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 extracts the design facts, tabulates with visible Python and drafts the brief; the survey researcher approves every denominator.
Upload questionnaire and responses
Extract facts, tabulate
Check ledger, rerun code
Draft the descriptive brief
Researcher checks denominators
Use: You
Have ready: questionnaire, response table, recruitment description, bounded question and permitted-use limits. Upload all to one ChatGPT chat with a source inventory.
Check every in-scope document is uploaded and readable, and scope and permission limits hold; flag missing or unreadable inputs.
Use: ChatGPT
Ask ChatGPT to extract design facts (recruitment, fieldwork dates, question wording) into a ledger, then tabulate with explicit denominators and missing counts, showing the Python.
What to paste into the tool
Task: Draft evidence ledger of design facts; First tabulation with visible code, denominators and missing counts. First, quote the recruitment description and each question's exact wording into a ledger with source locations. Then analyze the response table with Python: for each question give counts, percentages with the exact denominator used, and the number and share of missing responses. Show the code and state every assumption.
Use: You
Check the design-fact ledger against the original documents; inspect or rerun the generated code against the response table; ask for fixes in the same chat.
Compare extracted passages, dates, identifiers and numeric fields against the originals before relying on any number; check scope, source access and permission limits; flag missing or unreadable inputs. Checked ledger with source_id, source_date when known, passage_or_cell_location, extracted_value_or_quote, evidence_status, permission_note and unresolved_question fields (original units and time zones retained).
Use: ChatGPT
Ask ChatGPT to draft the brief with per-question denominators and wording, missingness and sampling limits; generalisation withheld unless recruitment evidence justifies it.
What to paste into the tool
Task: Draft descriptive survey brief with denominators, missingness and sampling limits; Source-linked uncertainty and exception notes. Draft the descriptive brief from the checked ledger and tabulation. For every reported percentage, name the denominator and the exact question wording. Summarize missingness per item. Describe the recruitment method and state the sampling limits. Withhold any population generalization unless the recruitment evidence supports it, and say which it is.
Use: The survey researcher
Check weighting and denominator choices; each percentage names denominator and wording; generalisation is justified from the recruitment evidence or withheld. Approve or hold.
Julius AI tabulates, builds the ledger and drafts the brief; you recompute the headline numbers and the researcher decides.
Upload table and documents
Tabulate, ledger, draft
Recompute headline numbers
Write the brief
Use: You
Have ready: questionnaire, response table and recruitment description. Upload all three to Julius AI.
Check each file uploads readably; flag missing or unreadable inputs.
Use: Julius AI
Ask for per-question tables with denominators and missing counts (export as CSV), then the design-fact ledger and a draft brief with sampling limits.
What to paste into the tool
Task: Exported CSV tables: per-question frequencies, percentages with denominators, missing counts. For each question column in this table: show the count and percentage of each response option with the exact denominator, and the count and share of missing responses. Then export all result tables as CSV. Task: Evidence ledger and draft descriptive brief. From the uploaded questionnaire and recruitment description, list each question's wording and the recruitment facts with their locations. Then draft the descriptive brief from the tabulation tables: per-question results naming the denominator and the question wording, a missingness section, and sampling limits drawn from the recruitment description. Keep original units. Say plainly where the documents do not state something. Do not send, buy, publish or change anything.
Use: You
Recompute headline counts and denominators from the raw table against the CSVs; compare the ledger and draft claims against the originals; resolve every difference.
Each percentage must name its denominator and question wording; no inspectable code is offered, so recompute before reuse.
Use: The survey researcher
Write the final brief by hand from the checked tables: per-question results naming denominator and wording, missingness, sampling limits; justify or withhold generalisation.
Retain unresolved findings instead of silently resolving them.
Gemini computes as tables inside the workbook, so every number is checked in place; you write the brief from it.
Import the response table
Build breakdown tables
Check tables in place
Build the summary sheet
Check summary, write brief
Researcher checks, decides
Use: You
Have ready: questionnaire, response table and recruitment description. Import the response table into a spreadsheet; keep the reference documents at hand.
Check the full table imported without truncation; check scope, source access and permission limits; flag missing or unreadable inputs.
Use: Gemini
Ask Gemini for per-question breakdown tables with counts, percentages and missing-per-item counts, produced as tables in the workbook.
What to paste into the tool
Task: Per-question breakdown tables and missing-count tables inside the workbook. For each question column in the response table, create a breakdown table with counts and percentages per response option, naming the denominator, and a table of missing responses per question.
Use: You
Check each table against the full raw data in place; add a ledger sheet with exact question wording, recruitment facts and source locations.
Gemini may read only a sample of rows; check every table against the full dataset.
Use: Gemini
Ask Gemini for one compact summary table of the key percentages with denominators, kept as a sheet the reviewer can audit.
What to paste into the tool
Task: Auditable summary sheet of key percentages with denominators. Create one summary sheet listing each key question with its top responses, percentages and denominators, drawing only from the checked breakdown tables in this workbook.
Use: You
Check the summary sheet against the checked breakdown tables, then write the brief: per-question results with denominators and wording, missingness, sampling limits.
Check the summary sheet against the checked breakdown tables before the brief cites it. Each percentage must name its denominator and question wording; generalisation is justified from the recruitment evidence or withheld. Checked ledger with source_id, source_date when known, passage_or_cell_location, extracted_value_or_quote, evidence_status, permission_note and unresolved_question fields (original units and time zones retained).
Use: The survey researcher
Check weighting and denominator choices against the workbook; confirm the generalisation wording; retain unresolved findings instead of silently resolving them. Approve or hold.
Best when: You want to inspect the exact code behind every number
What it does: ChatGPT tabulates and drafts; you audit the code
Strong at: Python code, outputs and assumptions are reviewable
Limit: Vendor warns the first method or chart may mismatch intent
Free plan: Yes
Best when: You want conversational tabulation plus ledger and brief drafting
What it does: Julius AI tabulates and drafts; you recompute numbers
Strong at: Plain-language requests with CSV export
Limit: No inspectable code shown; every number needs recomputing
Free plan: Yes, limited (75 one-time + 25 daily credits)
Best when: You want to audit every number inside the spreadsheet itself
What it does: Gemini computes in the workbook; you verify in place
Strong at: Computation and checking happen in one artefact
Limit: Gemini may read only a sample of rows; full-data checks needed
Free plan: Yes
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Gemini for this? See the comparison.
Research analyst verifies every tabulation against the raw table before the brief is written.
Survey researcher checks weighting and denominator choices and confirms the check, including whether population generalisation is justified or withheld.
OpenAI warns that exact values may be misread in image-based tables, scanned files or complex layouts.
Check the result against your original inputs. Approve it before sharing or acting.