Which brands LLMs name when buyers ask: the Q3 benchmark.

Every quarter we ask the same buying questions to the same assistants — and log who gets named, who gets cited, and what changed. This page carries the method in full; the data publishes as counts, per our counting rules.

The method

• Fixed question set. Buying-shaped questions (“which platforms should I look at for…”), phrased the way buyers phrase them, held identical between quarters.
• Fixed panel. The assistants buyers actually use: ChatGPT, Claude, Perplexity and Google's AI Overviews, each on default settings, clean sessions, no account history.
• Logged runs. Every answer is stored verbatim with a timestamp. We count mentions and citations; we do not interpret sentiment.
• Counts with denominators. “Named in 7 of 40 runs” — never “top of mind”.

What Q3 measures

Three questions drive this quarter's read:

1. Which brands do assistants name unprompted, per category?
2. Which sources do they cite for those names — and how often is that source a page the brand controls?
3. What moved since Q2 — and does the movement correlate with anything the brand published?

The Q3 tables land here

[ Q3 tables follow — publishing as counts with their denominators once the full run is cleared ]

We publish the tables when the quarter's runs are complete and re-checked. Data from a vendor measuring its own market is worth exactly as much as its method is transparent — hence the method first, in full, above the data.

What tends to move the needle

Across runs, three things correlate with getting named: a definition page assistants can quote safely, primary research they can cite, and consistency between what your pages claim and what third parties say about you. Vocabulary for the mechanics is in the AI-visibility hub.