UK SEO Summit

Mentions vs. Citations: What AI Search Uses as Evidence

Talk context

Being named and being cited are different outcomes. Watch Fernando Angulo explain how to measure both in this edited UK SEO Summit 2026 keynote, with English captions.

Recorded at UK SEO Summit on August 26, 2026. This edited edition includes re-recorded voice passages for clarity and English captions. The talk draws on Semrush research to compare brand mentions and source citations across AI search platforms. The comparisons are observational; they do not establish that a particular content change causes more citations.

UK SEO Summit 2026 event archive →

Three evidence-led takeaways

  • Measure mentions and citations separately. A brand can appear in an answer without supplying its evidence, while a cited page can contribute evidence without a brand mention.
  • Compare platforms and question types before combining scores. ChatGPT, Gemini, AI Mode, and AI Overviews draw on different source mixes, so a single visibility number can hide important differences.
  • Build evidence around a decision journey. Connect owned research with independent coverage, industry sources, and community discussion, then track how each contributes to answers.

Detailed talk summary

If an AI answer names your brand but cites another source, what have you gained? Fernando opens the keynote with that measurement problem. A mention makes the brand visible in the answer. A citation identifies a source used as evidence. Those outcomes can overlap, but they answer different questions for a marketing team. Treating them as interchangeable can conceal whether the brand is entering consideration, contributing useful information, or doing both.

The platform comparisons make the distinction practical. Fernando examines how the overlap between mentioned brands and cited sources changes across AI Overviews, AI Mode, ChatGPT, and Gemini. He also discusses ghost citations: cases where a source receives a link but its brand does not receive a corresponding mention. The implication is to keep platform, source, and attribution visible in reporting instead of hiding them inside one universal AI visibility score.

The middle of the session turns from measurement to the questions people ask. A broad discovery prompt and a specific evaluation question create different information needs. Topic ownership therefore requires coverage across the decision journey, including comparisons, constraints, costs, and practical use cases. Producing more pages is not the same as supplying the evidence an answer needs. Teams should identify gaps in the questions they can answer clearly before expanding their publishing volume.

The final section connects that coverage with an evidence portfolio. Owned research, specialist publications, industry relationships, and community sources each play a role in how information is corroborated. Fernando outlines a first-month operating approach: identify relevant topics, review the available evidence, assign work, distribute useful material, and measure mentions and citations separately. These are practical recommendations drawn from observed patterns, not a promise that a particular action will produce a citation. The edited recording retains that distinction between an observed association and a causal result.

About Fernando Angulo

Fernando Angulo is Senior Market Research Manager at Semrush, an Adobe company.

Learn more about Fernando →

Need a keynote for your event?

Review the current talks, audience outcomes, and practical information for conference programmers.

Explore current speaking topics