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 →
Claude classifies pairs, then assembles the two forms in the same chat, so the stage-1 table stays in the workspace.
Get ready
Classify pairs
Check the pairs
Assemble the forms
Solve and check
Use: You
The approved maths item bank, the blueprint and the permitted variation rules for two sittings. Upload them to one Claude chat.
Only material the data policy permits is uploaded. Check sizes and counts before upload: 500MB per file, 20 files per chat, 1000 PDF pages.
Use: Claude
Ask Claude to classify candidate item pairs against the matching criteria and list unmatched gaps.
What to paste into the tool
Task: Draft candidate pairing table and unmatched gaps. Classify candidate item pairs from this bank against the matching criteria: for each pair give the two item IDs, content target, solution-step count, difficulty cues and any mismatch. List unmatched gaps separately. Do not pair items the variation rules exclude.
Use: You
Check item exposure and allowable reuse against the pairing table. Correct it and keep unresolved items.
Check each pairing against the original bank. There are no inline citations; check against the bank directly.
Use: Claude
Ask for two draft forms with paired specifications and an equivalence-review checklist.
What to paste into the tool
Task: Assembled draft forms with paired specifications and equivalence-review checklist, not yet solved. Using only this corrected pairing table, assemble the two draft forms with their paired specifications and an equivalence-review checklist that records what was matched and what was not. Keep unresolved items visible. Matching specifications is not psychometric validation or proof of score equivalence.
Use: You
Solve both forms, compare wording, step count and difficulty cues, and check the checklist documents comparability limits.
Solve both and compare wording, step count and difficulty cues. Check every item matches the approved sources. Preparation only; do not submit learner work, publish material, alter official grades or make placement decisions automatically. Pair classification and form assembly rest on the lead's own checks; no official wording confirms these operations.
Gemini Notebook extracts item attributes. You pair items in Sheets, then it assembles the two forms.
Add sources
Build item table
Pair items in Sheets
Assemble the forms
Solve and check
Use: You
The approved maths item bank, the blueprint and the permitted variation rules for two sittings. Add them as notebook sources.
Only material the data policy permits is uploaded.
Use: Gemini Notebook
Generate a Data Table with one row per bank item: content target, solution steps, difficulty cues, exposure notes.
What to paste into the tool
Task: Item-attribute Data Table with one row per bank item. Extract an item-attribute table from these sources: one row per bank item with item ID, content target, solution-step count, difficulty cues and exposure notes as stated in the sources. Describe those rows and columns in the Data Table customization box. Do not propose pairings.
Use: You
Export to Google Sheets. Check rows against the bank. Pair items by content target, steps and difficulty cues. Add your pairing table as a source.
Check exposure and allowable reuse; keep corrections in the exported copy. Check the data policy allows re-adding the pairing table.
Use: Gemini Notebook
Ask for two draft forms with paired specifications and an equivalence-review checklist, using your checked pairing table.
What to paste into the tool
Task: Assembled draft forms with paired specifications and equivalence-review checklist, not yet solved. Using only the sources in this notebook, and following the pairing table source exactly, assemble the two draft forms with their paired specifications and an equivalence-review checklist that records what was matched and what was not. Keep unresolved items visible. Matching specifications is not psychometric validation or proof of score equivalence.
Use: You
Solve both forms, compare wording, step count and difficulty cues, and check the checklist documents comparability limits.
The evidence does not prove per-item attribute accuracy or that the forms assemble correctly, so the assessment lead verifies rows and solves both forms.
ChatGPT drafts a pairing table, then the two forms and checklist. You check every stage against the bank.
Prepare the list
Draft pairing table
Check the pairs
Draft the forms
Solve and check
Use: You
The maths item bank, blueprint and variation rules for two sittings. List candidate items with content targets, step counts and difficulty cues.
Check item exposure and allowable reuse. Check the list holds only what the task needs. Check your class data policy allows the upload.
Use: ChatGPT
Ask ChatGPT to draft the candidate pairing table against the matching criteria, listing unmatched gaps.
What to paste into the tool
Task: Draft a candidate pairing table. Using only the supplied candidate list, pair items that match on content target, step count and difficulty cues under the supplied matching criteria. List unmatched items as gaps; do not force a pair. Do not invent items and do not change the supplied cues. Matching specifications is not psychometric validation or proof of score equivalence.
Use: You
Check the draft pairing table against the bank and blueprint. Re-check exposure, allowable reuse and gaps, then correct it.
Check every pair against the bank. Form assembly is unevidenced.
Use: ChatGPT
Ask for two forms with paired specifications and an equivalence-review checklist, from the corrected table.
What to paste into the tool
Task: Assembled draft forms with paired specifications and equivalence-review checklist, not yet solved. Using only this corrected pairing table, draft the two forms with their paired specifications and an equivalence-review checklist that records what was matched and what was not. Keep unresolved items visible. Matching specifications is not psychometric validation or proof of score equivalence.
Use: You
Solve both forms, compare wording, step count and difficulty cues, and check the checklist documents comparability limits.
Solve both and compare wording, step count and difficulty cues. Check every item matches the approved sources. Preparation only; do not submit learner work, publish material, alter official grades or make placement decisions automatically.
Best when: You want the fewest moving parts
What it does: Claude pairs and assembles in one chat
Strong at: Pairing table never leaves the chat
Limit: 500MB per file, 20 files per chat, 1000 PDF pages
Free plan: Yes, limited (5-hour and weekly usage caps)
Best when: You want the matching judgement in a spreadsheet you control
What it does: Gemini Notebook extracts and assembles; you pair the items
Strong at: Clean item table exports to Sheets
Limit: Exporting and re-adding the table is manual
Free plan: Yes, limited (notebook and source caps)
Best when: You want AI to draft while you check every pair
What it does: ChatGPT drafts at each stage; you check each one
Strong at: Lead keeps the matching judgement by checking pairs
Limit: You check every pair against the bank
Free plan: Yes, limited (uploads and data analysis)
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Claude for this? See the comparison.
Never paste credentials or full payment details.
Assessment lead checks item exposure and allowable reuse before forms are assembled.
Assessment lead solves both forms and compares wording, step count and difficulty cues.
Forms must match agreed content and demand; solutions are checked but statistical equivalence is not claimed. Preparation only. You review the result; do not submit learner work, publish material, alter official grades or make placement decisions automatically.
Check the result against your original inputs. Approve it before sharing or acting.