---
title: PubMed Search troubleshooting
description: Zero results, a 45-second timeout, NCBI rate limits, and the missing abstract that advanced search never fetches — plus why you filter by date, not sort.
tags:
  - pubmed
  - troubleshooting
---

This page helps you diagnose the common problems with a [PubMed](/docs/pubmed) search — zero results, timeouts, rate limiting, missing abstracts, irrelevant hits, and query-construction errors — most of which trace back to how the E-utilities query is built.

## Common issues and fixes

PubMed Search uses the NCBI E-utilities API to retrieve biomedical literature. Most issues stem from query construction, rate limiting, or API availability.

## Empty or zero results

- **Overly specific query**: Reduce the number of AND-combined terms. Start broad and narrow down.
- **Spelling or terminology**: PubMed is sensitive to exact terms. Try synonyms or MeSH terms instead of free text. For example, "heart attack" may return fewer results than "Myocardial Infarction[MH]".
- **Date range too narrow**: Widen your date filters or remove them entirely to confirm results exist.
- **Wrong field tags**: Verify that field tags are correctly formatted (e.g., `[AU]` for author, not `[Author]`).
- **Non-indexed content**: Very recent articles may not yet be indexed. PubMed indexing can lag by days to weeks.

## Timeout errors

- PubMed Search uses a 45-second timeout for the multi-step ESearch + ESummary process.
- **Complex Boolean queries** with many OR branches can be slow. Simplify the query or reduce the result limit.
- **NCBI server load**: NCBI experiences peak traffic during US business hours. Try again during off-peak times.
- **Large result sets**: Requesting 100 results with a complex query is slower than requesting 20. Ask for fewer results.

## Rate limiting

- NCBI caps E-utilities at 3 requests/second without an API key, or 10 requests/second with one configured. The platform paces requests to stay under whichever cap applies and manages this automatically.
- If you see rate limit errors, it usually means multiple concurrent searches are active across your session. Wait a few seconds and retry.
- Rapidly chaining multiple PubMed queries (e.g., searching 20 different terms in quick succession) may trigger temporary throttling.

## Missing abstracts

- A plain PubMed search calls EFetch alongside ESummary and returns an `abstract` field for every article — but EFetch has nothing to return for records where PubMed itself has no `` (common for older or some non-English-language articles), so the field comes back as an empty string for those.
- A search filtered by author, affiliation, date range, or article type never calls EFetch, so its results never carry an `abstract`, `mesh_terms`, or `keywords` field. When you need the abstract text with a filter, ask Ivy to write the filter into the query as a field tag (for example `2024[DP]` or `Doudna JA[AU]`).

## Unexpected results or irrelevant articles

- PubMed results always come back sorted by relevance; there is no way to re-sort them by date. To bias toward recent articles, ask for a date range (or add a `[DP]` field tag to the query) instead.
- Use phrase searching with double quotes for multi-word concepts: `"immune checkpoint"` instead of `immune checkpoint`.
- Add `NOT review[PT]` to exclude review articles if you only want primary research.
- Use MeSH terms for precise concept matching. MeSH terms capture synonyms and related concepts automatically.

## Query construction errors

- **Unmatched brackets**: Ensure every `[` has a matching `]` in field tags.
- **Case sensitivity**: Boolean operators (AND, OR, NOT) must be uppercase. Field tags are case-insensitive.
- **Empty query**: An empty or whitespace-only query is rejected, and a filtered search needs at least 2 characters. Provide a meaningful search term.

## Still stuck?

When reporting an issue, include:
- The exact query text you entered (or the chat prompt)
- Whether you asked for filters (author, affiliation, date range, article type)
- The error message or unexpected behavior you observed
- Your session or job ID if available

## Read next