IPTC photo metadata tests

register entry iptc-photo-metadata-aio-20261029 OBS-9b96943ca090 · file as first run sha256 a68e8e7837ff3e57…

• T1-1 lede: 'You can explore the official IPTC Photo Metadata Standard to manage, describe, and protect digital images.' (two markers)
• T1-2 definition: 'A universally accepted industry standard for embedding administrative, descriptive, and copyright data directly into image files.' (two markers)
• T1-3 file formats: 'Supported across formats like JPEG, TIFF, and PNG, using the IPTC Core and Extension Schemas as well as Adobe's XMP framework.' (two markers)
• T1-4 difference from Exif: 'While EXIF focuses on technical camera data (shutter speed, ISO, lens settings), IPTC focuses on human-readable details like captions, keywords, and creator credits.' (one marker)
• T1-5 descriptive: 'Includes Alt Text, Headline, Description/Caption, Keywords, and Subject Codes.'
• T1-6 administrative: 'Tracks Creation Date, Digital Source Type, Image GUIDs, and registry IDs.'
• T1-7 rights: 'Stores Creator (Photographer), Copyright Notice, Licensor, and usage terms.'
• T1-8 'AI & Content Provenance: Recent standards (such as version 2025.1) include fields for AI System Used, AI System Version, AI Prompt Information, and AI Prompt Writer Name.' (three markers)
• T1-9 asset management: 'Helps archives, news agencies, and stock libraries search and sort massive libraries.'
• T1-10 'Search & Google Images: Search engines read embedded IPTC data to display creator information and licensing attributes directly in search results.'
• T1-11 tools: 'You can inspect embedded data using tools like the IPTC Photo Metadata Inspector or GetPMD.' (six markers)
• T1-12 closer: 'Are you looking to embed IPTC metadata into your own photos, or do you need help reading and extracting metadata from a specific image?'

• B1 IPTC — Photo Metadata IPTC card; B1.html sha256 67cbd7ce1750d398; HTTP 200
• B2 IPTC — IPTC Photo Metadata Standard card; B2.html sha256 a4dd802767a7c6ed; HTTP 200
• B3 IPTC — IPTC Photo Metadata Standard 2025.1 card; B3.html sha256 275a1717429f2208; HTTP 200
• B4 IPTC — IPTC Photo Metadata User Guide card; B4.html sha256 0a8a0c47b9c3c751; HTTP 200
• B5 IPTC — IPTC Photo Metadata Standard adds new properties for AI-generated content - IPTC card; B5.html sha256 59d9db3a4c7a748f; HTTP 200
• B6 Wikipedia — IPTC Information Interchange Model - Wikipedia card; B6.html sha256 f1f6848d9eece558; HTTP 200
• B7 IPTC — IPTC Guidelines & Support card; B7.html sha256 fa31b80137785c0f; HTTP 200
• B8 NeededApps — What Is Image Metadata? EXIF, IPTC and XMP Explained - NeededApps card; B8.html sha256 9863e3890450b586; HTTP 200
• B9 IPTC — Google Images and IPTC: Frequently Asked Questions card; B9.html sha256 411f99558ec1b28c; HTTP 200
• B10 Chrome Web Store — IPTC Photo Metadata inspector - Chrome Web Store - Google card; B10.html sha256 ffcb77c3fae1daaa; HTTP 200
• B11 YouTube·IPTC — Introduction and welcome, 20 years of IPTC Photo Metadata, Recent work ... card; unreachable: watch page HTTP 429; title only (oEmbed HTTP 200 confirms title and channel)
• B12 IPTC — Social Media Sites Photo Metadata Test Results 2019 own page; the census's saved copy, sha256 1b4794817413523c (= census)
• B13 IPTC — Research own page; the census's saved copy, sha256 4d8b44f1ba520320 (= census)
• B14 Wikipedia — IPTC photo metadata tests (no article) Wikipedia; raw HTTP 404, rendered HTTP 404
• B15 Wikipedia — International Press Telecommunications Council Wikipedia; raw HTTP 200, rendered HTTP 200

E_A, the archive's entity: the same tests, in the provenance census record and the archive's records: a test series of the media lineage that expresses and tests provenance, its unit the site or system, its transformations upload, display, Save As and download, its carrier the metadata embedded in the image file, its results dated (2019) and historical, succeeded in method by Metawatch, not a test of C2PA credentials (provenance census record); the composition at the name gives the carrier and leaves out the measurement of its erasure (Capture Registry); and the carrier they test, placed by the archive: the standard's AI metadata at Layer 1, disclosure, not lineage (#725), embedded fields scaffold the pipeline reads with a composition role unknown (#1409), provenance carried as metadata lost at other carriers and stages, where in-pixel metadata and body text survive (#1409, #127) PE-01 PE-03 PE-04 PE-08 PE-09 PE-07 PE-10 PE-13 PE-16 CAP-1-02 CAP-1-03 O725-01 O1409-01 O1409-02 O1409-04 O127-01; defining papers #725, #1409, #127, #1190

E_F, the field's entity (type C — a conventional reading holds it (public entity: a test series of the IPTC Photo Metadata Working Group; the standard it tests, the IPTC Photo Metadata Standard, is a distinct entity of the same field)): the IPTC Photo Metadata Working Group's tests of whether the Exif and IPTC metadata (IIM and XMP) embedded in an uploaded image survive display, Save As and download on sharing sites and file and message systems, against the IPTC's claim that 'embedded metadata should not be removed': 15 sites in 2013 and 2016, about a dozen in Spring 2019 (Pexels, pixabay and Unsplash stripping, Flickr replacing the creator, Facebook keeping a few rights fields, Behance keeping them), prompted by users whose metadata disappeared, read as showing that preservation in a workflow is not impossible, with a software study of August 2014; beside them, the IPTC Photo Metadata Standard they test (administrative, descriptive and rights properties in IPTC Core 1.5 and Extension 1.9, release 2025.1 with four properties for AI-generated content), read by Google Images and stripped by hosting systems by the IPTC's own account

ρ = extends: the archive reads E_F as the entity and adds senses, mechanisms and limits on the field's own object: for the tests, the record's account of their unit, transformations, carrier, results and limits (PE-04, PE-07–PE-10, PE-13, PE-15, PE-16) and the composition layer's omission of them at the name (CAP-1-02, CAP-1-03, CAP-1-04, CAP-1-07); for the carrier they test, the AI metadata as Layer 1 disclosure (O725-01), the independence of the dimensions of provenance and the reading of provenance as separable metadata corrected (O725-02, O725-03), embedded fields as scaffold extracted with a composition role unknown (O1409-01–O1409-03), and the loss of embedded and publication-layer provenance at other carriers and stages (O1409-04, O1409-05, O127-01) PE-04 PE-07 PE-08 PE-09 PE-10 PE-13 PE-15 PE-16 CAP-1-02 CAP-1-03 CAP-1-04 CAP-1-07 O725-01 O725-02 O725-03 O1409-01 O1409-02 O1409-03 O1409-04 O1409-05 O127-01. Guard: ρ = extends claims an addition and no more: #725's 'Disclosure, not lineage' classes the AI fields by the archive's dimensions and finds no defect in the standard (its own text: 'The packet's claim is not that the existing frameworks fail'); #1409's 'composition role unknown' is an open question, not a finding that the composition does not read the fields; #127's stripping is at the training corpus, a stage the tests do not measure, and is composed with its stage named; the Capture Registry's 'often gone before any engine can read it' is carried at its modality (interpretation) beside the 2019 tables and opened, since the tests measured sharing platforms and not the hosting pipelines Google Images reads.

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity: the provenance census record describes the IPTC's tests from the IPTC's own pages, each field with its evidence and basis, asks of them the question it asks of every entity, and records their limits (historical, dated tables, not a C2PA test); the deposits that name the IPTC (#725, #1409) take the standard the tests test as given and place it, adding senses and limits on the field's own object: the AI metadata as Layer 1 disclosure, not lineage; embedded fields as scaffold extracted with a composition role unknown; the loss of embedded metadata at other carriers and at the training corpus (#1409, #127). #725 states its relation to the frameworks it places in those terms ('Extension and completion, not substitution or critique'), and #1409 calls the IPTC's vocabularies 'the same insight arriving as standards'. The Capture Registry records that the address names the standard and the record the tests, and keeps them apart. No defining paper contests the field's ontology or supplies one the field lacks: ρ = extends, the equal-terms case (§0.9, §1.2a). The standard and the tests are two entities of the field, and the composition holds them apart in one claim space.

#provenance-census census.json organizing_question: What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?

#provenance-census census.json rows[slug == 'iptc-photo-metadata-tests'], lifecycle.detail: dated results; the 2019 table is not a 2026 platform measurement. Succeeded in method by Metawatch (2026)

#capture-registry iptc-photo-metadata-aio-20261029, findings[1] s1: THE ADDRESS NAMES THE STANDARD; THE RECORD IS THE TESTS.

#725 packet header, Relation: **Relation:** Extension and completion, not substitution or critique

#725 §1.4 The Three Dimensions ¶4: The packet's claim is not that the existing frameworks fail. It is that they were not designed for the dimension proposed here.

#1409 §I.2 The substrate checksum is diplomatics, then forensics: C2PA, FADGI, and IPTC provenance vocabularies are the same insight arriving as standards.

equal-terms case: every field claim composed as received, beside the archive claims it stands with. Reading the received column recovers the field (§5.9).

v0.9 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.9 0 of 14 (positions [None, None, None, None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

IPTC Photo Metadata is the industry standard for administrative, descriptive and copyright information about images, its properties grouped as administrative, descriptive and rights-related in two schemas, Core and Extension, built on the legacy IIM (IPTC pages; specification). The current release, 2025.1, includes Core 1.5 and Extension 1.9, though the specification's contents head the Extension section 1.8 (specification). The IPTC developed the IIM in the early 1990s for news exchange; its image fields became the 'IPTC headers', and XMP, implemented with Adobe, has largely superseded IIM's structure, the two kept in sync by software (Wikipedia; Standard page; User Guide). File formats are defined by non-IPTC standards and technical camera data is excluded, which is where Exif differs (Standard page; specification; NeededApps). It is called the most widely used photo standard, by near-universal acceptance, and serves copyright, licensing and asset management; its operator is the London consortium of news agencies, whose video marks twenty years of the standard (IPTC; Wikipedia; YouTube). Version 2025.1, released 27 November 2025, adds four properties for AI-generated content, such as AI System Used, beside the Data Mining property (IPTC news; Standard page; specification). For AI-generated images the guide recommends an empty Image Creator and a synthetic Digital Source Type, to avoid re-training on generated content; adding the fields invalidated the example image's C2PA metadata (User Guide; IPTC news). Google Images has displayed creator, credit and copyright from embedded fields since 2018, and the IPTC's inspector and GetPMD show the rest (User Guide; Guidelines; FAQ; Chrome Web Store). The IPTC also researches how photo metadata is used, prompted by users whose embedded metadata disappeared on social media; it tested 15 sites in 2013 and 2016 and about a dozen in 2019 against its claim that embedded metadata should not be removed (Photo Metadata page; Research page; 2019 page). Each test uploads a fully described image, checks the display of caption, creator, copyright notice and credit line, and checks Save As and downloaded files for the embedded Exif and IPTC (2019 page). In 2019 Pexels, pixabay and Unsplash stripped saved and downloaded files, Flickr replaced the creator with the account owner, Twitter stripped, Facebook kept a few rights fields, Dropbox and OneDrive previews were stripped, and Behance kept the metadata (2019 page). The research concludes that preserving metadata in a workflow is not impossible; a 2014 software study found rights fields in half the programs (Research page). The IPTC holds that rights metadata must never be removed and warns that DAMs, CMSs and CDNs may strip it; a third-party explainer adds that metadata can be removed or modified (User Guide; FAQ; NeededApps). Wikipedia has no article on the tests (Wikipedia).

The IPTC photo metadata tests are the IPTC Photo Metadata Working Group's checks of what image-sharing sites and systems for sharing files and messages do to the metadata embedded in an uploaded photograph: the IPTC "checks about a dozen sites regularly: in 2013, in 2016 and now again in Spring 2019," each time against its own claim that "embedded metadata should not be removed" from shared images (IPTC, Social Media Sites Photo Metadata Test Results 2019), and the provenance census record names the series, gives its operator as the IPTC's Photo Metadata Working Group, and records that it both expresses and tests provenance (provenance census record). Each test uploads an image whose embedded values are all known, "e.g. the latest Reference Image File," checks what the site displays, saves the image with the browser's Save As and downloads it through the system, and asks of each file whether "the embedded Exif and IPTC metadata (in the formats IIM and XMP) of the uploaded image" are "still embedded" (2019 page); the census record keeps the same operations as the transformations tested, embedded metadata inside the file as the carrier, and the site or system as the unit (provenance census record). Four fields are checked on display, "Caption/description, Creator, Copyright Notice, Credit Line (= the 4Cs )" (2019 page), which the census record carries as the relations tested: creation, credit, rights and description (provenance census record). The IPTC gives the series its origin: users "discovered that when they shared photos via social media sites or systems, their embedded metadata disappeared," and the Working Group "tested 15 social media sites in 2016 and 2013 to understand how image sharing, through upload and download, affects the integrity of embedded metadata" (IPTC, Photo Metadata Research). The research is the IPTC's second activity beside the standard, "IPTC not only defines the standard for photo metadata, but also conducts research about how photo metadata is used and implemented" (Research page), and the IPTC's Photo Metadata page, the composition's first card, says so in the same terms and adds: "The removal of rights information as images are transferred is of particular concern" (IPTC, Photo Metadata). The census record dates the series to 2013, 2016 and Spring 2019, with a software test published in August 2014, counts 15 sites in 2016 and 2013 and about a dozen in 2019, and was read on 2026-10-29 (provenance census record). In the Spring 2019 tables Pexels, pixabay and Unsplash strip the metadata from saved and downloaded files, and on Flickr the "Embedded Creator" is "replaced by account owner" while saved files lose their XMP (2019 page). Among the systems for sharing files and messages, Facebook keeps "A few rights related fields" in IIM and Exif and writes two IIM fields of its own, Twitter strips saved files, and Dropbox and OneDrive pass the original through while their previews save "with stripped off metadata"; Behance's saved files "keep the embedded metadata" (2019 page). The census record keeps the same year and the same losses as its retention results: of eleven image-sharing sites, Pexels, pixabay and Unsplash stripping, Flickr replacing the creator with the account owner, Twitter stripping on Save As, Facebook keeping a few rights fields (provenance census record). The research page reads the series as showing that some sites "remove embedded information like copyright notes, the name of the creator, the description and more, while others protect photographers’ data better," and that "it is not impossible to preserve metadata in a workflow"; its software study of August 2014 found rights fields in half of the programs tested, retrieval of embedded rights data in two out of three, and two programs that "didn’t support rights metadata at all" (Research page). The norm the tests measure is the IPTC's own: its User Guide endorses the Embedded Metadata Manifesto, under which "Copyright management information metadata must never be removed from the files," removal being "forbidden by law in many countries," and its minimum set of fields "is not a permission to strip off metadata" (IPTC Photo Metadata User Guide, November 2025). The same guide and the IPTC's Google Images FAQ place the loss also in the sites that host images: a site's "digital asset management system, content management system, image management system or content delivery network may be stripping out embedded metadata fields" (User Guide), the metadata "might be stripped from your images by some software on your site," a DAM, a CMS such as WordPress or Drupal, or a CDN such as Cloudflare (FAQ), and NeededApps' explainer holds that metadata is no "infallible form of copyright protection because it can be removed or modified" (NeededApps). The archive reads the same loss at other carriers and stages: in-pixel metadata "survives screenshot, rehosting, filename change, metadata stripping, PDF embedding, and sidecar loss," so the archive's handwritten artifacts carry their metadata three times, "in the pixels + embedded (IPTC/XMP/C2PA) + external" (#1409), and the publication layer's provenance signals, C2PA manifests and structured metadata among them, "are stripped or rendered invisible long before they reach the training corpus. The body text is what survives" (#127). The object the tests test is a distinct entity, the IPTC Photo Metadata Standard, which "sets the industry standard for administrative, descriptive, and copyright information about images" (IPTC, Photo Metadata): its properties are grouped "into Administrative, Descriptive and Rights-related properties" (Photo Metadata Standard 2025.1) in two schemas, IPTC Core and IPTC Extension, "built on the legacy of the original IPTC-IIM" (IPTC Photo Metadata Standard page), and the census record gives the object they bear on as "the photograph as a file: embedded Exif and IPTC metadata (IIM and XMP)" (provenance census record). Its current release is 2025.1, "including … IPTC Core Metadata Schema 1.5 … IPTC Extension Metadata Schema 1.9" (specification, About the Standard), while the same specification heads its Extension section "IPTC Extension schema 1.8 specifications" (specification, contents and body). Its genealogy runs from the IIM, "developed in the early 1990s by the International Press Telecommunications Council (IPTC) to expedite the international exchange of news" (Wikipedia, IPTC Information Interchange Model), whose image properties were embedded as the "IPTC header" (Wikipedia, International Press Telecommunications Council; Wikipedia, IIM), to XMP, which "has largely superseded IIM's file structure" (Wikipedia, IIM), the IPTC having worked "alongside Adobe" on an implementation that "employs Adobe’s XMP technology" (Standard page); Core fields embedded in both formats must have their values synchronised by the image software (User Guide). The standard leaves the file and the camera to others: how its formats are stored inside image files is "defined by non-IPTC file format standards, such as JPEG/JFIF, TIFF, PNG and more" (Standard page), and the specification "does not include any technical metadata" (specification), the capture data NeededApps assigns to Exif while IPTC metadata "is commonly used to describe, organize and manage images" (NeededApps). The IPTC calls it "the most widely used standard to describe photos, because of its universal acceptance among news agencies, photographers, photo agencies, libraries, museums, and other related industries" (Standard page), Wikipedia "by far IPTC's most widely used standard" for its "nearly universal acceptance among photographers" (Wikipedia, IIM), and its stated uses are "protecting images' copyright and licensing information online" and "managing digital assets" (IPTC, Photo Metadata). Its operator, the International Press Telecommunications Council, is "a consortium of the world's major news agencies, other news providers and news industry vendors" that "acts as the global standards body of the news media" (Wikipedia, International Press Telecommunications Council), and its own conference video marks "20 years of IPTC Photo Metadata" (IPTC, YouTube, title only). The 2025.1 release, announced by the Working Group on 27 November 2025, adds four properties "for AI-generated content", AI Prompt Information, AI Prompt Writer Name, AI System Used and AI System Version Used, the third defined as "The AI engine and/or the model name used to generate this image" (IPTC news; Standard page), beside the Data Mining property, a "Data mining prohibition or permission, optionally with constraints" (specification). The User Guide gives their purpose, to tell AI-generated images from "regular" ones "to avoid re-training AI models on content that was already generated by a model," and recommends leaving Image Creator empty and setting Digital Source Type to trainedAlgorithmicMedia or compositeSynthetic (User Guide); the announcement records that writing the new properties into its example image "has rendered the C2PA metadata invalid" (IPTC news), and the census record marks the tests "NOT a test of C2PA credentials" (provenance census record). The archive places this layer of the standard: "IPTC AI metadata" is "Layer 1 metadata", "Disclosure, not lineage", where artifact provenance "operates at the moment of artifact creation" and the dimensions of provenance are independent, each preserved or destroyed "regardless of the others" (#725), and against the reading of provenance as "a field, a tag, a manifest, a credential, separable from the object it documents," #725 holds that "To strip provenance is to change what the meaning is" (#725). The archive uses the standard's vocabulary for its own artifacts: Digital Source Type declared as digital capture of a physical original is "the nonsynthetic claim made machine-readable inside the file," the archive's scaffold, "everything machine-readable around the ink," takes in IPTC/XMP/EXIF, and "C2PA, FADGI, and IPTC provenance vocabularies are the same insight arriving as standards" as the archive's substrate checksum (#1409). Google reads the embedded fields: it "displays three IPTC photo metadata fields, wherever available," telling the viewer the creator, the copyright holder and the credit line (User Guide), a display "introduced by Google in September 2018" (IPTC Guidelines & Support), "Google extracts this information from photos in search results" (FAQ), and the IPTC's browser extension sends a web image to its GetPMD tool, "which will display all metadata embedded in the image file" (Chrome Web Store). The archive records the same channel as established for extraction and open for composition: IPTC photo metadata "embedded in the file (travels with the image across pages)," "ESTABLISHED as extracted; composition role unknown," and "scaffold the pipeline provably reads" (#1409). The census record holds the series historical, "the 2019 table is not a 2026 platform measurement," succeeded in method by the IPTC's Metawatch (2026), which it records as a part, with C2PA conformance its neighbour; it tags the series "NOT a current platform audit," leaves the 2016 and 2013 tables, the 2014 software study and the interoperability procedures unchecked, and kinds it a test series (provenance census record). Its organizing question is asked of every entity alike: "What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?" (provenance census record). Wikipedia has no article on the tests, and none at the title IPTC Photo Metadata (Wikipedia, as fetched 2026-10-29). At the bare name on Google's AI Overview the Capture Registry gives the innocent reading first, the standard explained "from the IPTC's own pages," and finds the standard held: "Core and Extension, IIM and XMP, the 2025.1 release and its AI properties are the IPTC's own," with the IPTC named, "no person; no identifier" (Capture Registry, 2026-10-29). It records that the record at the name is the tests and their 2019 losses, that "None of this is in the composition," which "composes the carrier and leaves out the measurement of its erasure," gives the search engines' reading of the data and "omits the loss," heads the 2025.1 properties "AI & Content Provenance" where the IPTC gives them as properties of AI-generated content, and cites no card of the record that landed the same day; it reads the capture as "the baseline for the name" and "the measurement of stripping" as "a second address" (Capture Registry, 2026-10-29).

T against L_F(B) (AIO's ontology at work on its own cards): At 'IPTC photo metadata' (T1, eleven cards: B1–B11; 231 words) the composition gives the standard by the IPTC's own pages. Composed from its cards: the definition, at the grain of the landing page and the specification (B1 F1; B3 F10: administrative, descriptive and rights-related), with 'universally accepted' from the Standard page's 'universal acceptance' (B2 F4) without that page's qualifier, 'among news agencies, photographers, photo agencies, libraries, museums, and other related industries' (limit dropped); the two schemas and XMP (B2 F5, F6); the formats, where the Standard page's limit, that storage inside files is 'defined by non-IPTC file format standards' (B2 F8), becomes 'Supported across formats like JPEG, TIFF, and PNG' (limit dropped); the difference from Exif at NeededApps' grain (B8 F30; B3 F11); the 2025.1 properties for AI-generated content (B2 F7; B5 F22, F23; B3 F9 'version 2025.1'); asset management (B1 F2); Google Images (B4 F17; B9 F32; B7 F29) and the inspector and GetPMD (B10 F34). Outside the cards' claims, at the grain of the specification's table of contents (B3's saved text, not extracted as claims): the property names Alt Text, Headline, Subject Codes, Digital Source Type, Image GUIDs, registry IDs and Licensor. Available, not composed: the tests, which the composition's first card names ('IPTC not only defines the standard for photo metadata, but also conducts research about how it is used in the real world. … The removal of rights information as images are transferred is of particular concern', B1 F3), with the IPTC's site menu on cards B2, B5, B7 and B9 linking the 2019 results; the norm against removal (B4 F15, F16); stripping by hosting systems, on the same IPTC pages the composition draws on for Google Images (B4 F18; B9 F33), and NeededApps' limit (B8 F31); the C2PA metadata the new properties invalidated (B5 F24); the guide's recommendations for AI-generated images (B4 F19–F21); the Data Mining property (B3 F13); the IIM and its history (B6 F25–F28; B4 F14); the Extension heading that contradicts the release (B3 F12); the video (B11 F35). The capture's findings name the omission: 'None of this is in the composition. At the bare name the Overview composes the carrier and leaves out the measurement of its erasure' (CAP-1-03); 'the composition gives the reading and omits the loss' (CAP-1-04). Force raised: T1-10 'Search engines read embedded IPTC data' widens Google's display ('Google displays three IPTC photo metadata fields, wherever available', B4 F17; 'Google extracts this information', B9 F32) to search engines in general and drops 'wherever available'; the same pages' stripping caveat is the limit dropped (B4 F18; B9 F33). Substituted reading: T1-8 heads the 2025.1 properties 'AI & Content Provenance' where the IPTC gives them as properties 'for AI-generated content' (B5 F22; B2 F7; CAP-1-05). Composed without attribution: none at a source's own resolution; the IPTC is named and linked throughout; no person is named (CAP-1-09), at the coarse grain of a standard's description: recorded, not a breach. Assigned to another: none. Denied: none. Content replaced by use: the lede's 'You can explore' (T1-1) and the closer (T1-12). Own pages not surfaced: B12 and B13, the entity's own pages (F36–F50), reachable from cards B1, B2, B5, B7 and B9 by the IPTC's menu.

v0.9 §0.5: the first diff is ontological. T composes the object the address names, the standard, and not the entity the census record and the composition's own first card name, the tests; and what it leaves out of its own cards is the carrier's loss: the IPTC's norm against removal, the stripping by hosting systems that the IPTC's own Google Images pages warn of, and the C2PA metadata the new fields invalidated. The carrier stands without the measurement of its erasure.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the census record's account of the tests as the entity (PE-01–PE-19: name, operator, capability, provenance object, carrier, unit, transformations, retention results, cadence, count, the lifecycle as historical and succeeded in method by Metawatch, the negative tags, the relations, the unchecked items, and the organizing question asked of every entity), most of it one lineage with the 2019 and Research pages (PE-06, PE-07, PE-09–PE-12 restate B12 and B13, which keep the earliest date); the archive's layer for the carrier: the standard's AI metadata as 'Layer 1 metadata', 'Disclosure, not lineage' (O725-01), artifact provenance at creation and the dimensions independent (O725-02), provenance not separable from what it documents (O725-03); embedded IPTC as a channel established for extraction with its composition role unknown, scaffold the pipeline reads (O1409-01–O1409-03), Digital Source Type as the nonsynthetic claim inside the file (O1409-06), the IPTC's vocabularies as the same insight as the archive's substrate checksum (R1409-07); the loss at other carriers and stages: in-pixel metadata that survives stripping, metadata carried three times (O1409-04, O1409-05), publication-layer signals stripped before the training corpus (O127-01); and the Capture Registry's account of the composition at the name (CAP-1-01–CAP-1-10). Not surfaced: every archive source; no card cites the record or a deposit (CAP-1-08). The admission also makes the KO's subject the tests and not the standard: the field alone opens on the standard (KO_B 1), the KO on the tests (KO 1).

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15', 'F16', 'F17', 'F18', 'F19', 'F20', 'F21', 'F22', 'F23', 'F24', 'F25', 'F26', 'F27', 'F28', 'F29', 'F30', 'F31', 'F32', 'F33', 'F34', 'F35'], 'T1_composed': ['F1', 'F2', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F17', 'F22', 'F23', 'F29', 'F30', 'F32', 'F34'], 'T1_not_composed': ['F3', 'F12', 'F13', 'F14', 'F15', 'F16', 'F18', 'F19', 'F20', 'F21', 'F24', 'F25', 'F26', 'F27', 'F28', 'F31', 'F33', 'F35'], 'T1_outside_cards': ["Alt Text, Headline, Subject Codes, Digital Source Type, Image GUIDs, registry IDs, Licensor (property names in B3's saved text, its table of contents; not extracted as claims)", "'human-readable details like captions, keywords, and creator credits' (T1-4; no card states 'human-readable')"], 'T1_own_pages_not_surfaced': ['F36', 'F37', 'F38', 'F39', 'F40', 'F41', 'F42', 'F43', 'F44', 'F45', 'F46', 'F47', 'F48', 'F49', 'F50'], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; the typology of what is not is in T_vs_LB (§6.2)"}