Authority signals

What are authority signals?

Authority signals are observable indicators practitioners use to assess whether a source is likely to be trusted, retrieved, cited, or ranked. The signal set has expanded meaningfully in the AI era: classical SEO weighted backlinks heavily, while AI-era practitioner discussion adds entity recognition, brand mentions across channels (web and video platforms), structured-data presence, author transparency, content freshness, and per-engine citation history. No AI engine has published a per-signal weighting, and the strongest available empirical evidence on what correlates with AI visibility is third-party telemetry, not vendor disclosure.

The most useful empirical anchor in 2026 is Ahrefs' December 2025 follow-up study of 75K brands across ChatGPT, Google AI Mode, and AI Overviews¹. Headline findings:

• YouTube mentions ~0.737 correlated most strongly with AI visibility across all three engines, beating branded web mentions
• Branded web mentions 0.656-0.709 depending on engine (highest for AI Mode, lowest for ChatGPT)
• Content volume ~0.194: almost no relationship; the "more pages = more authority" assumption is not supported by this dataset
• DR vs ChatGPT visibility ~0.266: classic backlink-authority metrics underperform expectations on the newest AI surface
• Ahrefs' own disclaimer: "the usual disclaimer applies: correlation isn't causation." Large brands have both more mentions and more AI visibility; brand strength may confound the relationship

No single signal is sufficient and no single dataset answers the question definitively. Authority emerges from signal stacking, and the stack has to be evaluated per engine: Bing AI Performance reports something different from what a Profound dashboard reports from what a Wikidata record demonstrates.

Status in 2026

Shifting and contested. Classical SEO authorities (Moz, Ahrefs) still publish backlink-weighted authority scores; AI-search analytics (Profound, Otterly.AI, AthenaHQ) emphasize citation history and entity recognition. The two camps describe overlapping but non-identical authority surfaces. Practitioners in 2026 typically track both (backlink authority for classic Google ranking, citation-share authority for AI surfaces) and treat them as related but distinct.

How to apply

Authority is stacked, not bought. Three concrete moves for a new or mid-stage brand:

• Audit your signal stack across four layers (this four-layer grouping is editorial to this glossary, useful for organizing the practitioner discussion; it is not a published industry-standard taxonomy and no engine documents a corresponding internal categorization): backlinks (classical SEO), entity recognition (Knowledge Graph, Wikidata where notability passes), author transparency (Person schema + visible bios + verifiable credentials), and content quality (sourced claims, accurate freshness metadata, DefinedTerm coverage where applicable). FAQPage markup still belongs in this layer where you have real Q&A content, but ship it for the underlying Q&A structure rather than SERP rich-result hopes; Google fully deprecated FAQ rich results for all sites on May 7, 2026.
• Prioritize entity-layer signals first for a new brand: backlink accumulation has a long natural latency; entity recognition can move faster because the signals (Organization schema, consistent sameAs profiles across LinkedIn / Crunchbase / GitHub / Wikidata, third-party brand mentions) are publisher-controlled. Practitioners commonly observe entity clarity producing recognition before equivalent backlink authority compounds, though specific timelines depend heavily on category and competition. The operational chain is the same as on the E-E-A-T page: schema → entity legibility → Knowledge Graph node → downstream authority-aligned heuristics become tractable; without entity recognition, the underlying signals have nowhere to land.
• Track per-engine authority separately: a brand may have strong Perplexity citation but weak ChatGPT presence even with identical content. Set per-engine goals rather than chasing a universal authority score. For Microsoft Copilot surfaces, the only vendor-native per-engine measurement currently available is Bing Webmaster Tools' AI Performance dashboard (public preview since February 10, 2026); for ChatGPT, Claude, Perplexity, and Google AI Overview, measurement still requires third-party tools (Profound, Otterly.AI, AthenaHQ, Brand Radar) or manual sampling.

What to skip: paid "authority score" dashboards in month 1. Moz DA and Ahrefs DR are dominated by backlink-graph signals; they may not fully explain AI citation visibility (the Ahrefs December 2025 data above puts DR at ~0.266 with ChatGPT visibility), but they remain useful for the classic Google ranking surface and are not strictly inferior to entity-recognition signals in a controlled comparison (no public study has done that comparison). Track raw observable signals (citation count, recognition state, mention growth across web and YouTube) alongside DA/DR rather than substituting one for the other, until your own baseline data tells you what predicts your citations.

How it relates to other concepts

• Operationalization layer for E-E-A-T in the practitioner reading. E-E-A-T is Google's human-rater framework; authority signals are the observable inputs practitioners optimize against. Google has stated E-E-A-T itself is not a direct ranking signal, so this is a practitioner mapping rather than an engine-internal mechanism.
• Plausible contributor to citation share outcomes alongside content relevance, retrieval access, query fit, and per-engine retrieval logic; authority is one input among several, not the deciding factor.
• Operationalized through the Knowledge Graph entity recognition chain: schema markup + consistent sameAs profiles → entity edges in the KG → downstream authority-aligned heuristics in Google's ranking systems become tractable. Without entity recognition, the underlying signals have nowhere to land in the algorithmic layer.
• One input among many to GEO programs. Content quality, retrieval access (crawler-bot allow rules + index inclusion), query fit, and source authority all influence whether content is cited; authority signals are necessary but not sufficient.