---
title: The right fact, the wrong paper | Scholar Sidekick
description: An AI Overview cited our page's correct fact, then attributed PMID 37905776 to the paper at PMID 39075776 - the same digits, shuffled. A second run fixed it.
doc_version: "1.0"
last_updated: "2026-08-15"
---

# The right fact, the wrong paper

> The search below was me checking how a page we had shipped the day before was ranking. The answer engine cited it - correctly - and then attached the wrong paper to the identifier. I have skin in this game, so every receipt on this page is a public registry record you can check without trusting me.

*Published 2026-08-15. Canonical: https://scholar-sidekick.com/citation-integrity/right-fact-wrong-paper*

On 15 August 2026, the day after we shipped a page answering “does PMID 37905776 have a PMCID?”, Google's AI Overview cited it. The fact it took from the page was right. The paper it attached to the identifier was wrong - a real study that lives at PMID 39075776, the same eight digits in a different order. A second run of the same query, later the same day, got the pairing right. Both answers are preserved below, next to the registry records and a live verifier verdict, because the pair demonstrates something a single run could not: answer engines are probabilistic about the one thing a reference list cannot afford to be probabilistic about - which paper an identifier names.

## Three captures, one day

- **Google AI Overview, first run** (2026-08-15 (AEST), query “37905776 pmcid”) - *wrong pairing.* Cites the Scholar Sidekick page for the (correct) fact that PMID 37905776 has no PMCID, then states the identifier “references” a rural-health study that actually lives at PMID 39075776. Two sources about two different papers, joined into one claim about one identifier.
- **Bing organic results, same query** (2026-08-15 (AEST), query “37905776 pmcid”) - *correct.* No answer layer involved: the Scholar Sidekick page ranks first, and Europe PMC's record of the correct paper (the Kezirian letter) ranks second. Retrieval had the right association available.
- **Google AI Overview, second run** (2026-08-15 (AEST), same query, later the same day) - *correct pairing.* Now pairs PMID 37905776 with the Laryngoscope letter and DOI 10.1002/lary.31132, citing the same Scholar Sidekick page for both the fact and the DOI. One edge stays blurred: the title shown drops the letter's “In reference to” prefix, so the letter's PMID appears under the title of the meta-analysis it comments on.

## What the first run said

> The identifier 37905776 is actually a PubMed ID (PMID) rather than a PubMed Central ID (PMID 37905776 has no PMCID assigned). It references the following medical study: Title: Describing a medical school's rural activity footprint. Journal: Rural and Remote Health (Published July 2024).

Transcribed from a screenshot captured 2026-08-15 (AEST); AI Overviews have no shareable URL and no version stamp. The fact in the first sentence is correct and correctly attributed. The study described after it is the paper at PMID 39075776, not the paper at the searched identifier.

## What the two identifiers actually name

*Two real papers whose PMIDs contain the same eight digits. Records as returned by PubMed and Crossref on 2026-08-15 - unlike an AI answer, these lookups are repeatable.*

|  | PMID 37905776 (searched) | PMID 39075776 (described by run 1) |
| --- | --- | --- |
| What it is | A letter in *The Laryngoscope* commenting on a meta-analysis of orofacial myofunctional therapy | A research article on mapping a medical school's rural activity area |
| Title on record | In reference to Orofacial Myofunctional Therapy for Obstructive Sleep Apnea: A Systematic Review and Meta-Analysis | Describing a medical school's rural activity footprint: setting selection and workforce distribution priorities |
| Authors | Kezirian EJ | Fuller L, Beattie J, Versace VL, Rogers GD, McGrail MR |
| Journal | The Laryngoscope 2024;134(3):E10 | Rural and Remote Health 2024;24(3):8316 |
| DOI | [10.1002/lary.31132](https://doi.org/10.1002/lary.31132) | [10.22605/RRH8316](https://doi.org/10.22605/RRH8316) |
| PubMed record | [pubmed.ncbi.nlm.nih.gov/37905776](https://pubmed.ncbi.nlm.nih.gov/37905776/) | [pubmed.ncbi.nlm.nih.gov/39075776](https://pubmed.ncbi.nlm.nih.gov/39075776/) |
| PMCID | None assigned | None assigned |

The two PMIDs - 37905776 and 39075776 - contain exactly the same eight digits; three of them have changed position. Digit transposition is one of the oldest transcription errors in bibliography, which is precisely why an identifier should be resolved, not trusted on sight.

## The same claim, through a deterministic check

Feeding the first run's claim - the rural-footprint title paired with PMID 37905776 - to the [citation verifier](https://scholar-sidekick.com/tools/citation-verifier) on 2026-08-15 returned:

- **Verdict: ambiguous** (confidence: high) - the identifier resolves to one paper while the claimed title matches a different one.
- **title mismatch** - resolved record reads “In reference to Orofacial Myofunctional Therapy for Obstructive Sleep Apnea: A Systematic Review and Meta-Analysis” (similarity 0.7).
- **container mismatch** - resolved record reads “The Laryngoscope” (similarity 0.48).
- **Full-score candidate for the claimed title** - PMID 39075776, DOI 10.22605/RRH8316, *Rural and remote health* (score 1).

The same input produces the same output tomorrow, because each side of the comparison is a registry lookup. That is the property the answer layer lacked.

## The second run corrected it

> PMID 37905776 does not have an assigned PMCID (PubMed Central ID) because the article has not been deposited into the PubMed Central open-access repository. … Title: Orofacial Myofunctional Therapy for Obstructive Sleep Apnea: A Systematic Review and Meta-Analysis. Journal: The Laryngoscope. Author: Eric J. Kezirian. Digital Object Identifier (DOI): 10.1002/lary.31132.

Same query, later the same day: the pairing is now correct, with the same Scholar Sidekick page cited for both the fact and the DOI. One edge stays blurred - the displayed title drops the letter's “In reference to” prefix, which leaves the letter's PMID under the exact title of the meta-analysis it comments on (a different paper, PMID 37606313). Whether that is truncation or conflation, a reader cannot tell from the answer alone.

## What this does and does not show

- **Two runs, two answers.** The same engine, the same query, the same day: one run paired the identifier with the wrong paper, the next run got it right. Neither run announced its own reliability - both carried the identical boilerplate disclaimer. The non-determinism is the finding: a reference list cannot afford a probabilistic answer to “which paper does this identifier name?”
- **One observation of the failure, not a rate.** We saw the wrong pairing once, and it did not reproduce. This page makes no claim about how often AI answer surfaces mispair identifiers - measuring that would need a sampled corpus of queries scored against registry records, which we have not run.
- **The correct answer was inside the cited source.** The first run cited a page that states the resolved title and DOI for PMID 37905776 - and still described a different paper. Retrieval found the right source; the wrong pairing happened at the synthesis step, where facts from different sources are joined into one answer.
- **Organic retrieval had it right.** Bing's plain result list for the same query put the Scholar Sidekick page first and Europe PMC's record of the correct paper second. The failure mode is specific to the layer that composes an answer, not to search retrieval as such.
- **Even the corrected run blurred an edge.** Run 2's “Title” drops the letter's “In reference to” prefix, which is exactly the title of the meta-analysis the letter comments on - a different paper (PMID 37606313). A small slip, but title-versus-record exactness is the entire game in citation integrity.
- **We benefit from telling this story.** The page the Overview cited is ours, and a post about being cited is good for us - that conflict of interest is why every claim here resolves to a public PubMed or Crossref record rather than to our say-so.

## Frequently asked questions

### Which paper does PMID 37905776 actually identify?

A letter by Eric J. Kezirian in The Laryngoscope (2024;134(3):E10), titled “In reference to Orofacial Myofunctional Therapy for Obstructive Sleep Apnea: A Systematic Review and Meta-Analysis”, DOI [10.1002/lary.31132](https://doi.org/10.1002/lary.31132). The PubMed record is at [pubmed.ncbi.nlm.nih.gov/37905776](https://pubmed.ncbi.nlm.nih.gov/37905776/). The rural-health study the first AI answer described lives at PMID 39075776.

### Did the AI Overview get the PMCID fact wrong?

No. In both runs the highlighted fact - PMID 37905776 has no PMCID because the paper is not deposited in PubMed Central - was correct, and correctly attributed to the page that states it. As it happens the fact is also true of the wrongly described paper: neither PMID has a PMCID. The error was the pairing of identifier and paper, not the fact.

### Is this the fabrication pattern from Topaz et al.?

No. The Topaz pattern is a real, resolvable identifier stitched to an invented title. Here both papers are real; a real identifier was paired with the wrong real record. That is the “wrong identifier for a real paper” class - the quieter, more common error - and it is why the verifier returns [ambiguous](/citation-integrity/verified-citations-can-still-be-wrong) for it rather than a flat real/fake verdict.

### Will I see the wrong answer if I search the same query now?

Probably not - a second run of the identical query on the same day already showed the correct pairing. AI answer surfaces are session-bound and non-deterministic, which is why this page preserves transcriptions and dated captures rather than a link to the answer. The registry records it checks against are the durable part.

### Could a human have made the same mistake?

Easily. 37905776 and 39075776 are the same eight digits in a different order, and digit transposition is one of the oldest transcription errors there is. That is the practical lesson: whether a citation was assembled by a person, a reference manager, or a language model, the identifier you publish should be resolved against its registry and compared with what you claim it names.

### How do I check an identifier-paper pairing like this myself?

Resolve the identifier at its own registry (for a PMID, its PubMed record) and compare the resolved title with the claimed one. Or let the [citation verifier](/tools/citation-verifier) do both sides: given the first run's exact claim, it resolves PMID 37905776 to the Laryngoscope letter, flags the title and journal disagreement, and surfaces the paper the claimed title really belongs to - PMID 39075776 - as a full-score candidate. Verdict: ambiguous, high confidence.

## Related

- [Citation integrity in the age of AI (the parent explainer)](https://scholar-sidekick.com/citation-integrity)
- [A verified citation can still be wrong (the error-class explainer)](https://scholar-sidekick.com/citation-integrity/verified-citations-can-still-be-wrong)
- [What four AI search engines admit about recommending citation tools](https://scholar-sidekick.com/citation-integrity/ai-evaluator-bias)
- [Citation Verifier (run the check yourself)](https://scholar-sidekick.com/tools/citation-verifier)
- [PMID / PMCID / DOI Converter (the tool family the cited page belongs to)](https://scholar-sidekick.com/tools/pubmed-id-converter)
- [PubMed record for PMID 37905776 (the searched identifier)](https://pubmed.ncbi.nlm.nih.gov/37905776/)
- [PubMed record for PMID 39075776 (the paper the first run described)](https://pubmed.ncbi.nlm.nih.gov/39075776/)

## Sitemap

See the full [sitemap](https://scholar-sidekick.com/sitemap.md) for all pages.