A Semrush analysis of 283,215 citations and 76,493 brand mentions across 1,094 categories (January–June 2026) reveals the critical distinction that most GEO strategies miss: being cited in an AI answer and being recommended by one are not the same thing. In categories where a brand has shallow coverage, 50% of appearances are citations — passing references in an answer about a broader topic. In categories where a brand has genuine depth, 44% of appearances are direct brand mentions — the AI actively naming the brand as a relevant option.

The practical implication: citation volume is a vanity metric until it converts to brand mentions. Kevin Indig's analysis identifies the threshold — brands need repeat appearances (5/5 prompts sampled in a category, not just 1/5) before the pattern shifts from citation to recommendation. Finance and real estate show the strongest returns from broad coverage; legal and healthcare show negative mention associations even with high citation rates, likely reflecting AI caution around professional advice in liability-sensitive verticals.

Lily Ray's analysis of 11 sites hit by Google's January 2026 core update reveals a counterintuitive finding: ChatGPT citations dropped -27.8% on average — more severely than Google's own AI Mode (-23.8%) and far more than Perplexity, which showed resilience with just -2.9% average decline. The implication is structural: despite being a separate product, ChatGPT appears to rely heavily on Google's organic index for retrieval, making Google ranking health a dependency for ChatGPT visibility. Perplexity's independence from Google's index makes it a more stable citation channel after Google algorithm volatility.

"The citation-versus-mention distinction from Kevin Indig's analysis is the most practically useful finding I've seen in months. We've been telling clients that citations are a lagging indicator — they show up after you've built real category authority, not before. This data makes that concrete: in categories where brands have depth, 44% of AI appearances are direct brand mentions. In shallow categories, you're just background noise at 50% citation rate. The Lily Ray data adds another layer — if you're leaning on ChatGPT as your primary AI channel and your Google rankings slip, expect a -27.8% citation drop with almost no warning. Perplexity's independence from Google's index is undervalued as a diversification argument right now."