Is your brand in the answer? Tracking four assistants for $0.08 a check
More buyers ask an assistant before they search. Visibility tools charge hundreds a month to tell you whether the assistants name you. This is the way we know to measure it yourself: ask the four assistants your buyers' questions on a schedule, and keep a row of facts per answer.
What you get
One row per prompt and engine: brand_in_answer and brand_first_match, competitor_in_answer and competitor_first_match, citations_count, cited_domains and cited_urls, whether a page on your domain is cited (target_cited), the pages the assistant read while searching (retrieved_domains) and the search_queries_used it ran. The answers come live from ChatGPT, Gemini, Claude and Perplexity with web search on, through the DataForSEO AI Optimization API; the actor keeps facts about each answer and never stores the text.
- prompt
- What are the best running shoes for flat feet?
- chatgpt
- brand named (Brooks) · first competitor ASICS · 9 citations · bmcsports.ie, theflatfeet.com, paceandpavement.com
- gemini
- brand named (Brooks) · first competitor ASICS · 30 citations · runrepeat.com, alastairrunning.com, runweekly.com
- claude
- brand named (Brooks) · first competitor ASICS · 9 citations · jdsports.com, roadrunnersports.com, fleetfeet.com
- perplexity
- brand named (Brooks) · first competitor ASICS · 10 citations · rtings.com, runrepeat.com, asics.com · brooksrunning.com cited
The cost
| Item | Price | 20 prompts × 4 engines, weekly |
|---|---|---|
| Checks | $0.08 per prompt and engine | $6.40 a week |
| First 4 rows of a run | free | −$0.32 |
| Total | about $6 a week, $26 a month |
Billing happens on your Apify account. One prompt across all four assistants costs nothing, so a first look is free.
Steps
- Open LLM Brand Visibility on Apify and sign in or create a free Apify account.
- Write 10 to 30 prompts the way a buyer would ask an assistant, one per line: "best CRM for a small agency", "is X worth the money", "which running shoes for flat feet".
- Put your names in brandTerms, rivals in competitorTerms and your site in targetDomain. Pick a country for the assistants' web search.
- Keep all four engines on, or switch any off. Press Start; each prompt takes a few seconds per engine.
- Schedule the same run weekly. Chart the share of rows with brand_in_answer true per engine, and read cited_domains to see which pages the assistants trust; those are the pages to earn mentions on.
- Download CSV, Excel or JSON, or call it as an MCP tool through mcp.apify.com from your own agent.
Things worth knowing
- Assistants answer differently every time, as they do for real users. Trends over weeks are the signal; one run is a sample.
- Models used: ChatGPT gpt-4o-mini and Claude claude-haiku-4-5 with web search forced, Gemini gemini-2.5-flash with Google grounding, Perplexity sonar. Gemini returns redirect links for its citations, so its cited_domains come from the citation labels.
- Brand and competitor matching is whole-word and case-insensitive, so "Brooks" does not match "brookside".
- If the supplier is unreachable the run waits and retries for several minutes; if it still gets nothing, it fails with the reason and you are not charged.
- You can bring your own DataForSEO credentials in the two optional inputs if you prefer the supplier charge on your own account; the row price stays the same.
Run it on Apify First 4 checks free.