The Scenario
A postdoctoral researcher is running a meta-analysis and has 12 seed papers already confirmed for inclusion. The next step in the protocol is candidate expansion: find papers similar to the seeds that the team hasn't encountered yet. The standard move is to hand the list to a research assistant, ask them to mine Semantic Scholar for recommendations, and wait two days. There is no research assistant this quarter.
The bad version:
- Open the Semantic Scholar page for seed paper 1, scroll to the Recommended Papers section, copy the first 10 titles, switch to the workbook, paste — then reformat the paste because the clipboard included links and author names in a single column.
- Repeat for seed paper 2. Notice the recommendations for paper 2 overlap heavily with paper 1, but you have no way to deduplicate until you've done all 12 and manually compared.
- After 90 minutes you have a messy, partially overlapping candidate list with inconsistent formatting across 12 pastes.
The expansion step is supposed to surface candidates the team missed — not occupy a full morning of mechanical copying.
The Easy Way: One Prompt in SheetXAI
SheetXAI is an AI agent that lives inside your Excel workbook. It reads the seed paper IDs in the SeedPapers worksheet, calls Semantic Scholar's recommendation engine for each one, and writes the suggested papers as rows in a new worksheet — ready for triage.
Here is the prompt for this task:
Use each paper ID in the SeedPapers sheet as a positive example, get Semantic Scholar recommendations for each, and write all results to a new Excel sheet called ExpandedReadingList with one recommended paper per row
What You Get
- An ExpandedReadingList worksheet with one row per suggested paper.
- Columns: Seed Paper ID, Title, Year, Authors.
- All 120 candidates (12 seeds x 10 recommendations) land in a flat table — ready for deduplication and triage in one sort-and-filter operation.
- Duplicates across seeds appear as multiple rows so you can see which papers the recommendation engine independently surfaced from different starting points.
What If the Data Is Not Quite Ready
The SeedPapers worksheet has titles, not Semantic Scholar paper IDs
For each paper title in the SeedPapers worksheet, resolve it to a Semantic Scholar paper ID, then call the recommendation engine and write the top 10 recommended papers with title, year, and authors into the ExpandedReadingList worksheet with a Seed Title column
You want to exclude papers already in the SeedPapers worksheet from the recommendations
For each paper ID in the SeedPapers worksheet, get the top 10 Semantic Scholar recommendations, filter out any recommended paper whose ID appears in the SeedPapers worksheet, and write the remaining results to the ExpandedReadingList worksheet with title, year, authors, and seed paper ID
You need recommendations filtered to a specific field of study
For each paper ID in the SeedPapers worksheet, call Semantic Scholar's recommendation engine, filter results to papers tagged under the Biology field of study, and write the top 5 matching recommendations per seed to the ExpandedReadingList worksheet with title, year, and authors
Resolve seed IDs, get recommendations, deduplicate, and rank by citation count in one pass
For each paper ID in the SeedPapers worksheet, fetch the top 10 Semantic Scholar recommendations, exclude papers already in the SeedPapers worksheet, write each unique recommendation to the ExpandedReadingList worksheet with title, year, authors, and citation count, and sort the full table by citation count descending so the most-established candidates appear first
Try It
Get the 7-day free trial of SheetXAI and open any Excel workbook with a worksheet of confirmed seed paper IDs. Ask SheetXAI to expand the reading list into a structured candidate table — and start triage the same afternoon instead of waiting on a research assistant.
See also: Bulk Search Research Topics and the Semantic Scholar hub overview.