X this year has circulated a formal pitch document to advertisers outlining how the platform now deploys its Grok artificial intelligence system as a core component of brand safety operations - a significant strategic shift that positions xAI's model not as a chatbot, but as an automated content scoring engine sitting between advertisers and the posts their campaigns appear next to.

The document, titled the Brand Suitability Playbook, was obtained and first reported by Adweek. It offers a detailed account of the controls X has assembled since Elon Musk's $44 billion acquisition of the platform in late 2022, when a mass departure of advertisers triggered a decline in annual ad revenue from $2.43 billion in 2021 to an estimated $1.25 billion in 2025, according to projections from Emarketer. The gap - nearly half the platform's pre-acquisition revenue - has driven X to systematically rebuild its advertiser toolkit over the past three years.

Grok replaces rule-based filtering

The most technically notable element of the playbook is the deployment of Grok for brand safety scoring. According to the document, Grok has been integrated into X's pre-bid and adjacency controls infrastructure, enabling the AI model to evaluate posts for suitability before ads are served next to them. The playbook states that Grok's involvement has enabled "enhanced contextual understanding of posts," "advanced brand suitability detection to identify sensitive topics based off of cultural trends on X," and "superior text recognition across content formats."

The outcome metrics cited in the document are sourced from X, Integral Ad Science, and DoubleVerify on a global basis. According to the playbook, the platform now achieves an average brand suitability score exceeding 97% and an average brand safety score exceeding 99.99%. These figures claim to exceed industry benchmarks, though no specific external benchmark figure is provided for direct comparison in the document.

The distinction between brand safety and brand suitability matters here. Brand safety represents a baseline floor - ensuring ads do not appear next to illegal, hateful, or explicitly harmful content. Brand suitability is the layer above that, where individual advertiser preferences come into play: an alcohol brand may accept content a children's toy company would not. Grok's deployment, X claims, handles both tiers simultaneously, operating across what the company describes as profile allowlists, reply enhancements, and pre-bid vertical video inventory.

The prior integration of IAS and DoubleVerify into X's measurement stack dates to January 2023, when the platform - then still operating as Twitter - announced feed-based measurement solutions that allowed the two verification providers to classify tweets directly above and below ad placements. The current playbook extends that architecture by layering Grok's contextual classification on top of third-party measurement. Ads in replies, according to the document, "will appear under focal posts that have been Grok-reviewed for Brand Suitability, as well as posts directly above and below." For profile placements, ads "will only appear on profiles fully vetted by Grok."

The product stack in detail

The playbook organises X's advertiser controls across three categories: platform transparency, brand safety controls, and third-party measurement. The transparency section covers X's open-source algorithmic approach, its accreditation from the Trustworthy Accountability Group (TAG), its bi-annual Global Transparency Report, Community Notes, and Grok for brand safety.

TAG certification is presented as independent validation. According to a statement from Mike Zaneis, CEO of TAG, quoted directly in the playbook: "We worked extensively with the X team and their auditor to review their policies, processes, and brand safety products to ensure X is meeting the industry's high standards. [...] We encourage brands to take advantage of X's brand safety offerings."

The advertiser controls section spans seven tools. The Placement Filter allows brands to select or deselect ad surfaces across home timelines, profiles, search results, replies, and media viewer. Adjacency Controls - described as X's "most powerful tool" for managing where brand ads appear - allow negative targeting of up to 4,000 keywords and up to 2,000 author handles. Sensitivity Settings reduce adjacency to varying levels of sensitive content. The document notes that standard sensitivity is the default, while "limited" sensitivity is the recommended setting for all campaigns. Content that violates X's rules is excluded regardless of the sensitivity level selected.

A Vertical Video Pre-Bid product offers a partnership-specific solution with Integral Ad Science, enabling advertisers to run vertical video ads against pre-screened, brand-suitable inventory. According to the playbook, this is described as a "unique offering that only X & IAS can offer," currently available in the United States only. The IAS coverage expanded previously to include vertical video adjacency classification aligned to the GARM framework, giving advertisers what X characterizes as maximum control over where their ads appear in the vertical video feed.

Trend Genius is documented as a contextual targeting product that monitors real-time conversations on X and triggers ads when key terms or hashtags begin trending. The playbook cites The National Lottery as a case example. Standard brand safety controls for Trend Genius include a first-party platform solution monitoring trending and sensitive topics, X's adjacency controls and sensitivity settings, and third-party measurement from DoubleVerify or IAS.

The Amplify Video product connects advertisers to premium publisher content via two methods: direct sponsorships with individual vetted publishers, or category-based buying across IAB content categories using curated publisher lists. The playbook also describes an Always-On Brand Suitability Protections layer, which includes a first-party blocklist that dynamically removes real-time trends that may affect suitability, an allowlist containing only eligible, brand-safe, and commercially relevant search terms, and Grok-reviewed profile and reply placements.

Community Notes as a brand signal

A substantial section of the playbook is dedicated to Community Notes, X's crowd-sourced fact-checking mechanism. The document frames Community Notes not merely as a content moderation tool but as an active brand safety signal. Research from Sorbonne Université Paris, Cornell University, and the University of Washington is cited: users are 80% more likely to delete a post that receives a Community Note, and 60% less likely to repost content with a note attached.

The mechanism operates through what the playbook calls a "bridging algorithm," which only displays notes that users with historically divergent viewpoints agree are helpful. This design, according to the document, reduces the risk of ideologically biased notes reaching broad audiences and ensures that only notes with cross-spectrum support appear on posts. The playbook describes the notes as open-source, specifying that the code is publicly available, that notes only appear where users with diverse views agree they are helpful, and that note adjacency ensures no brand adjacency to untrustworthy content.

The Community Notes program has reached one million contributors across 197 countries. A notable new development documented in the playbook is the AI Note Writers API - a pilot that allows AI tools to propose Community Notes. If those AI-proposed notes are rated helpful by people from different perspectives, they can appear on X posts. The document frames this as a "future" enhancement over the current system, with the flow shifting from: "Misleading Post → Note Writers → Proposed Notes → Note Raters → Scored Notes" to a model incorporating "Note Writers [with optional AI assistance]" using "Automated Note Generation" and "Reinforcement Learning From Community Feedback."

The playbook does not disclose which AI models or providers are involved in the AI Note Writers API pilot beyond referencing it as part of X's broader product roadmap.

Revenue context and advertiser returns

The commercial stakes behind the playbook are substantial. X received a €120 million fine from the European Commission in December 2025 for three violations under the Digital Services Act: deceptive design of its blue checkmark verification system, an inadequate advertising repository, and blocking researcher data access. The fine was the first non-compliance decision issued under the DSA.

On the advertiser side, a Grok safety failure on December 25, 2025 - in which the chatbot generated prohibited visual content - prompted significant debate within the marketing community about the reliability of automated safeguards on the platform. The incident added complexity to X's narrative that Grok is now a trustworthy brand safety tool.

Despite this, Monique Pintarelli, who was elevated in January to global head of advertising at xAI, stated at CES on January 13 that 97 of the top 100 advertisers have returned to the platform, with some spending at or above pre-acquisition levels. Large clients in the US grew by over 40% in 2025, though Pintarelli declined to share specific revenue figures. She told Digiday that X is "more focused on advertising than ever before" and described X's role in sports culture as one of its core advertising strengths going into the 2026 World Cup cycle.

The 2025 eMarketer estimate of $1.25 billion in X ad revenue compares starkly to the $2.43 billion the platform generated in 2021. Major advertisers including IBM, Disney, and Apple departed following Musk's acquisition in October 2022 but have since returned. The playbook itself does not reference the advertiser exodus period, presenting its brand safety timeline as a progression of investment and improvement starting from 2022.

What the measurement data shows

The playbook's appendix includes an advertiser recommendations table mapping brand safety products to placements and markets. For Adjacency Control (1P pre-bid), X recommends application to all campaigns across timeline, search, profiles, replies, and media viewer globally. For Sensitivity Settings (1P pre-bid), it recommends the "Limited" setting across all placements globally. For Inventory Vetting (3P pre-bid via IAS), coverage is limited to the media viewer for vertical video in the United States only, with a recommendation for "more conservative" advertisers.

Third-party measurement coverage from IAS and DoubleVerify covers brand safety rates, brand suitability rates, viewability rates, and invalid traffic rates. According to the playbook, IAS and DoubleVerify have reported that 99% of measured ad impressions appeared adjacent to content meeting the Brand Safety Floor criteria. If advertisers want to validate placement performance further, X recommends using post-bid brand safety adjacency measurement partners to externally validate the prevalence of ad placements adjacent to content deemed unsafe or unsuitable.

The DoubleVerify study published in November 2025 surveying 22,000 consumers and 1,970 marketing decision-makers across 21 countries found that nearly two-thirds of marketers advertising on social media expressed brand suitability concerns. That research - conducted independently of X - establishes the continuing anxiety advertisers carry about platform adjacency, regardless of any individual platform's claims.

Advertising attention data cited

The playbook also includes a section on ad performance, drawing on eMarketer data from April 2024 and GWI data from Q2 2024 through Q1 2025. According to these sources, X ranks first for ad attentiveness among social media platforms, and X users are 45% more likely to buy products they have seen advertised compared to non-X users. The eMarketer source cited is a specific report by Daniel Konstantinovic titled "Industry KPIs: YouTube and X emerge as leaders for social media advertising attention," published April 1, 2024.

These performance figures are presented alongside the brand safety stack, positioning X's pitch to advertisers as a combined argument: that the platform is both adequately safe and unusually effective at converting attention into purchase intent. Whether advertisers weigh these claims against the broader regulatory and safety incidents of 2025 will determine how quickly X's ad revenues close the gap with their pre-acquisition peak.

A separate incident highlighted in January 2026 by martech consultant Jonathan D'Souza-Rauto revealed that advertisers who had explicitly avoided X were inadvertently purchasing X inventory through Google Ads Search Partners, Video Partners extensions, and Performance Max campaigns - because X had made a substantial portion of its ad auctions available programmatically, with Google Ads serving as the primary buyer of that inventory. For brands with explicit policies against advertising on X, this created an unintended compliance gap that has added to the urgency of understanding exactly where ads appear on the platform.

Timeline

Summary

Who: X (formerly Twitter), operating under Elon Musk's xAI umbrella, with Monique Pintarelli serving as global head of advertising. The playbook is directed at brand advertisers, agencies, and media buyers evaluating the platform for campaign investment.

What: X has published a formal Brand Suitability Playbook that details how Grok AI is now deployed as a pre-bid and adjacency scoring engine across the platform's advertising stack, alongside a suite of advertiser controls including Placement Filters, Adjacency Controls covering up to 4,000 keywords and 2,000 author handles, Sensitivity Settings, Vertical Video Pre-Bid with IAS, Trend Genius, Amplify Video, and Always-On Brand Suitability Protections. The playbook cites 99%+ average brand safety scores and 97%+ average brand suitability scores, sourced from X, Integral Ad Science, and DoubleVerify.

When: The pitch document was obtained and reported by Adweek on March 3, 2026. The brand safety infrastructure it describes was built across 2022 to 2025, following the Musk acquisition and the subsequent advertiser exodus.

Where: The playbook applies globally across X's advertising surfaces, with certain products - particularly Vertical Video Pre-Bid inventory vetting through IAS - currently limited to the United States market.

Why: X's total 2025 advertising revenues reached an estimated $1.25 billion, compared to $2.43 billion in 2021 before the Musk acquisition. The playbook is part of a broader effort to rebuild advertiser confidence that was damaged by post-acquisition content moderation changes, high-profile safety incidents, regulatory fines, and concern about ad adjacency to controversial content. By positioning Grok as a brand safety infrastructure tool - rather than solely a consumer chatbot - X is attempting to convert its AI investment into a commercial argument for advertiser return.

Share this article
The link has been copied!