How to find papers that cite a study in ChatGPT

Get the list of papers that cite a study, see how many there are, and pick out the ones that matter for your question. With the Firecrawl plugin, ChatGPT can run a forward citation search through Firecrawl Alexandria, using the Firecrawl Research Index and the citation links in PubMed.

• Systematic reviewers screening the work that built on a key trial.
• Researchers checking who cited their own paper and in what context.
• Students looking for recent follow-up studies to a paper they were assigned.

Plugin: the Firecrawl plugin. It lets ChatGPT search papers and follow citation links in Firecrawl Alexandria, billing any paid call to the Firecrawl team you sign in with.

Explore Firecrawl Alexandria data providers across categories

Use your own study in place of the vaccine trial, ideally with its PMID:

@firecrawl Find papers that cite "Safety and Efficacy of the BNT162b2 mRNA
Covid-19 Vaccine" (NEJM 2020, PMID 33301246). Use PubMed and PMC and the
Firecrawl Research Index in Firecrawl Alexandria, and show me the price of
each tool before you run it.
 
Give me:
- How many PubMed articles cite it, and the 20 newest with journal, date
  and PMID.
- From the Firecrawl Research Index, the citing papers most relevant to
  booster doses in older adults, with a one-line summary of each abstract.
- Which of the citing papers are reviews.
 
List only papers the tools returned, each with its identifier.

• Get Firecrawl into ChatGPT. The button below takes you to Firecrawl's listing in ChatGPT.
• Link a Firecrawl account. Paid lookups are charged to it, and you can sign up for free.
• Have an identifier ready. A PMID, DOI or exact title helps ChatGPT land on the right seed paper, since many papers share similar titles.

A forward citation search depends on an index of who cited whom, and ChatGPT doesn't carry one. These six problems follow from that.

1. References that don't exist. When researchers asked ChatGPT to reproduce the literature searches behind published systematic reviews, 28.6% of GPT-4's references and 39.6% of GPT-3.5's were hallucinated, and the authors advise validating every generated reference (JMIR, 2024). OpenAI also notes ChatGPT can make up citations (OpenAI).
2. Recent citing papers fall after the cutoff. Without a live tool, ChatGPT answers from training data that stops at a cutoff date (OpenAI), and the papers that cited a study last month are exactly what a forward search is for.
3. Google Scholar can't be pulled in bulk. Scholar says it can't provide bulk access to records, shows at most 1,000 results for a query and asks automated tools to respect its robots.txt (Google Scholar). A web search that lands on a "Cited by" page sees only a slice of it.
4. The main citation databases need a subscription. Scopus and Web of Science are subscription based (UT Southwestern), so ChatGPT can't query them unless you have access yourself.
5. Every database gives a different count. The same guide notes that because these sources index different journals, times-cited counts will vary. UC San Diego adds that Google Scholar counts theses and some non-peer-reviewed sources that Web of Science leaves out (UC San Diego).
6. Preprints and journal versions split the count. Metrics for a preprint and its published version may not be merged, and each can carry its own citation count (UC San Diego). A search for one version can miss papers that cited the other.

Three pieces do the work when you ask for citing papers:

• Firecrawl Alexandria: Firecrawl's route from AI agents to structured data. It holds more than 100 official providers, each on a paid agreement, beside Firecrawl's four indexes (web, Research, Developer and Government). Firecrawl tested this on 845 tasks while holding the model and prompts fixed, and answer quality came out 21% higher for agents using Firecrawl Alexandria compared with agents relying on their default web tools.
• The Firecrawl Research Index: Firecrawl's own index of scientific papers, covering biomedical, life-science and clinical literature from PubMed, bioRxiv and medRxiv alongside arXiv. Its related-papers tool follows citation links from a seed paper in three directions: citing papers, references and similar work.
• PubMed's citation links: the PubMed and PMC tool reads NCBI's record of which PubMed articles cite a given PMID. It reports the full count, and when we queried it directly on Oct 9, 2026, the newest citing articles came back first.
• Identifiers on every result: each paper comes back with a PMID, PMCID, DOI or arXiv ID, so you can open it and check it yourself.

Write for the job the citing papers will feed. For a systematic reviewer screening a trial's follow-up work:

@firecrawl How many PubMed articles cite PMID 33301246? List the 50 newest
with title, journal, publication date, PMID and publication type, and flag
any that are reviews or meta-analyses.

For a machine learning researcher tracing a paper's lineage:

@firecrawl Find "Attention Is All You Need" in the Firecrawl Research Index
and list the papers it cites that relate to neural machine translation,
with their arXiv IDs and a one-line summary of each abstract.

Naming the paper exactly and listing the fields you want keeps ChatGPT from guessing at either.

Rank citing papers by your question

@firecrawl Use the Firecrawl Research Index to find papers citing the
BNT162b2 phase 3 trial (NEJM 2020) and rank them for relevance to lipid
nanoparticle delivery. Give title, identifier and the abstract's main claim.

Check a paper before you cite it

@firecrawl Inspect this paper in the Firecrawl Research Index and give me
its exact title and abstract: "A Multi-Scale Attention-based Reconstruction
Fusion Network for Motor Imagery Classification."

Pull passages from a paper's full text

@firecrawl Read "Attention Is All You Need" in the Firecrawl Research Index
and quote the passages that explain multi-head attention.

Price check first

@firecrawl Which tools can find papers that cite a study? List each one
with its price per call before running anything.

Finding a tool and seeing its price costs nothing. A priced tool takes Firecrawl credits only at the moment ChatGPT runs it; see Firecrawl pricing for what each plan includes. Every Firecrawl Research Index tool shows a price of 0 credits in the catalog, and our direct Oct 9, 2026 API calls to them were billed nothing.

ChatGPT tells you, lookup by lookup, which tool ran and what it was billed.

• How to search arXiv and PubMed in ChatGPT: find the seed paper and related work by topic before following its citations.
• PubMed and PMC references: the same PubMed tool can list a paper's own references that are in PubMed, the backward half of a citation search.
• OpenReview: search the public reviews and discussion on machine learning conference papers, for venues such as ICLR that publish them.

• PubMed counts only PubMed. The cited-in count covers citing articles indexed in PubMed, so a thesis, a book or a paper in a journal PubMed doesn't index won't be counted. One call returns the count and up to 50 records.
• The Research Index's citing lists are partial. For the BNT162b2 trial it returned 130 citing papers against PubMed's 9,028, all records added in 2026. For "Attention Is All You Need" it returned 21 citers, all biomedical journal articles and no arXiv papers. Use it to rank and screen, and use PubMed for the count.
• Full text varies by paper. Passages came back for an arXiv paper, but for the PubMed records we tried the passage tool returned nothing, so you may only get the abstract.