Gemini names brands four times more than it links to them

Semrush published a how-to piece on tracking brand visibility inside Google Gemini. The piece includes a number from Semrush's own Ghost Citations study. Gemini names a brand in 83.7 percent of answers that mention one. It cites the brand's site in only 21.4 percent of those answers. ChatGPT, Semrush says, does the opposite: it cites more often than it mentions.

The piece was written by Carlos Silva and published on 3 November. Semrush sells an AI Visibility Toolkit, so the piece is promotional. The data behind it is Semrush's own, not an independent audit.

The six measurements Semrush uses

Semrush lists six things to measure in Gemini. The first is brand mentions: how often the model names your brand. The second is citations: how often it links to your site as a source. The third is an overall AI visibility score. The fourth is share of voice against the competitors you name. The fifth is sentiment, in three bands of positive, neutral and negative. The sixth is average position, meaning where your brand lands in a list.

The first two are the ones Semrush says most teams get wrong. A team that only counts citation links misses most of Gemini's brand exposure.

One worked example

Semrush uses Asana to show the gap. Over the window Semrush measured, Gemini named Asana 6,300 times and cited asana.com 553 times. That is a ratio of about eleven mentions for each citation. Semrush says Asana's share of voice in its category is eight percent on Gemini, against eighteen percent for Trello. Asana's sentiment rating is 91 percent positive. Gemini carries 24 percent of Asana's total AI visibility across the models Semrush tracks.

Every one of these numbers is Semrush's. They are measured against the prompts Semrush runs, not against Google's own analytics.

How to track it without buying anything

Semrush's own manual method takes a spreadsheet. Pick 20 to 50 prompts across four kinds of buyer intent: branded, bottom-of-funnel, feature-based and educational. Run each prompt in Gemini. Log four columns per prompt. Was the brand named. Was it cited with a link. How was it framed. Which rivals appeared beside it. Rerun the same prompts each month.

This is the part of the piece that does not need Semrush's tool. The rest of the piece describes what the toolkit adds: daily data across 317 million prompts, source citation mapping, and competitor benchmarking.

What this means for a GEO team

The honest reading is that a dashboard built only on citation data is counting Gemini wrong. The reader hears the brand mention even when the model gives no link. The team's job is to track both, and to know that Gemini and ChatGPT behave in opposite directions.

A caveat: Semrush is a vendor and has a stake in making brand owners worry about this. No independent party has published the same 83.7 and 21.4 figures. Any team planning a decision around this should rerun the manual method first. Do it on your own prompts before paying for a toolkit.

Source

• Carlos Silva, "How to track your Gemini visibility with and without Semrush", Semrush