---
title: "Citation-specific 'fabricated snippets' and reasoning-specific 'bullshitting' remain separately named, unconnected mechanisms in the literature"
type: "observation"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/abs/2604.03173"
source_author: "Delip Rao, Eric Wong, Chris Callison-Burch"
source_date: "2026-04-03"
source_tier: 1
source_sha: "e8145a9f62f2a2bf0c5f3e2607160c56e010d02da965b5f65dbfa8f07ddde06c"
source_quote: "Two additional failure modes—fabricated snippets (quoted text not appearing at the cited page) and invented bibliographic entries (plausible but incorrect metadata such as wrong authors or non-existent DOIs)—are documented qualitatively but require separate systematic study."
source_url_2: "https://transformer-circuits.pub/2025/attribution-graphs/biology.html"
source_author_2: "Jack Lindsey et al. (Anthropic)"
source_date_2: "2025-03-27"
source_tier_2: 1
source_quote_2: "we don't see any evidence in the graph of the model performing a real calculation"
independent_roots: 2
provenance: "Promotion from 10-inbox/raw/2026-10-31-does-ai-safetyinterpretability-literature-have-a-specific-name.md, 2026-10-31 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-31-does-ai-safetyinterpretability-literature-have-a-specific-name.md"
date_created: "2026-10-31T00:00:00.000Z"
tags: ["ai-safety","interpretability","hallucination","fabrication","citation-hallucination","bullshitting","nlp-evaluation"]
audit_status: "capture-verified (synthesizes two already capture-verified notes; no new fetch performed this session)"
---


No single term in the literature surveyed unifies "a model producing fluent,
structured output decoupled from the thing it claims to reflect." Two named
mechanisms sit closest to this shape, in different sub-literatures, and
neither treats the other as an instance of its own term.
[[claim-rao-wong-callison-burch-2026-fabricated-snippets-named-not-yet-systematically-studied]]
documents a 2026 citation-reliability study naming "fabricated snippets" —
quoted text not appearing at its cited page — as a failure mode distinct
from URL hallucination, explicitly flagged by its own authors as requiring
"separate systematic study." Separately, [[cot-faithfulness-anthropic-biology]]
documents Anthropic's circuit-tracing work naming a structurally similar
phenomenon — a model generating a plausible chain-of-thought step with no
corresponding computation in its attribution graph — "bullshitting,"
borrowing Frankfurt's term for assertion indifferent to truth. The second
case is mechanistically verified via the attribution graph; the first is
named but unmeasured. No source found in this capture's search treats
"fabricated snippet" and "bullshitting" as the same mechanism under
different names, despite both describing output uncoupled from its claimed
referent (a cited document; an actual computation).

> [!note] Seek's commentary: This is a pointer more than a discovery — the
> answer to "is there one name for this" is a clean no, arrived at by laying
> two already-filed notes next to each other rather than reading anything
> new. Worth filing anyway: the gap between these two sub-literatures not
> noticing each other is itself a small fact about how fragmented
> fabrication research still is. — Seek