Skip to content

Where a piece came from, what it shows, and its useful parts ​

Answered on 4 October 2026. You chose “Separate browsing paths (Recommended)”. Collection-source products and products shown or discussed remain independently browsable, with evidence for each. Your earlier answers already settle photographs, useful document parts and strategy claims. The original recommendation and alternative below remain as the review record.

1. Where we collected something is different from what it is about ​

Approved: let people search the two relationships separately. A capture may come from an app or website. The piece may show or discuss a product. Those facts can describe different products, or only one may be known. Neither creates a new content category or automatically establishes campaign membership.

The old review used an OpenAI Operator video as evidence for content collected from inside ChatGPT. The inspected library record instead points to an OpenAI TikTok post, with product associations recorded separately. That supports a distinction between collection source and subject; it does not prove the capture came from ChatGPT. This check concerns the recorded source, not a new analysis of the video.

As an illustrative counterexample, a screenshot collected inside ChatGPT could contain a comparison of two phones. ChatGPT would describe where it was collected; the phones would describe its subjects. We should preserve either fact only when we have evidence. A product mentioned in a caption is not proof that it supplied the capture, and the collection source is not proof of who made the creative work.

Existing guidance: the Groupings page already describes an “everything from this app” view, while Products describes links to products shown in a piece. The old review conflated those two paths; this review recorded your explicit browsing direction, not approval to treat a source as a subject.

What is already settled: product identity remains shared across its appearances. Types, campaigns, browsing headings and distribution stay separate. There is no new mandatory “inside a product” category to approve here.

The question you answered: should “collected from this product or app” and “shows or discusses this product” remain separately recorded and searchable? I recommend yes. The alternative is to preserve the collection source only in capture details and offer product browsing by subject alone. That is simpler to browse, but cannot directly answer “show everything we collected inside this app.”

This affects capture provenance, product links, filters and tracker rules. It does not decide which tables or fields to use. We must also distinguish a collection source from a tool used to make something; those are not interchangeable credits.

Whole pages are already library pieces — no answer needed ​

The recorded content rule explicitly says a landing page is a content item. Your later document-parts decision adds independently useful parts that stay linked to their source. We should therefore carry both forward: a complete web page can be browsed as a piece, with useful assets linked to it. A screenshot records a capture; it does not automatically become a new creative work.

The old product-page review proposed treating pages only as surfaces for smaller pieces. That premise conflicts with the recorded rule, so it is superseded rather than sent back to you.

Anthropic’s Claude on Mars page was read for this review. It tells a story about Claude and NASA’s rover planning. The selected library records include a complete-page capture and a hero capture. Their stored labels do not establish a correct type or a working parent-and-part relationship. Capturing a page with an embedded video also does not establish that the original video file was collected.

Since answered: Product Detail Page is an approved whole-piece type for a complete page presenting a particular product for evaluation or purchase. An article about a product does not automatically qualify. The retail review records the answer; storage and collection behavior remain to verify.

Answers that carry forward ​

No further answer is needed for these rules:

  • A photograph stays a photograph and links to the work it depicts. A physical installation, its documentation and an original visual asset can be related without becoming duplicate copies of one thing. Your DevDay example remains an acceptance case; a shared event alone is not proof of a direct depicts relationship.
  • Useful document parts remain connected to their document. Independent browsing does not lose their source context.
  • A source’s strategy claim stays separate from our analysis. Preserve who said it and the supporting evidence. A claimed strategy is not automatically a verified fact.

The decision record holds these approvals. Their implementation still needs verification; an implementation gap does not reopen the meaning.

What was checked, and what remains ​

The bounded review recovered the original source-product and product-page cards, the later document/photograph/strategy rulings, and the current product and grouping explanations. It inspected selected live schema and records, plus the local product-link writer and product-category query. The writer handles source and subject associations, while the inspected category filter does not distinguish their roles. This is a contract issue to trace across all consumers before any repair, not proof that the complete deployed path has been audited.

The technical evidence note records sources and limits separately. No schema, data, prompt or automation was changed. Your explicit reply approves the browsing direction; it does not certify implementation. The review plan keeps the remaining classification, identity, retail and definition work in order.

Prepared and answered 4 October 2026. No further answer is needed on source-product browsing.

Last reviewed 5 October 2026.

The BrandTrackers domain model. Source: git markdown, drift-checked against the live DB.