Two claims, one from Fox News, one from CNN, walk through the full MediaReceipts pipeline from broadcast to published verdict. The same methodology, the same standard, applied identically regardless of network or political lean.
Iran announced that the Strait of Hormuz is fully open and ready for business.
"Iran announced that the Strait of Hormuz is fully open and ready for business."
Is this claim worth the cost of deep verification?
The AI research assistant assessed the claim across all four Gate 1 dimensions: claim structure, attribution quality, domain source availability, and verification output. It produced a nomination (APPROVE) and a reviewer summary. The human editor then reviewed the AI's assessment and made the final call.
Gate 1 confirmed that an Iranian announcement about the Strait verifiably occurred. But Gate 1's job is viability triage, not verdict assignment. The question of whether Host A's characterization of the announcement ("fully open and ready for business") accurately reflects what Iran actually said is Gate 2's responsibility. Gate 1 approved the claim forward with the note that preliminary verification surfaced a gap between the announcement's declaratory language and operational reality.
What does the evidence actually show?
Gate 2 conducted an independent research pass, starting from scratch rather than relying on Gate 1's preliminary findings. The research established the following evidence:
On April 17, 2026, Iranian Foreign Minister Abbas Araghchi posted on X: "In line with the ceasefire in Lebanon, the passage for all commercial vessels through Strait of Hormuz is declared completely open for the remaining period of ceasefire, on the coordinated route as already announced by Ports and Maritime Organisation of the Islamic Rep. of Iran."
Iran's actual statement was time-limited (tied to a Lebanon ceasefire), conditionally routed (requiring coordination with Iranian maritime authorities), and subject to exclusions. IRGC-affiliated media simultaneously described conditions for safe passage that resembled the restrictions Iran had imposed for weeks.
The phrase "fully open and ready for business" does not appear in Iran's announcement. It originates from President Trump's Truth Social post on the same day: "THE STRAIT OF HORMUZ IS COMPLETELY OPEN AND READY FOR BUSINESS AND FULL PASSAGE." Host A's on-air language tracks Trump's social media characterization of the Iranian announcement, not the announcement itself.
Within approximately 24 hours, Iran reversed course. On April 18, Iran's joint military command stated that "control of the Strait of Hormuz has returned to its previous state… under strict management and control of the armed forces." IRGC gunboats fired on vessels attempting to transit the Strait. India summoned Iran's ambassador over attacks on Indian-flagged ships.
The post-broadcast reversal is corroborating context. It confirms the announcement was fragile and conditional, but it does not carry the verdict. The mischaracterization was complete at airtime, based on what was knowable on April 17.
Source independence: INDEPENDENT. Multiple outlets reporting from different institutional origins. Source strength: MEDIUM, reflecting the active military/diplomatic domain.
The corrective operation test determines which misleading trigger is primary by asking: what is the simplest fix that would make this claim accurate?
Fix 1: Replace the characterization. Replace "fully open and ready for business" with what Iran actually said: "completely open for the remaining ceasefire period under specified conditions." This is a characterization replacement, the FRAMING pattern. The claim took a real event and reframed it with language that creates a false inference about its scope and nature.
Fix 2: Add missing context. Add the Iranian parliament speaker's warning that the Strait would close again if the U.S. continued its blockade, and the IRGC conditions requiring coordination. This is context addition, the OMISSIONpattern. But even with this context added, the core characterization "fully open and ready for business" would still mischaracterize Iran's statement.
Because the characterization replacement is the more fundamental fix, FRAMING is primary. OMISSION is recorded as a secondary, contributing mechanism.
The editor forms an independent verdict before seeing the AI's recommendation.
Under the staged disclosure protocol, the human reviewer reads the evidence summary and sources first, then records their own preliminary verdict beforeseeing what the AI nominated. This prevents the AI's recommendation from anchoring the human's judgment.
"Preliminary assessment was MOSTLY TRUE because I read 'fully open and ready for business' as an overstated characterization of a real announcement, which maps to a factual inaccuracy in degree rather than a structural distortion. After reviewing the full sourcing, the problem isn't a wrong detail in a correct picture. Every specific condition attached to the announcement (ceasefire linkage, coordination requirements, hostile-nation exclusions, same-day re-closure warnings, 24-hour reversal) was stripped out, and the language itself traces to Trump's framing rather than Iran's actual statement. That's right details, wrong picture: MISLEADING with OMISSION primary and FRAMING secondary, not MOSTLY TRUE."
The reviewer's preliminary-to-final shift is a textbook example of the MOSTLY TRUE / MISLEADING boundary in action. The boundary test asks: where does the error live?
At first read, the reviewer treated "fully open and ready for business" as an overstated characterization, a wrong detail in an otherwise correct picture (the announcement did happen). That's the MOSTLY TRUE pattern.
After examining the full sourcing, the reviewer concluded that the error isn't in a detail; it's in the picture itself. The claim's specific facts are all technically grounded (an announcement occurred), but the overall impression (unconditional, operational reopening) is demonstrably false. That's the MISLEADING pattern: right details, wrong picture.
The staged disclosure protocol is designed to produce exactly this kind of independent reasoning. The reviewer reached MOSTLY TRUE without seeing the AI's MISLEADING nomination, then independently worked through the evidence to arrive at the same conclusion the AI reached, for their own documented reasons, not because the AI told them to.
Gate 2 confirmed Gate 1's MISLEADING verdict but upgraded confidence from MEDIUM to HIGH. Gate 2 also identified a secondary FRAMING trigger not caught by Gate 1: specifically, that Host A attributed Trump's Truth Social language to Iran rather than Iran's own qualified statement. The additional wire-service and expert-analysis sourcing at Gate 2 strengthened the evidence base beyond what Gate 1's triage-scoped pass had found.
Every specific fact in the claim has a kernel of truth: Iran did make an announcement about the Strait on April 17. But the characterization "fully open and ready for business," sourced from a presidential social media post rather than from Iran's actual statement, creates a false impression of unconditional reopening. Iran's announcement was time-limited to a ceasefire period, required coordination with Iranian maritime authorities, and excluded "hostile nations." The claim's construction leads a reasonable viewer to a conclusion substantially different from what the complete picture would convey.
This claim demonstrates three things about the MediaReceipts process. First, the Misleading category captures claims where every cited fact is technically grounded in reality, but the overall impression is demonstrably false. Second, the corrective operation test determines the primary distortion mechanism (here, FRAMING (characterization replacement) over OMISSION). Third, and most importantly, the human reviewer initially disagreed with the AI, formed an independent preliminary verdict, then changed their mind based on the evidence, not because the AI told them to.
The staged disclosure protocol produces genuine independent judgment, not rubber-stamping.
RFK Jr. directed the CDC to stop recommending the combined MMRV vaccine for toddlers under four.
"Kennedy, who directed the CDC to stop recommending the combined MMRV shot for toddlers under four, MMRV measles, mumps, rubella…"
Is this claim worth the cost of deep verification?
The AI research assistant assessed the claim across all four dimensions and produced an APPROVE nomination with a lifecycle flag. The human editor then reviewed the assessment, assigned the claim type, and made the final approval decision.
Gate 1's methodology requires a lifecycle check on all OFFICIAL_FINDING claims. A finding that has been stayed, appealed, or overturned must be flagged, because presenting a suspended policy as currently operative is exactly the kind of omission that the Misleading category exists to catch.
In this case, the ACIP vote happened (September 2025), but a federal court subsequently stayed it (March 16, 2026). This is a textbook lifecycle-currency case: the finding existed, but its current status is materially different from what a present-tense characterization implies. Gate 1 approved the claim forward with this flag prominently noted for Gate 2.
What does the evidence actually show?
In June 2025, HHS Secretary Kennedy fired all 17 existing members of the CDC's Advisory Committee on Immunization Practices (ACIP) and replaced them with his own appointees. On September 18–19, 2025, this reconstituted panel voted 8-3 to stop recommending the combined MMRV vaccine for children under age 4. The CDC adopted the recommendation.
Host B's factual core, that Kennedy directed a change to MMRV recommendations, is substantially accurate.
Since 2009, CDC guidance had stated a preference for separate MMR and varicella vaccines for the first dose in children aged 12–47 months, with the combined MMRV offered only as a parental option; by 2025, 85% of children already received separate first doses. The September 2025 vote eliminated an option. It did not reverse an affirmative recommendation. Host B's phrasing "stop recommending" overstates the scope of the change.
On March 16, 2026, 32 days before the April 17 broadcast, U.S. District Judge Brian Murphy issued a preliminary injunction that stayed Kennedy's 13 ACIP appointments and invalidated all votes taken by the reconstituted panel, expressly including the MMRV restriction. The prior vaccine schedule was restored pending litigation.
The injunction was widely covered at the time of its issuance by Reuters, the American Academy of Pediatrics, Axios, The Hill, STAT News, Roll Call, NPR, and others. Host B's own program covered the injunction on the day it was issued. As of the April 17 broadcast, the MMRV policy change Host B described was not in effect. A viewer hearing "Kennedy directed the CDC to stop recommending MMRV" on April 17 would form the impression that this is current CDC policy. It was not.
Source independence: INDEPENDENT. Multiple sources from different institutional origins. Source strength: HIGH. Full public institutional record with court filings, meeting transcripts, and multi-outlet coverage.
Fix 1: Add missing context. Add the fact that the policy was judicially stayed on March 16, that the prior vaccine schedule was restored, and that the MMRV restriction is not currently in effect. This is context addition, the OMISSION pattern. Adding this context reverses what a reasonable viewer concludes about current operative policy.
Fix 2: Replace the characterization. Replace "stop recommending" with "eliminated as an option," because since 2009, MMRV was already only a parental fallback, not the affirmative default. This is a characterization replacement, the FRAMING pattern. It meets the materiality threshold but does not address the more fundamental distortion (the currency problem).
The currency omission is dominant because it reverses what a reasonable viewer concludes about the current state of CDC policy. OMISSION is primary. FRAMING is recorded as secondary.
The editor forms an independent verdict before seeing the AI's recommendation.
The same staged disclosure protocol applies. The reviewer reads the evidence, forms a preliminary verdict independently, then sees the AI's nomination and makes a final determination.
The reviewer independently reached MISLEADING before seeing the AI's nomination, then confirmed the verdict after reviewing the full disclosure. When the AI and the human reviewer agree at every stage, no override rationale is required. The agreement itself is the signal. Both identified the March 16 injunction as the dispositive omission.
The AI classified this claim as CONTESTED, but the contested designation reflects two secondary boundary questions, not the primary verdict:
First, whether the secondary FRAMING trigger (the pre-2009 baseline context) clears the materiality test on its own. The AI noted this is meaningful but less decisive than the injunction omission alone. Second, whether characterizing the ACIP vote as Kennedy personally "directing" the CDC is itself a framing inaccuracy that might push toward MOSTLY TRUE. The AI resolved this by noting that Kennedy's personal causal role (firing all 17 prior ACIP members and appointing vaccine-skeptic replacements) makes "directed" a reasonable editorial shorthand, not a material inaccuracy.
Neither boundary question threatened the primary verdict. The dispositive basis for MISLEADING was and remained the injunction omission.
Gate 2 fully concurred with Gate 1's MISLEADING/OMISSION verdict and HIGH confidence assessment. Gate 2 added a secondary FRAMING trigger (the pre-2009 baseline context that reframes the scope of Kennedy's action) and confirmed through direct sourcing that Host B's own program had covered the injunction on March 16, 2026, strengthening the finding that the omission was material.
The September 2025 ACIP action and Kennedy's role in driving it are accurately characterized. But on April 17, 2026, the policy Host B described had been judicially stayed for 32 days. The prior vaccine schedule was back in effect. Host B's own program had covered the injunction the day it was issued. A viewer hearing this claim would conclude that the MMRV restriction is current CDC policy. It was not. Every specific fact is technically grounded, but the omission of the injunction creates a false impression about what policy is actually operative.
This claim demonstrates the lifecycle check in action: official findings that have been stayed, overturned, or superseded cannot be presented as current policy without noting their changed status. It also shows what full agreement looks like in the pipeline. When the human reviewer independently reaches the same verdict as the AI, the agreement is recorded as a data point for inter-rater reliability analysis. Agreement is not rubber-stamping; it is two independent assessments converging on the same conclusion for documented reasons.
These two claims come from opposite ends of the cable news spectrum. Both received the same two-gate review with the same staged disclosure protocol. Both were assessed against the same source hierarchy, the same corrective operation test, and the same materiality threshold. Both produced Misleading verdicts with dual-trigger documentation. The methodology does not adjust its rigor based on network or political lean.
Critically, the human reviewer is not a formality. In the first example, the reviewer's independent preliminary verdict disagreed with the AI, and the reviewer changed their mind based on the evidence, documenting exactly why. In the second example, the reviewer independently reached the same conclusion the AI did, and the agreement was recorded as a calibration data point. Both outcomes, override and agreement, are expected and healthy. The system is designed to produce independent human judgment, not to ratify the AI's output.
Every published verdict in the MediaReceipts database went through this process. The full methodology is available on the Methodology page, and every individual claim's sources, reasoning, and review record are published on its claim page.