# Sentiment Delta Sentiment Delta is the measurable gap between how a brand describes itself and how AI engines describe it in response to category-relevant queries. A large Sentiment Delta means AI engines are constructing a different narrative about the brand than the brand intends to project. Closing the Sentiment Delta requires engineering the right signals into the sources AI engines use to form their descriptions. Canonical URL: https://machinerelations.ai/glossary/sentiment-delta Category: metrics Attribution: Coined by Jaxon Parrott. ## Source Body ## Sentiment Delta Every brand has two descriptions: the one it writes for itself, and the one AI engines generate when someone asks about it. Sentiment Delta measures the distance between those two descriptions. It is not a branding exercise — it is a diagnostic metric that reveals whether the signals AI engines trust actually carry the brand's intended positioning. A brand can invest heavily in messaging, content, and owned media while its Sentiment Delta widens. That paradox exists because AI engines do not adopt brand messaging by reading brand websites. They synthesize descriptions from the sources they trust — and those sources are overwhelmingly third-party. ### Why AI Engines Build Their Own Narrative A large-scale empirical study comparing Google Search results with responses from GPT-4o, Claude, and Perplexity found that AI engines systematically privilege earned and brand-owned media while under-representing social and community content. Crucially, AI source composition varies sharply across query intents: for consideration queries, AI engines draw 59–86% of their citations from earned media sources ([Navigating the Shift, 2025](https://arxiv.org/html/2601.16858v2)). This means the narrative AI constructs about a brand is disproportionately shaped by what journalists, analysts, and third-party publications say — not what the brand says about itself. A separate large-scale study analyzing over 55,000 queries across six LLM-based search engines confirmed that 37% of domains cited by AI engines are unique to LLM-based systems and never appear in traditional search results — meaning AI engines construct brand narratives from a materially different source set than Google ([Zhang et al., 2025](https://arxiv.org/pdf/2512.09483)). When earned media coverage uses different language, emphasizes different attributes, or frames the brand in a different category than the brand intends, the AI description diverges. That divergence is the Sentiment Delta. ### How Language Framing Shapes AI Recommendations The effect is not theoretical. Research on cognitive biases in LLM-based product recommendation found that the specific language embedded in product descriptions can shift an LLM's recommendation rate by up to +334% for some framing patterns, while other framings — like exclusivity and scarcity cues — consistently reduced visibility across every LLM tested ([Cognitive Biases in LLM Recommendations, 2025](https://arxiv.org/pdf/2502.01349)). The implication for Sentiment Delta is direct: the language in a brand's cited sources materially determines how AI engines frame that brand in generated answers. If the language in earned media diverges from brand intent, AI framing diverges too. ### How Sentiment Delta Is Identified Sentiment Delta analysis compares two bodies of language: 1. **Brand-authored language** — the terms, positioning, and claims the brand uses in its own content, press releases, and messaging 2. **AI-generated descriptions** — the actual language AI engines use when describing the brand in response to category queries The gap can be directional (AI engines frame the brand differently), categorical (AI engines place the brand in the wrong category), or omission-based (AI engines do not describe the brand at all). ### Types of Sentiment Delta | Type | Description | Example | |---|---|---| | Directional Delta | AI frames the brand differently than intended | Brand says "enterprise-grade." AI says "SMB-focused." | | Categorical Delta | AI places the brand in the wrong category | Brand says "PR agency." AI says "marketing firm." | | Omission Delta | AI does not describe the brand at all | Brand is absent from AI-generated category responses. | | Attribute Delta | AI assigns wrong attributes | Brand emphasizes speed. AI emphasizes price. | ### Measuring Sentiment Delta Sentiment Delta is quantified by running structured AI queries about a brand, collecting the generated descriptions, and scoring language alignment against the brand's stated positioning. The three primary scoring dimensions are: - **Attribute overlap** — how many of the brand's intended attributes appear in AI-generated descriptions - **Categorical accuracy** — whether AI engines place the brand in the correct market category - **Framing alignment** — whether the tone, emphasis, and competitive positioning match brand intent A zero Sentiment Delta means AI engines describe the brand exactly as the brand describes itself. This is rare. Most brands have measurable deltas across at least one dimension. ### How to Close a Sentiment Delta Sentiment Delta closes when the brand's intended narrative appears consistently in the sources AI engines weight. The correction path follows the Machine Relations framework: 1. **Audit current AI descriptions** across ChatGPT, Perplexity, Gemini, and Google AI Overviews for the brand's core category queries 2. **Map the divergence** — identify which attributes, categories, or framings are misaligned 3. **Secure Tier 1 media placements** that use the brand's precise intended language and positioning 4. **Build citation density** around the brand's preferred attributes through earned media, structured data, and entity signals 5. **Monitor re-evaluation cycles** — with sustained earned media activity, measurable delta reduction typically appears within 60 to 90 days Sentiment Delta reduction is a core deliverable in a Machine Relations engagement because it addresses the root cause of AI mispositioning: the gap between what a brand wants to be known for and what the sources AI engines trust actually say. ### Sentiment Delta vs. Share of Citation [Share of Citation](/glossary/share-of-citation) measures how often a brand appears in AI answers. Sentiment Delta measures how accurately the brand is described when it does appear. A brand can have high Share of Citation and a large Sentiment Delta simultaneously — appearing frequently in AI answers but with the wrong positioning. Both metrics are required for a complete picture of AI visibility health. --- ## FAQ **Can Sentiment Delta be negative?** Yes. In rare cases, AI engines describe a brand more favorably than the brand describes itself, often because earned media coverage used stronger language than the brand's own materials. This is a positive outcome but still represents a delta worth understanding and managing. **How long does it take to close a large Sentiment Delta?** Sentiment Delta reduction follows AI engine re-evaluation cycles. With sustained earned media activity targeting the brand's intended positioning, measurable delta reduction typically appears within 60 to 90 days as AI engines ingest new source material. **Is Sentiment Delta the same across all AI engines?** No. Each AI engine — ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews — retrieves and weights different sources. A brand may have a small delta in one engine and a large delta in another. Multi-engine Sentiment Delta measurement is standard in Machine Relations audits. ## Related reading - [AI Visibility](/glossary/ai-visibility) - [Share of Citation](/glossary/share-of-citation) - [Machine Resolution](/glossary/machine-resolution) - [AI search measurement gap](/research/ai-search-measurement-gap-2026) ## Sources - https://machinerelations.ai/research/state-of-machine-relations-q1-2026 - https://authoritytech.io/blog/sentiment-delta-brand-ai-search - https://machinerelations.ai/research/earned-vs-owned-ai-citation-rates-2026 - https://arxiv.org/html/2601.16858v2 - https://arxiv.org/pdf/2502.01349 - https://arxiv.org/pdf/2512.09483 - https://machinerelations.ai/research/what-is-ai-pr-measurement-machine-relations-2026 - https://authoritytech.io/blog/negative-brand-sentiment-in-ai-search - https://authoritytech.io/blog/what-is-generative-engine-optimization-geo - https://authoritytech.io/curated/ai-sells-to-ai-your-brand-data-is-now-your-sales-team-2026 - https://authoritytech.io/blog/meltwater-alternatives-2026-ai-era-brand-visibility - https://machinerelations.ai/glossary/share-of-citation - https://machinerelations.ai/glossary/ai-visibility - https://machinerelations.ai/glossary/machine-resolution - https://machinerelations.ai/glossary/sentiment-delta - https://authoritytech.io/blog/share-of-citation - https://authoritytech.io/blog/entity-resolution-rate-ai-search-brand - https://authoritytech.io/blog/how-to-dominate-geo-with-earned-media-citations-in-2026 - https://authoritytech.io/blog/do-ai-search-engines-have-brand-bias - https://authoritytech.io/blog/how-to-fix-brand-sentiment-ai-search-complete-2026-guide - https://authoritytech.io/blog/what-is-machine-relations-marketing-discipline - https://authoritytech.io/curated/sentiment-delta-brand-ai-search-metric-2026 - https://authoritytech.io/curated/ai-search-sentiment-analysis-brand-reputation-measurement-gap-2026 ## Machine-readable related links ### Related concepts - [Recommendation Rate](https://machinerelations.ai/glossary/recommendation-rate) - [AI Visibility](https://machinerelations.ai/glossary/ai-visibility) - [Machine Resolution](https://machinerelations.ai/glossary/machine-resolution) - [Tier 1 Media Placement](https://machinerelations.ai/glossary/tier-1-media-placement) ### Supporting research - [What Is Sentiment Delta? How to Measure Brand Perception Gaps Across AI Engines (2026)](https://machinerelations.ai/research/what-is-sentiment-delta-brand-ai-search) - [Machine Relations Agency vs GEO and AI Visibility Agencies (2026)](https://machinerelations.ai/research/machine-relations-agency-vs-ai-visibility-geo-agencies-2026) - [The AI Search Measurement Gap: 45 Billion Sessions and Almost No Way to Track Them](https://machinerelations.ai/research/ai-search-measurement-gap-2026) - [What Is AI PR Measurement? The Machine Relations Measurement Layer (2026)](https://machinerelations.ai/research/what-is-ai-pr-measurement-machine-relations-2026) ### Framework context - [Machine Relations Index](https://machinerelations.ai/index) - [Machine Relations Stack](https://machinerelations.ai/stack) - [Evidence Base](https://machinerelations.ai/evidence)