NewsGuard AI Tracking Center

register entry newsguard-ai-tracking-center-aio-20261029 OBS-f7965ff6f906 · file as first run sha256 4efaa7d3d28fc9ff…

• T1 definition: 'The NewsGuard AI Tracking Center is an ongoing investigative project by NewsGuard that monitors and catalogs artificial intelligence-generated misinformation, unreliable AI news sites, and AI tool vulnerabilities.' (markers [1], [2])
• T2 content farms: 'AI Content Farms: Identified 3,749+ unedited AI content farm sites operating across 16 languages that mimic human news layouts and monetize via programmatic advertising.' (marker [1])
• T3 audits: 'Chatbot & Tool Audits: Found that leading generative AI tools repeat known news false claims in 28% to over 33% of prompts, fail to recognize AI-generated video (such as Sora) in 78% to 95% of prompts, and occasionally propagate state-sponsored disinformation.' (marker [1])
• T4 blocking: 'Publisher Blockades: Approximately 67% of top global news sites actively block AI crawlers and chatbots, pushing AI systems toward lower-quality data sources.' (marker [1])
• T5 the series as a tool: 'AI False Claims Monitor: Launched to measure AI reliability using a journalistic red-teaming approach and a "False Claim Fingerprint" database.' (marker [1])
• T6 a product as a tool: 'False Claim Fingerprints: A continuously updated database tracking false narratives across text, images, audio, and video to help safeguard AI models and brands.' (markers [1], [2])
• T7 a product as a tool: 'Brand Misinformation Monitoring: Real-time tracking and early-warning alerts designed to help advertisers and enterprises avoid funding or amplifying AI-generated falsehoods.' (marker [1])
• T8 closer: 'Would you like to explore specific data from the AI False Claims Monitor or learn how NewsGuard audits generative AI chatbots?'

• B1 NewsGuard — Tracking AI-enabled Misinformation: 3,749 AI Content Farm sites (and Counting), Plus the Top False Claims Generated by Artificial Intelligence Tools - NewsGuard fetched 2026-10-29T05:07:16Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 b8b3e0e8b8f2aea4939149fd4c1c69773427bc39a5da77788bbce7f004896c38; text sha256 f1762ff5b7fe3e9578d1af936f47da0f82d052bd6873a8aa572c8e3a000180ad
• B2 NewsGuard — AI False Claims Monitor - NewsGuard fetched 2026-10-29T05:08:00Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 18cce1f6eab47c0bbdc041084c01eaa999cc11acdf8f17f1ff87c300cc14062c; text sha256 c5a60b0da5631ec13a9198d976d82f2d160f3fc89849f7fbc4857ab336e84d2c
• B3 NewsGuard — Artificial Intelligence - NewsGuard fetched 2026-10-29T05:08:01Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 96f64dfa3dd77551b6bf376d25c37af762c94a0763477e3371a17dc92c1a7820; text sha256 9cf722cabe36140285d9f5c776cc55645bfddc8eb48db3ebf8e0a5f41d8829b4
• B4 NewsGuard — NewsGuard Launches Tracking Center for AI-Generated News fetched 2026-10-29T05:08:02Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 2ba67f597de4afb3f7f1c6a08cf2a9de8f31a508d39fff24140bfc00916f9dcb; text sha256 ad201e628d3868a36ab3aaef798d33bfa52da64f5bfc77f3692ed76111ba0eda
• B5 NewsGuard — Brand Misinformation Monitoring Service - NewsGuard fetched 2026-10-29T05:08:03Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 0d53455b4cc4b485c7173f8abf7afb4d62b1ebfea6214a130a1746b5076d5b1c; text sha256 1f11ed98d0f500f75bfa9306dbc64145ac10fbf7ae964db50234b6ae1e96ca8e
• B6 Wikipedia — NewsGuard (the operator's article; the center has none) fetched 2026-10-29T05:07:16Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw sha256 c08d687b7d20af61938d252914e0a7c0e5303e7c343cc7f59a2d50b7dcd0db72; rendered page and two Special:Search responses saved beside it

E_A, the archive's entity: The AI Tracking Center as the MMRS meta-aggregator records it (editor Lee Sharks, 2026-10-28, corrected 2026-10-29): a monitor, run by NewsGuard, that catalogues AI content farm websites, unreliable news sites run with little or no human oversight, by four criteria, 3,749 in 16 languages as of 23 June 2026; its selection leaves out what discloses itself; it records the supply side of machine-mediated reception, the sites that answer surfaces may retrieve and cite, and its series, the AI False Claim Monitor, a distinct entity and its part, records the composition side; active, kept apart from the observation that read it; held, by the Capture Registry, as the catalogue and not as NewsGuard's offer. In the archive's class reading, content farms compress journalism into SEO-optimized paraphrase, and the compression burns authorial context and citational provenance (#641). MMRS-01 MMRS-11 MMRS-12 MMRS-13 MMRS-03 MMRS-04 MMRS-05 MMRS-06 MMRS-07 MMRS-02 MMRS-19 MMRS-16 MMRS-17 CAP-1-04 CAP-1-07 O641-01; defining papers #641

E_F, the field's entity (type C — a conventional reading holds it (public entity: a monitor, NewsGuard's special report cataloguing AI content farm websites, with a selection of NewsGuard's AI reports; on its operator's site among its commercial products)): The NewsGuard AI Tracking Center, launched on 8 June 2023 as the 'Unreliable AI-Generated News Tracking Center' (B4), intended 'to highlight the ways that generative AI has been deployed to turbocharge misinformation operations and unreliable news' and including a selection of NewsGuard's AI reports, insights and debunks (B1); its unit the AI Content Farm, by four criteria (substantial AI-produced content, no significant human oversight, a presentation an average reader could take for human journalism, no clear disclosure), sites with effective human oversight excluded (B1); 3,749 such sites in 16 languages by 23 June 2026, 150 at launch (B1, B4); funded in many cases by programmatic advertising (B1, B4); the full list on request (B1, B4); run by NewsGuard, whose pages also carry its AI audit series (B2) and its commercial products for AI companies and brands (B3, B5). No Wikipedia article; NewsGuard's article reports its AI audits and has no section on the center (B6).

ρ = extends: The archive reads E_F as the entity and adds senses, limits and relations on it: what the selection leaves out (MMRS-04, MMRS-05), the supply side and the composition side (MMRS-06, MMRS-07), the series as a distinct part (MMRS-19), the catalogue kept apart from NewsGuard's products (CAP-1-06, CAP-1-07), the lifecycle apart from the observation (MMRS-16, MMRS-17), the class reading of content farms (O641-01), and the reception record (CAP-1-01, CAP-1-05). MMRS-04 MMRS-05 MMRS-06 MMRS-07 MMRS-19 CAP-1-06 CAP-1-07 MMRS-16 MMRS-17 O641-01 CAP-1-01 CAP-1-05. Guard: ρ = extends claims an addition and no more (§1.2a). Where an archive claim and a field claim differ in grain (the record's single exit by disclosure, MMRS-04, beside the field's two, F16 and F14; 'hosts', MMRS-02, beside 'a selection of NewsGuard's reports', F17), both are composed with their sources (P_coh) and the difference is carried open. #641's claim is said of content farms as a class, not of the center by name; it enters by O-class on a class sourced from the field (B1, B4).

The case: equal terms, ontology included (v0.9). The archive's account of the AI Tracking Center extends the field's ontology on the field's own object. The MMRS record holds the entity the field holds: NewsGuard's catalogue of AI content farm websites by four criteria, 3,749 in 16 languages, with its series, every field fact restated from the center's own page with source and basis 'operator' (census.json evidence). It adds senses the field does not have: what the selection leaves out, read as disclosure (MMRS-04); the center's place on the supply side of machine-mediated reception and its series on the composition side (MMRS-06, MMRS-07); the lifecycle kept apart from the observation (MMRS-16, MMRS-17); the catalogue held as the entity against the composition that widens it to NewsGuard's offer (CAP-1-06, CAP-1-07); and a class reading of content farms (#641). Nothing in the archive denies a field presupposition, and the archive supplies no ontology the field lacks: the law of the suppressed witness (§0.10) does not apply.

#MMRS articles.py ARTICLES['newsguard-ai-tracking-center'], kernel: NewsGuard's AI Tracking Center catalogues AI content farm websites, unreliable news sites run with little or no human oversight: 3,749 in 16 languages as of 23 June 2026.

#MMRS articles.py ARTICLES['newsguard-ai-tracking-center'], section 'What its selection leaves out': Disclosure takes a site out. A site that says its content is machine-made is outside the catalogue whatever it publishes.

#MMRS articles.py ARTICLES['newsguard-ai-tracking-center'], section 'Place in the field': It records the supply side of machine-mediated reception: the sites that answer surfaces may retrieve and cite. Its series, the False Claim Monitor, records the composition side.

#MMRS census.json rows[slug == 'newsguard-ai-tracking-center'], evidence.unit: {"src": "newsguardtech.com/special-reports/ai-tracking-center", "basis": "operator"}

#capture-registry newsguard-ai-tracking-center-aio-20261029, analysis: The center is composed from its maker's pages, and the maker's products come with it: the entity widens from a catalogue to a company's offer. The record keeps the catalogue as the entity.

#641 § 'The extraction crisis': Content farms compress journalism into SEO-optimized paraphrase.

equal case: each field claim is composed as received beside the archive claims on its lineage; no receiving text stands between them (v0.9 §4.3). 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 10 (positions [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.

NewsGuard's AI Tracking Center, launched on 8 June 2023 as the "Unreliable AI-Generated News Tracking Center" (the launch release), "is intended to highlight the ways that generative AI has been deployed to turbocharge misinformation operations and unreliable news" and includes a selection of NewsGuard's reports, insights and debunks related to artificial intelligence (the center's page). Generative AI "has been a boon to content farms and misinformation purveyors alike" (the center's page), and Steven Brill likened the new sites to the Macedonian content farms of a few years before, cheaper to produce and more prolific (the launch release). A domain enters when it meets four criteria, a substantial portion of AI-produced content, publication without significant human oversight (chatbot error messages are a sign), a presentation an average reader could take for human journalism, and no clear disclosure, and sites that use AI with effective human oversight are not AI Content Farms (the center's page); in 2023 the unit was the "Unreliable AI-Generated News" site (the launch release). NewsGuard counted 49 sites in May 2023 and 150 in 12 languages at the launch (the launch release) and 3,749 in 16 languages by the page's update of 23 June 2026, the page signed by ten NewsGuard analysts, the sites bearing generic names and churning out dozens of articles a day (the center's page). In many cases the sites live on programmatic advertising, which places ads "through ad-tech firms such as Google" (Crovitz, the launch release); unless brands exclude them their ads create "an economic incentive for their creation at scale" (the center's page), and generative AI is expected to grow the $2.6 billion a year directed to misinformation sites (the launch release). The full list of domains is given on request (the center's page; the launch release). The page lists NewsGuard's AI reports, among them the monthly and then quarterly AI False Claims Monitor, a journalistic red-teaming of the leading chatbots on false claims from the Fingerprints database that found them repeating false claims more than 28% and then more than one third of the time, and debunks from the False Claim Fingerprints catalog (the center's page; the series' page). Other reports find Sora videos unrecognized in 78-95 percent of prompts, Pravda network claims repeated 33 percent of the time, DeepSeek failing 83 percent of the time, and 67% of top news sites blocking AI chatbots, which may then rely on lower-quality sources (the center's page), the claims increasingly from state influence operations (the series' page). NewsGuard sells AI companies a Source Reliability Ratings datastream on 35,000 sources, a False Claim Fingerprints datastream and bespoke red-teaming, the offer resting on AI products erring 40+% of the time on news (NewsGuard's AI page; the series' page; the launch release), and sells brands a Brand Misinformation Monitoring service that tracks false narratives across social platforms, GenAI outputs and websites (the service's page). NewsGuard, launched in 2018 by Steven Brill and Gordon Crovitz, rates news websites and licenses its ratings (the launch release); Wikipedia calls it "a rating system for news and information websites" and reports its AI audits, 35% false claims on average in August 2025 and Sora 2 at 80%, and has no article or section on the center (Wikipedia).

NewsGuard's AI Tracking Center catalogues AI content farm websites, "unreliable news sites run with little or no human oversight", and the MMRS record files it as a monitor, its role monitoring, its unit the AI content farm website, at its own address on newsguardtech.com (MMRS record); in its own words it "is intended to highlight the ways that generative AI has been deployed to turbocharge misinformation operations and unreliable news" (the center's page). NewsGuard launched it on 8 June 2023 as the "Unreliable AI-Generated News Tracking Center," "cataloging the ways that generative AI has been deployed to turbocharge misinformation operations and low-quality content farms" (NewsGuard's launch release), and it includes a selection of NewsGuard's reports, insights and debunks related to artificial intelligence (the center's page). The MMRS record gives the center's lifecycle as active and keeps it apart from its own observation, a reading of the center's page on 28 October 2026 for its criteria, count and series (MMRS record). The rollout of generative AI tools "has been a boon to content farms and misinformation purveyors alike" (the center's page); at the launch NewsGuard's co-CEO Steven Brill likened the new sites to "the misinformation-spreading Macedonian content farms from a few years ago", cheaper to produce and more prolific (the launch release); and compression studies reads content farms as a class of the semantic economy's compression: they "compress journalism into SEO-optimized paraphrase", and in each such compression "authorial context, experiential density, citational provenance, the labor of witness" are burned (#641). A domain enters the catalogue when it meets "all four of the following criteria": clear evidence that a substantial portion of its content is produced by AI; strong evidence that the content is published "without significant human oversight", of which error messages or other language specific to chatbot responses are a sign; a presentation under which "an average reader could assume that its content is produced by human writers or journalists", by layout, generic or benign name or other news-like content; and no clear disclosure that the content is produced by AI (the center's page), the four restated in the MMRS record's article and its unit field (MMRS record). At the launch the unit was named the "Unreliable AI-Generated News" website (UAIN): "low-quality news sites that are mostly or entirely produced by artificial intelligence and that operate with little to no human oversight" (the launch release). What the selection leaves out is stated from two sides: sites that use AI tools "but also deploy effective human oversight" "will not be considered AI Content Farms" (the center's page), and, in the MMRS record's reading, "Disclosure takes a site out. A site that says its content is machine-made is outside the catalogue whatever it publishes" (MMRS record). NewsGuard counted 49 such sites at its first count in May 2023 and 150 in 12 languages at the launch, a number that had "more than tripled in past month" (the launch release); by the page's last update, 23 June 2026, its team had identified 3,749 AI Content Farm news and information websites spanning 16 languages, from Arabic to Turkish (the center's page), which is the MMRS record's count, as of that date (MMRS record). The page carries the names of ten NewsGuard analysts, McKenzie Sadeghi first, and describes the sites as typically bearing generic names such as Times Business News and Business Post, churning out dozens of articles a day, and "often the originators of false claims including about top brands, public health, political leaders and celebrities" (the center's page). The catalogue itself is not on the page: from the launch the full list of domains has been offered on request to researchers, platforms, advertisers, government agencies and other institutions ("Request domain list", the launch release), and the page now offers it beside "our services for generative AI companies" (the center's page). "In many cases", the center's page and the launch release say alike, the sites' revenue model is programmatic advertising, which delivers ads "without regard to the nature or quality of the website"; NewsGuard's co-CEO Gordon Crovitz said at the launch that this model "seems to be working", the ads appearing "through ad-tech firms such as Google" (the launch release). Two predictions ride on the model: unless brands take steps to exclude untrustworthy sites, their ads "will continue to appear on these types of sites, creating an economic incentive for their creation at scale" (the center's page), and generative AI "is expected" to let more players tap into or grow the estimated $2.6 billion a year that NewsGuard and Comscore found unintentionally directed toward misinformation news sites (the launch release). Below the criteria the page lists "a selection of NewsGuard’s reports and research related to AI" and a selection of false narratives originating from AI tools that NewsGuard has debunked in its False Claim Fingerprints catalog, the first an AI-edited image of an Iranian missile (the center's page). Among the reports is the recurring audit of leading chatbots, which the MMRS record names as the center's series, the AI False Claim Monitor, and relates to it as a part (has_part) (MMRS record): its editions run from a monthly monitor in July 2024 to a quarterly one by January 2026, which found 11 tools repeating false claims "more than 28%" of the time, and one audit found the rate "nearly doubled", to "more than one third of the time" (the center's page; the series' page). The series' page says that since its launch in July 2024 the monitor "has set a new standard for measuring the trustworthiness of the AI industry in the domain of news", by a journalistic "red-teaming" of AI on 30 prompts per model drawn from 10 false claims a month in the Fingerprints database, its analysts evaluating the top models monthly "to identify vulnerabilities in AI systems", and its own tests finding that this red-teaming "reduces failures by two thirds or more" (the series' page). The page's other reports find that three chatbots could not tell Sora videos in 78-95 percent of prompts, that ten tools repeated false claims of the pro-Kremlin Pravda network 33 percent of the time, that DeepSeek's chatbot failed 83 percent of the time, and that AI chatbots "Are Blocked by 67% of Top News Sites": "Most high-quality news sites request to block AI chatbots", and the chatbots "may be forced" onto lower-quality sources (the center's page); the series' page finds false claims "increasingly generated by Russian, Chinese, and Iranian state-sponsored influence operations" (the series' page). The MMRS record places the center on the supply side of machine-mediated reception, "the sites that answer surfaces may retrieve and cite", and its series on the composition side (MMRS record), and the blocking report reaches the same pairing from the other end: chatbots that "may be forced to rely on lower-quality and misinformation-prone sources" (the center's page). The center is NewsGuard's, and NewsGuard sells beside it to AI companies: a Source Reliability Ratings datastream of human-labeled data on 35,000 news and information sources, for training and RAG and to "avoid citing unreliable sources in responses", and a False Claim Fingerprints datastream, "A continuously updated database of labeled data about false claims" across text, images, video and audio, offered as protection against false claims entering responses through training data or RAG (NewsGuard's AI page); bespoke red-teaming and a licence to the Fingerprints database (the series' page); and NewsGuard for AI, launched in February 2023 (the launch release), the offer grounded in the claim that without quality sourcing AI products err "40+% of the time" on news (NewsGuard's AI page). To brands it sells Brand Misinformation Monitoring, a real-time service "to identify and respond to emerging false narratives before they cause reputational harm", tracking false narratives about a brand across social platforms, GenAI outputs and websites, with early warning alerts and "actionable insights into how clients are represented—or misrepresented—by AI models" (the service's page). The Capture Registry holds the center to the catalogue: False Claim Fingerprints and the Brand Misinformation Monitoring Service "are NewsGuard's commercial products", the composition at the center's name files them under the center as "Core Tools & Resources", and so "the entity widens from a catalogue to a company's offer", while "The record's center holds the content-farm catalogue and the False Claim Monitor series" and "The record keeps the catalogue as the entity" (Capture Registry, 2026-10-29). NewsGuard was launched in March 2018 by Steven Brill and Gordon Crovitz and provides credibility ratings and "Nutrition Labels" for news and information websites, licensed by browsers, news aggregators, education companies, and social media and search platforms (the launch release); the MMRS record gives NewsGuard as the center's operator, on the basis of the center's own page (MMRS record). Wikipedia's article on NewsGuard opens "NewsGuard is a rating system for news and information websites", files it in its short description as a "Browser plugin", dates the company's founding to 2018 by Brill and L. Gordon Crovitz, and says the company also provides misinformation tracking and brand safety (Wikipedia, NewsGuard). Its section on disinformation in AI reports that NewsGuard "has audited the output of various AI models to identify disinformation", that leading generative AI tools repeated false news claims 35% of the time on average in August 2025, that Sora 2 produced videos advancing fake claims in 80% of cases, and that, according to NewsGuard, Russian propaganda networks are attempting to "infect" large language models (Wikipedia, NewsGuard); it has no section on the center, and a search of English Wikipedia's articles for the phrase AI Tracking Center answers "There were no results matching the query." (Wikipedia, 2026-10-29). At the center's name on Google's AI Overview on 29 October 2026 the Capture Registry reads a composition that "gives what the center tracks, its headline figures, and the NewsGuard tools around it, all from NewsGuard's pages", with "3,749 sites in 16 languages" the record's figure, checked that day against the MMRS record and the entity's own pages, and states that "The video-recognition failure rates and the 67% blocking figure are not verified here" (Capture Registry, 2026-10-29); the MMRS record links that composition as "the composition layer's account of the entity" (MMRS record).

T against L_F(B) (AIO's ontology at work on its own cards): One transcript, five cards, every card NewsGuard's (B1 the center's page, B2 the series' page, B3 NewsGuard's AI page, B4 the 2023 launch release, B5 the brand service). Composed from card 1 at its grain: the count, 3,749 sites in 16 languages (T2; F4), the news-like layouts (T2; F15), programmatic advertising (T2; F7), Sora videos unrecognized in 78% to 95% of prompts (T3; F21), the 67% blocking figure (T4; F24), the series' journalistic red-teaming and the Fingerprints database (T5; F29, F30). Limit dropped: 'monetize via programmatic advertising' (T2) where both NewsGuard texts say 'In many cases' (F7, F49). Force raised: 'actively block AI crawlers and chatbots, pushing AI systems toward lower-quality data sources' (T4), where the report says 'Most high-quality news sites request to block AI chatbots' and that the chatbots 'may be forced' onto lower-quality sources (F25, a hypothesis); 'crawlers' is in no field source. Substituted reading: 'occasionally propagate state-sponsored disinformation' (T3), where the field's figure is 33 percent for the Pravda network (F22) and the series finds the claims 'increasingly generated' by state operations (F34); the range '28% to over 33%' (T3) joins the quarterly 'more than 28%' (F18), 'more than one third' (F19) and the Pravda 33 percent (F22), three audits, as one finding. Composed without a source in the field: 'an ongoing investigative project' (T1); neither 'investigative' nor 'ongoing' is in the saved texts; 'unedited' (T2) is the card's snippet word, the page's terms being 'without significant human oversight' and 'without adequate editing' (F12, F13). Assigned to another (the series' and the products' senses composed as the center's): the center 'monitors and catalogs … AI tool vulnerabilities' (T1), where vulnerabilities are the series' object (F31) and the center's own statement is to 'highlight the ways that generative AI has been deployed to turbocharge misinformation operations and unreliable news' (F1); the three 'Core Tools & Resources' (T5–T7) file the series and two commercial products (F33, F39, F40, F55–F58) under the center, which the Capture Registry records (CAP-1-03, CAP-1-06). Substituted reading: Brand Misinformation Monitoring 'designed to help advertisers and enterprises avoid funding or amplifying AI-generated falsehoods' (T7), where its own page gives it 'to identify and respond to emerging false narratives before they cause reputational harm', across social platforms, GenAI outputs and websites (F55, F56); 'funding' is the content farms' economics (F8), and the narrowing to 'AI-generated' is T's. Available, not composed: the four criteria as admission and their governing limit (F10–F16, F14), the launch of 8 June 2023 and its first name and unit (F42, F43; card 4's snippet carries the date), the count's history (F41, F44, F45), the list on request (F9, F51), the predictions on the economics (F8, F50), Brill's and Crovitz's statements (F46, F47, F48), the compilers (F5), the debunks (F26, F27), DeepSeek (F23), the series' launch date and self-assessments (F28, F32), Source Reliability Ratings, card 3's own snippet (F37, F38), the 40+% claim (F36), NewsGuard for AI (F53), NewsGuard as operator (F52, F54). Content replaced by use: the closer (T8). Attribution: the body names NewsGuard (T1) and its cards are NewsGuard's; no grain breach.

v0.8 §0.5: the first diff is already ontological. In the field's own ontology the center is a special report: a catalogue of AI content farms by four criteria, with a selection of NewsGuard's AI reports, on a company's site among its products. T composes the center as a company's toolkit: its definition takes the series' object (vulnerabilities), its 'Key Findings' are the reports held, and its 'Core Tools & Resources' are the series and two products; the catalogue's criteria and limits, which make it a catalogue, are not composed. The row records this as observed and claims nothing about the composer's internals (§0.5).

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, at the archive's own force: the center's kind and role, a monitor (MMRS-11, MMRS-12); what its selection leaves out, 'Disclosure takes a site out' (MMRS-04, MMRS-05), beside the field's own second exit by effective oversight (F14); its place in machine-mediated reception, the supply side, 'the sites that answer surfaces may retrieve and cite', with its series on the composition side (MMRS-06, MMRS-07); the series as a distinct entity related as a part (MMRS-02, MMRS-19); the catalogue kept as the entity and NewsGuard's products kept apart from it (CAP-1-03, CAP-1-04, CAP-1-06, CAP-1-07); the lifecycle kept apart from the observation (MMRS-16, MMRS-17, MMRS-18); the archive's reading of content farms as a class of compression that burns citational provenance (O641-01, #641, reached beside the string pass); and the reception record, the composition of 29 October read against the record (CAP-1-01, CAP-1-02, CAP-1-05, CAP-1-08, MMRS-20). Not surfaced: the MMRS record, the capture and #641 (no card cites them). Restated, one lineage with the field: the count and its date (MMRS-01, MMRS-14, MMRS-15; F4, F5), the criteria (MMRS-03, MMRS-13; F10–F16), the operator (MMRS-10). No archive claim is marked; none contradicts a field claim. Qualified: 'hosts' (MMRS-02) beside the page's 'a selection of NewsGuard's reports' (F17), carried open. Every field claim is composed as received (ontology.translation). The measure: 0 of 10 KO_B sentences survive in the KO (§4.10); no opening run, no block (measure).