Citation or mention: what really boosts a brand in ChatGPT and Perplexity answers

A Russian author published results in July 2026 on Habr from months of monitoring ChatGPT, Perplexity, Gemini, and Yandex's "Alisa" responses to brand-related queries. Analysis of data from Ahrefs, Profound, and personal measurements showed that textual mentions of a brand in the body of an AI response occur significantly more often than link citations, and it is these mentions that determine actual AI visibility.

Quote and Mention — These Are Different Metrics

In the practice of AI marketing, a persistent confusion has developed, carried over from classical SEO: "appearing in an AI response" is considered equivalent to "getting a link in sources." In reality, these are two fundamentally different outcomes, measured differently and affecting brand perception differently.

Citation — a link to a page as a source of information. It appears in footnotes or in a "Sources" block next to the model's response. Marketers habitually count it as a sign of success: "there's a link — we made it into the AI output."

Mention — when the model names a brand or product directly in the response text, without a single link. The user sees a recommendation like "Service A and Service B would work," but there are no footnotes.

According to public research from Ahrefs and Profound, textual mentions occur significantly more often than link citations. The author's own monitoring confirmed this pattern in the Russian market as well, including Yandex's "Alisa."

Why

Recommendation Answers Matter More Than Informational Ones

AI tools respond to two fundamentally different types of queries. Informational — "what is RAG," "how does a vector database work" — often receive responses with source links: models cite documentation, articles, Wikipedia. Recommendation queries — "which service to choose for analytics," "what works for small business" — the model usually answers with text without a single footnote.

It is the recommendation scenario that is commercially more important: this is where the decision to try or buy a product forms. A brand either appears in the recommendation — or it isn't there, and no citations in other contexts compensate for that.

"There's a link — victory, no links — failure" — this is SEO thinking,

which doesn't work in an AI context.

Additionally, the author noted differences between platforms: ChatGPT, Perplexity, Gemini, and "Alisa" mention brands differently in responses to identical queries — presence in one tool doesn't guarantee it in others. In the Russian market, "Alisa" behaves noticeably differently from Western platforms.

What to Measure During AI Optimization

The author concludes that working with AI visibility (in terms of AEO and GEO) requires a separate toolkit — not SEO positions and backlink mass, but systematic monitoring of brand mentions in response text. Metrics that really matter:

• Frequency of brand mentions in recommendation responses to thematic queries
• Coverage of queries where the brand is named in text (regardless of link presence)
• Position in the listing — mentioned first or at the tail of the list
• Context and wording — how exactly AI describes the product or company
• Platform coverage — how stable the brand's presence is across ChatGPT, Perplexity, Gemini, and "Alisa"

This approach fundamentally differs from classical SEO monitoring: you need to track not positions in search results, but what exactly AI says about the brand in responses to thematic queries.

What This Means

For AI promotion specialists, this is a signal of the need to change metrics. Optimization should be aimed not only at link citations, but primarily at content, reputation signals, and authority in the niche — the things that prompt models to name the brand in recommendations. It is this visibility — the kind a user sees at the moment of deciding to choose a tool or vendor — that truly matters.