  1. Have ready: report, appendix, rights, objective

    Use: You

    Collect and clear the approved report, methods appendix, usage rights and campaign objective for permitted use.

    Check

    Check the report is the approved version, usage rights are attached, and every input carries a source ID and date.

  2. Build the finding-to-message table

    Use: ChatGPT

    Upload the cleared inputs and ask for the finding-to-message table with denominators, limitations and quotation permissions, plus source references and uncertain facts.

    Details

    What to paste into the tool

    Task: Finding-to-message table with denominators, limitations and quotation permissions; Source references; Uncertain or absent facts. Using only the supplied report and methods appendix, build a finding-to-message table. For each finding: the exact statistic with its denominator and population, the methods limitations that constrain it, the quotation permissions from the usage rights, and the candidate message angle. Keep the source ID, date and page or section reference on every row. List missing or uncertain facts. Do not add facts beyond the uploads and do not generalise any statistic beyond its stated scope.

  3. Check the table against the originals

    Use: You

    Compare the working record with the original inputs before drafting.

    Check

    Compare the record with the originals, confirm permitted use, and resolve or retain missing facts before drafting.

  4. Draft the campaign package

    Use: ChatGPT

    In the same conversation, turn the findings into campaign messages, keeping denominators, methods limitations and permitted quotations, with open questions retained.

    Details

    What to paste into the tool

    Task: Draft evidence-led campaign package with message angles and required methodological disclosures; Open questions and approval dependencies retained beside the recommendations. Using only the reviewer-checked finding-to-message table, draft an evidence-led campaign package: message angles built only on the listed findings, with the required methodological disclosures beside every promoted statistic - its denominator, scope, caveat and quotation permission. No causal claim may exceed what the report states. Keep the source ID and date beside each statistic exactly as supplied. List open questions and approval dependencies beside the recommendations.

  5. Check the package and decide

    Use: You

    Check the draft against the checked record and decide on corrections.

    Check

    Check every promoted statistic keeps its denominator, scope and caveat, no causal claim goes beyond the report, every required methods disclosure is present, and every open question is still listed. Decide approval and any subsequent execution separately.

Before you paste into any AI tool, remove any confidential information. Put it back at the end.

Tools in this workflow

Deciding between ChatGPT and Claude for this? See the comparison.

Final check

The Content lead and research owner compare the working record with the original inputs and confirm permitted use before drafting.

The Content lead and research owner apply the check: Every promoted statistic keeps its scope and caveat and no causal claim exceeds the report. Human confirmation by Content lead and research owner: evidence and unresolved questions reviewed.

No option sends, publishes, deploys or changes any account; approval and any execution are separate human decisions.

Check the result against your original inputs. Approve it before sharing or acting.

What still needs checking
  • No vendor page evidences statistical fidelity of any assistant; the statistic-level human check is load-bearing in every path.
  • The unresolved questions reach the draft through the explicit handoff from the checked record, and the constraints come from the owner-supplied constraints named in the starting input.
  • Vendor pages checked 2026-10-28. Check plan terms and limits with the vendor, and check each tool on your own data.
  • The options were chosen from the 154-tool catalogue by documented upload, extraction and drafting operations; purpose-built catalogue tools such as a source-grounded notebook or an SEO content tool were not compared against these options.

Share this page