Bounded acquisition closed on 9 September 2026. Three primary records were read: GDI's April 2024 parliamentary evidence, its privacy/research policy updated April 2026, and IAS's dated announcement that it was discontinuing GDI input. Two original HTML files and one official-reader derivative were retained. The exact procedures differ in date and scope; none supplies an executed distribution agreement or proves an individual appeal changed advertising decisions.
Historical publisher appeal and the decision split
GDI CEO Clare Melford's written evidence FON0071, dated 29 April 2024, expressly says publishers can appeal their ratings through a linked review form on GDI's contact page. Under GDI's Review Process, it describes a two-to-three-week review and regular site reassessment, with ratings able to change when content changes. This is an explicit publisher-facing procedure described by the organisation to Parliament, stronger evidence than an invitation for academics to critique methodology. It is not a tribunal finding, a tested service-level guarantee, or proof of a successful appeal.
The preceding GDI's Service section describes risk data produced using AI and human review. It says advertising technology companies use that data in purchasing, and attributes ultimate purchasing and data-use control to licensees. GDI also says it charges for the data. These are GDI's statements about its commercial mechanism, not independently read licence terms. They support separating classification, licensing and the recipient's advertising decision; they do not establish that GDI could compel a customer to undo a decision after a rating changed. The evidence does not publish an evidentiary test for appeal, reviewer independence, reasons requirement, interim suspension, deletion deadline, update delivery protocol or recipient acknowledgement.
Source: [FON0071 official parliamentary publication](http[local research file] headings GDI's Service and GDI's Review Process; retained reader lines 41–61, date at line 74. Full substantive reader text was read. The submission refers to a screenshot and links a review form; neither the screenshot nor the target form was acquired or exercised. Do not infer that the same form or review duration remains current in September 2026. The statement is not a detailed, versioned DEL methodology or an identification of the criteria governing the 2024 Oracle segment.
Current research-output complaints: broader standing, qualified correction
GDI's Privacy Policy, last updated 20 April 2026, identifies Disinformation Index Ltd as controller. Section 15 expressly accepts questions, concerns or complaints about personal-data processing or any research outputs, and says it will consider representations and, where appropriate, review, clarify or update findings. Section 19.3 acknowledges that some projects may have no advance notice to platforms or stakeholders; it describes a post-publication enquiry route, prompt review/response and corrective action where appropriate. This gives a public route to raise a research-output concern, although it does not promise the requested substantive result.
The boundaries matter. Section 1 excludes personal-data processing for services under a separate agreement from this notice's coverage. Sections 14–16 describe data-subject rights and supervisory complaints, subject to applicable law and limitations; those do not automatically turn a corporate publisher's disagreement with a rating into a successful data-protection claim. Section 19.1 promises deletion of unnecessary or inaccurate personal data without undue delay, which is not an explicit promise to delete a domain rating or notify every advertising recipient. Section 19.3 says dedicated project notices describe retention and sharing but are not routinely published. No such project notice was obtained, and no contact was made.
This policy does not supply a numerical research-complaint deadline equivalent to the historical two-to-three-week account, an independent appeal tier, or a rule requiring downstream customers to replace cached classifications or resume advertising. It must not be retroactively applied as the DEL's 2024 operating rules. Its statement in section 18 about personal data not being subject to automated decision-making with legal or significant effects is GDI's current self-description, not an audited conclusion about historical commercial outcomes.
Source: [GDI Privacy Policy](http[local research file] sections 1, 14–16, 18, 19.1–19.4 and final update date. Entire substantive text sections 1–20 was read through the official reader. Two embedded collection/retention tables were not visually read; no table contents are inferred.
A distributor says it ended the input, rather than documenting an appeal result
IAS's IAS Enhancements to Context Control Avoidance announces a decision to discontinue the limited GDI input in that product and says its past use of GDI would not affect that solution going forward. The page describes customers selecting exclusions by categories and risk levels, but it does not publish the exact former GDI fields, a customer migration log, deletion protocol or an individual corrected classification. This is a directly named distributor's reported product change, not an observed technical removal or evidence that a particular publisher won an appeal. Its stated benefit claims, including neutrality and independence, are provider assertions rather than independent judgments.
The headline and dateline say 18 December 2025. Retained HTML explicitly separates datePublished 2025-12-18T18:17:32Z from dateModified 2026-07-06T16:30:34Z, and the posted-date elements display both dates. The September 2026 captured wording cannot be assumed identical to the original December wording; no version comparison was acquired. Do not assign an exact implementation date from those page dates alone.
Source: [IAS announcement](http[local research file] full substantive announcement, especially the paragraph beginning its continuous-improvement explanation. Full body was reader-read; original HTML date elements and JSON-LD were locally inspected. This is distinct from Oracle's announced 30 September 2024 service end in the prior accepted packet, and from current GDI research products.
Historical funding disclosure present in the same evidence
FON0071's GDI and Government Funding section says GDI had received no UK government funding since 2022; describes FCDO Russia-unit Countering Disinformation and Media Development work in 2018–2022 to develop large-scale detection and country studies; and says the supporting grant expired in March 2023. The receipt-period statement and later grant-expiry date are not a self-evident contradiction. It also attributes then-current resources to private foundations, commercial data licensing and a small EU Horizon Europe grant. This is a historical recipient disclosure, not donor-side confirmation, acquired conditions, proof of payment dates or government approval of a particular rating. No funding genealogy was pursued.
Discriminating gap and stopping boundary
The evidence now defeats the broad claim that no publisher appeal was publicly described. It also leaves a distinct question unanswered: what obliges a data recipient to stop using or replace a successfully challenged signal? GDI may revise a rating while an advertiser independently declines the inventory; the acquired documents do not reconcile those steps or prove any such instance occurred.
The most useful expected instruments are held by GDI and its relevant licensee: the dated publisher-review procedure behind the 2024 form; a service-specific notice or licence annex defining corrections, updates and deletion; and an actual recipient instruction/acknowledgement following a resolved rating. IAS's versioned discontinuation/migration specification would test its broader no-continuing-effect statement. These are missing records, not proof that no rules or corrections exist.
The 2019 GDI methodology PDF reader returned an unsafe-open non-retryable error and was closed without retry. It was not used as methodological evidence. The previously closed 2022 policy PDF and Jounce routes were not reopened even where search results resurfaced them. No named-site score audit, customer system, operational list, form submission, account, or harmful-content endpoint was accessed. The packet closes at three read primary records; acquisition and custody details are in recourse-captures.json.