AI share of voice index · updated 2026-09-23
A public, monthly measurement, category by category. We ask both assistants the twelve questions a buyer types when choosing a vendor, then count who gets named and which websites the assistants read to decide. No vendor is involved; the data is CC BY, reuse it with a link.
6 categories · 168 live answers so far · new categories added weekly, existing ones re-run monthly.
The share of answers naming each vendor, split by assistant; the largest disagreements between ChatGPT and Perplexity; and the source sites the answers were built from. The recurring finding across categories is that a small number of roundup and review pages decide who gets recommended, and they are rarely the vendor's own site.
Twelve buyer-intent prompts (best, top options this year, for a small business, most affordable, for a team of ten, what professionals recommend, compare the leaders, what to look for, best reviews, alternative to the most popular, best for startups, leading providers) are sent live to ChatGPT (gpt-5, web search on) and Perplexity (sonar) through the DataForSEO LLM Responses API. Each answer is checked for every vendor name with aliases and old names merged; source links are counted once per domain per answer. Assistants are non-deterministic, so single-run shares carry a few points of noise; the monthly re-runs are what make the trend meaningful. Full method on each category page.
The free AI visibility audit runs the same twelve questions for any brand in any category and sends you the report. Vendors on any page can have their per-question breakdown free by writing to audits@agentbuilt.dev.