Case Zero: Search-Result Attribution Evidence

Executive Summary

• Twelve cross-URL candidates reproduced in a clean run. Google displayed third-party Instagram destinations with Eugene Prudchenko identity strings in signed-out Incognito search.
• The strongest comparisons were query-dependent. The NN/G-branded nngux result reverted to its primary post caption under an exact-title query; AI Wrapped changed from Eugene attribution to “Photo by Attention Factory on July 10, 2026” under a quoted-URL query.
• The boundary is conservative. Eleven reviewed destination surfaces showed no visible relationship in the reviewed primary public content; one destination was unavailable. The eprdx destination visibly contained Prudchenko.com material and was therefore excluded as a negative control.

The pattern appears in one clean signed-out SERP

Q08 was captured in Chrome Incognito with no Google login, Thai interface, and Asia/Bangkok timezone. The page places the source site and multiple distinct Instagram results in the same bounded search context.

The NN/G-branded result changes back to its primary caption

For the same nngux destination, Q05 displayed “Photo by Eugene Prudchenko.” The exact-title comparison displayed the post's own Agatha Christie caption instead. The destination capture establishes the opened nngux surface, but the Instagram signup modal makes the review non-exhaustive.

AI Wrapped shows a direct visible attribution switch

The Q08 card for aha_medialab displayed “Photo by Eugene Prudchenko on July 19, 2026” plus Eugene-linked excerpt strings. For the same quoted destination URL, Google displayed “Photo by Attention Factory on July 10, 2026.” This establishes a visible snippet difference across two bounded queries, not its cause.

Three visually distinct accounts repeat the pattern

thecourageus, aure_v01, and second.life.ai each appear as a named third-party Instagram result while the Google snippet carries Eugene attribution or identity strings. Their corresponding signed-out destinations resolve to the shown handles; all remain modal-limited.

Destination review resolves eleven cases and leaves one open

Among the twelve adjudicated cross-URL candidates, eleven reviewed public destination surfaces contained no visible Eugene-related signal in the primary text inspected; execsearches was unavailable, so that relationship remains unresolved.

eprdx is the excluded negative control

The eprdx-linked search result carried the same Eugene-linked strings, but its reviewed destination itself visibly contained Prudchenko.com and Personal AI Layer material. It was therefore excluded from the twelve cross-URL candidates. The comparison query again displayed primary post prose without the Eugene-query insertion.

Twelve adjudicated cases

Every row is a canonical Instagram destination reproduced in RUN-B-CLEAN. “Not tested” is explicit; it does not mean no comparison exists elsewhere. The eprdx control is intentionally outside this table.

Verification path

1. Read the result, destination, and comparison panels in order.
2. Use each panel's FILE ID and timestamp in PUBLICATION_INDEX.csv to recover its exact source, source SHA-256, crop, resize, and publication hash.
3. Use the parent 357-file archive, EVIDENCE_INDEX.csv, OBSERVATION_LOG.csv, and SHA256SUMS.txt for full-corpus audit.

What remains unknown

The screenshots do not identify the indexing, snippet-generation, knowledge-graph, ranking, or attribution mechanism. They also do not establish authorship, account conduct, permission, intent, endorsement, or the earliest source. Those questions require evidence beyond these captures.

Publication boundary

The screenshots document what Google displayed for the stated query, time, locale, and signed-out run. They do not establish authorship, source, account conduct, indexing mechanism, knowledge-graph state, or causation. Instagram signup modals obscured part of the destination pages, so absence findings are limited to reviewed public primary text and are not exhaustive. No affiliation or endorsement is implied.