Entity Resolution Rate

Entity resolution rate answers the question "can AI search correctly resolve this brand": it is the percentage of AI-engine queries in which a brand is identified, attributed, and represented as the entity it actually is, rather than confused with a competitor, a different product line, or omitted outright. When an AI system encounters a brand-relevant query, it runs a resolution process: cross-referencing names, descriptions, categories, founding details, and publication mentions to determine whether signals from across the web all refer to the same entity. Brands that pass this process confidently get cited. Brands that don't get omitted, hedged, or misrepresented.

The term was coined by Jaxon Parrott, founder of AuthorityTech, as part of the Machine Relations measurement framework. It is designed to replace share of voice as the primary brand metric for the AI search era, because raw mention volume is no longer the relevant signal. A brand can appear in thousands of documents and still fail entity resolution if those documents describe the brand inconsistently, belong to low-trust domains, or accumulate without forming a coherent entity profile that AI systems can anchor to.

What entity resolution rate measures

Entity resolution rate captures whether an AI system can confidently resolve a brand across queries, not just whether it mentions the brand at all. It is measured by running a representative set of brand-relevant queries against multiple AI engines and tracking how often the brand is correctly identified, accurately described, and cited by name.

High resolution rate means the AI consistently recognizes the brand, attributes claims correctly, and surfaces the brand in relevant answers. Low resolution rate means the AI either omits the brand, hedges with vague mentions, or surfaces it with incorrect attributes.

Harvard Business Review's March 2026 analysis of brand readiness for agentic AI documented how LLM data on brands was "often incomplete or incorrect" in ways companies only discovered after AI systems had already begun influencing buyer decisions. In one case, a major spirits brand found a popular AI model had miscategorized an affordable product as a prestige offering. The brand had not failed at marketing. It had failed at machine legibility.

The confidence threshold

AI engines operate with resolution confidence thresholds below which they will not surface a brand by name, even when the brand is genuinely relevant to the query. The design is intentional: a hallucinated brand recommendation is worse than an omission. Below roughly 60% resolution confidence, brands get passed over regardless of their actual relevance.

Research by Dong Liu and Sreyashi Nag (arXiv, February 2025) on query brand entity linking in e-commerce documented this directly: resolution fails most often when a brand's signals are inconsistent or when the gap between a brand's documented identity and its real identity is wide. The AI cannot resolve what it cannot reconcile.

The GEO-16 framework analysis by Kumar et al. (arXiv, September 2025) showed the parallel in citation quality: pages scoring above a structural quality threshold of 0.70 and hitting at least 12 quality pillars achieved a 78% cross-engine citation rate, with an odds ratio of 4.2 for quality as a predictor of citation. Below the threshold, pages were omitted even when directly relevant. The same selectivity governs entity resolution for brands.

Resolution confidence	What drives it	AI behavior
High (80%+)	Multiple high-DA sources; consistent signals; editorial Tier 1 coverage; Wikidata anchor	Cited accurately by name with correct specifics
Medium (60-80%)	Some editorial coverage; basic structured data; partial third-party corroboration	Cited in some contexts; inconsistent across engines
Low (<60%)	Sparse third-party coverage; contradictory descriptions; primarily owned-channel signals	Omitted or hedged even in directly relevant queries

Why earned media moves the rate

AI engines weight third-party editorial sources over brand-owned content when building entity confidence. A brand's own website, social profiles, and press releases carry low resolution weight because they are self-reported. Independent coverage in publications that AI systems have indexed and trust produces the corroboration needed to cross the confidence threshold.

This is why entity resolution and entity resolution rate are both downstream of earned media strategy. Earning placements in Tier 1 publications does more than drive referral traffic. Each placement is a corroboration node: an independent, high-trust source that confirms the brand's identity, attributes, and category. Multiple corroboration nodes pointing to the same entity description create the signal density that moves resolution confidence above the citation threshold.

Brands with high entity resolution rates have, almost without exception, built that rate through consistent earned media in publications AI engines treat as authoritative, not through owned content volume.

Brand entity contradictions across AI platforms

The failure mode that drives resolution rate down has a specific shape: contradiction, not absence. A brand rarely disappears from AI training and retrieval data entirely; it more often exists as several partial, conflicting descriptions that no engine can merge into one entity. A directory listing says one category, a press mention says another, an old acquisition rumor never corrected sits beside the current ownership structure, and a review site uses a product name the company retired two years ago. Each engine resolves the contradiction differently, which is why the same brand can be cited accurately on Perplexity and mischaracterized on Gemini in the same week: the underlying entity graph each platform built from public signals simply disagrees with itself.

Per the Machine Relations Index (release mri_score_v2.0+2026-09-30+f16ab66ee5f1, window 2026-05-10 to 2026-09-30), authoritytech.io ranks #8 of 529 cited sources on the "is it worth it" question shape in the AI Visibility & GEO category, a 10.24% cited-run rate across 8 distinct run dates — evidence that a single, consistently-described entity profile compounds into citation share on the exact question types where contradictory brand data usually costs a brand its place in the answer.

Managing brand entity resolution in AI search citations

Brand entity management for AI search citations is an ongoing maintenance discipline, not a one-time fix. The description that resolves cleanly today degrades as the brand changes products, leadership, or category positioning, and as new third-party sources publish their own (sometimes outdated) version of the brand. Three practices keep resolution rate from decaying:

• Audit for drift on a schedule. Run the brand's core facts (category, founding details, leadership, flagship product names) against its own structured data and against the most-cited third-party sources at least quarterly; correct the owned side first, then request corrections on corrigible third-party listings.
• Anchor with a single structured identity. A Wikidata entry or consistent Organization/Person schema, referenced the same way across every owned property, gives every platform one canonical node to reconcile external mentions against instead of rebuilding the entity from scratch each time.
• Treat every earned placement as an entity-definition event, not just a mention. A placement that repeats the same category language, founding facts, and product names as the brand's own structured data reinforces the entity graph; a placement that describes the brand loosely or inconsistently adds to the contradiction instead of resolving it.

Entity resolution rate vs. share of voice

Share of voice measures how often a brand appears relative to competitors. Entity resolution rate measures how accurately the brand is understood when it does appear. The two metrics diverge sharply in the AI search era.

A brand can win share of voice, accumulating mentions across low-trust or self-owned domains, while losing entity resolution rate if those mentions don't form a coherent, corroborated entity profile. Conversely, a brand with a smaller footprint can achieve a high resolution rate if its coverage is concentrated in trusted editorial sources that AI engines weigh heavily.

Entity resolution rate is the metric that actually predicts AI citation behavior. Share of voice predicts how often a brand name appears in text. In AI-mediated discovery, those are different outcomes.

Frequently asked questions

How is entity resolution rate calculated? Run a defined set of brand-relevant queries across target AI engines (ChatGPT, Perplexity, Gemini, Claude) and score each response: correctly identified, attributed, and represented counts as resolved. Misidentified, omitted, or hedged counts as failed. Rate = resolved responses / total queries. The query set should include informational, comparative, and recommendation-type prompts to capture resolution across contexts.

What is the minimum resolution rate a brand needs to appear in AI answers? Based on documented AI system behavior, brands operating below approximately 60% resolution confidence face consistent omission from AI-generated answers, even when directly relevant. Crossing into the 60-80% range produces inconsistent but improving citation behavior. Above 80% produces reliable, accurate citation across most engines.

Can entity resolution rate be improved without changing the product or positioning? Yes. The primary lever is earned media strategy, not product changes. Earning consistent placements in Tier 1 publications that AI engines treat as authoritative, and ensuring those placements describe the brand consistently using the same category terms, founding details, and product attributes, builds the corroboration density that moves resolution confidence. Structured data (schema.org, Wikidata entries) accelerates the process by giving AI engines a machine-readable anchor to cross-reference against editorial coverage.

How does entity resolution rate relate to share of citation? Share of citation measures how often a brand is cited relative to competitors across AI engines. Entity resolution rate is the prerequisite that determines whether the brand gets cited at all. A brand with a low resolution rate will have a low share of citation almost by definition. Improving resolution rate is the first-order fix; measuring share of citation then tracks whether the improvement is compounding across competitive queries.