When someone asks an AI assistant "what's the best project management tool for a small agency?", the brands in that answer get a shot at the sale and the rest don't. This guide explains how to measure that systematically: choosing prompts, collecting answers from several AI models, and turning them into a mention rate and share of voice you can track over time.

Disclosure: ManiGinaM, the company behind aileapers, built this Actor. ManiGinaM built and sells the AI Brand Visibility Actor on Apify Store. We earn money every time it runs. It is one way to do this; if you'd rather have a managed dashboard, see our comparison of the best AI brand visibility trackers, which covers subscription tools such as Otterly.AI, Peec AI, and Semrush.

What You Are Measuring

Three numbers do most of the work:

  • Mention rate: the share of AI answers that name your brand at all.
  • Position: when you are named, where you appear relative to competitors (first, second, and so on).
  • Share of voice: your share of all brand mentions across the answers, compared with the competitors you track.

A fourth signal, citations, shows which web pages the AI used as sources. That tells you which sites to earn coverage on. Perplexity returns citations with its answers; other models often answer from training data without sources unless you use a search-enabled model.

Before You Start

You need:

  1. A list of buyer questions (prompts). Write the questions real customers ask before they buy, such as "best CRM for a 10-person sales team" or "alternatives to [competitor]". Ten to twenty is a good start. Avoid prompts that include your own brand name; they tell you little.
  2. Your brand, aliases, and domain. Include common misspellings or product names that should count as you.
  3. Three to ten competitors to benchmark against.
  4. API keys for the AI providers you want to check. The method below uses the providers' developer APIs, which are billed separately from ChatGPT Plus or other chat subscriptions. Set a monthly spending limit in each provider's dashboard before you begin. Keys come from OpenAI, Perplexity, Google AI Studio (Gemini), and Anthropic (Claude).
  5. An Apify account. Apify's free plan includes $5 of monthly usage, according to Apify's pricing page.

Step 1: Estimate the Cost

The AI Brand Visibility Actor charges per answer: $25 per 1,000 answers ($0.025 each), with lower rates on Apify's higher discount tiers. Failed answers and the summary record are free. You pay the AI providers separately for the tokens each answer uses.

The number of answers is prompts × repeats × providers. For example:

  • 10 prompts × 1 repeat × 4 providers = 40 answers = $1.00 in Actor charges.
  • 20 prompts × 2 repeats × 3 providers = 120 answers = $3.00 in Actor charges.
  • Run that second setup weekly (about 4.3 times a month) and Actor charges come to roughly $13 a month, plus API costs.

If you plan to track many prompts every day, compare against a flat-fee dashboard; at high volume a subscription can work out cheaper per answer.

Step 2: Set Up the Input

Open the Actor in Apify Console and fill in the form. The listing documents these main fields:

Field What to enter
brand Your brand as people write it (required)
brandAliases Up to 10 other names that count as your brand
brandDomain Your website, used to check whether AI answers cite it
competitors Up to 25 competitor names
prompts Up to 100 buyer questions
repeats 1–5; how many times to ask each prompt per provider
Provider keys Paste each key into its secret field; leave out providers you don't use

Semrush pricing last checked Nov 2, 2026. See its price history.

Here is an input in the format the listing documents, for a fictional coffee-subscription brand (all names are made up):

json { "brand": "Northbean", "brandAliases": ["Northbean Coffee"], "brandDomain": "northbean.example", "competitors": ["Roastly", "BeanCrate", "MugClub"], "prompts": [ "What is the best coffee subscription for beginners?", "Which coffee subscriptions let you choose roast level?" ], "repeats": 2, "perplexityApiKey": "pplx-...", "openaiApiKey": "sk-..." }

That asks 2 prompts, 2 times each, of 2 providers: 8 answers, or $0.20 in Actor charges.

According to the listing, keys entered in the secret fields are stored encrypted by Apify, sent only to their own provider, and kept out of logs and output. Default models are inexpensive ones (such as gpt-4.1-mini for OpenAI and sonar for Perplexity); you can choose others in the input.

Tip: if your brand name is also a common word, add a more specific alias (for example "Northbean Coffee") and spot-check answers, because matching is done by name.

Step 3: Run It and Read the Output

Click Start. When the run finishes, the Output tab holds one record per answer and a final summary record.

Per the listing's documentation, each answer record includes the provider and model, the prompt, the full answer text, brandMentioned, brandPosition, competitorsMentioned, cited sources, and whether your domain was cited. The summary record includes, for your brand and each competitor, mentionRate, shareOfVoice, averagePosition, and firstMentions, plus a per-provider breakdown.

How to read it:

  • Low mention rate everywhere: AI models don't associate you with the category yet. Work on being covered by the sources they cite and on clear category language on your own site.
  • Mentioned but rarely first: compare the answers where a competitor leads; look at what the AI says about them.
  • Big differences between providers: focus on the providers your audience uses.
  • Cited domains: in the Perplexity answers, note which publications and review sites appear repeatedly. Those are your outreach list.

Export to CSV or Excel from the Output tab if you want to work in a spreadsheet.

Step 4: Track It Over Time

One run is a snapshot. AI answers shift with model updates and new web content, so the value is in the trend.

  1. In Apify Console, save your input as a task.
  2. Add a schedule to the task (weekly is a sensible default).
  3. Optionally add an integration (Google Sheets, Slack, Zapier, Make, or a webhook) so each run's results land where your team looks.
  4. Compare the summary records week to week. Keep the prompt list stable so the numbers are comparable; add new prompts as a separate set.

Limitations to Keep in Mind

These caveats apply to this method and, in various forms, to every AI visibility tool:

  • APIs aren't the chat apps. The Actor asks the API models behind the assistants, not ChatGPT, Gemini, or Claude as a signed-in person sees them. The APIs don't know who is asking or where, so answers can differ.
  • Answers are random-ish. The same prompt can produce different answers. Use repeats of 2–3 for important prompts and read trends, not single runs.
  • Citations mostly come from Perplexity. The default OpenAI, Gemini, and Anthropic models answer from training data and often give no sources.
  • It's new. The Actor was listed in November 2026, has no user reviews yet, and its listing notes it has not yet been run against live provider keys. Report problems in the Actor's Issues tab on Apify.
  • No dashboard. You get data, not charts. If you want a managed interface, a subscription tool may suit you better.

FAQ

Q: Can I track my brand in ChatGPT without paying for a tool?

A: You can ask ChatGPT your buyer questions manually and note the results in a spreadsheet. It works for a handful of prompts, but it's slow to repeat and hard to compare across models and weeks, which is what tracking tools automate.

Q: Does this method use my ChatGPT Plus subscription?

A: No. It uses each provider's developer API, which needs an API key and is billed separately from chat subscriptions.

Q: How many prompts should I track?

A: Start with 10 to 20 questions your buyers actually ask, covering your main category, key use cases, and "alternatives to" searches. Add more once you know which ones move.

Q: How much does it cost to run?

A: The Actor charges $0.025 per answer, and you pay the AI providers for your own API usage. For example, 20 prompts asked twice of three providers is 120 answers, or $3.00 in Actor charges per run.

Q: Is my API key safe?

A: According to the Actor's listing, keys go into encrypted secret input fields, are sent only to their own provider, and are never written to logs or output. Setting a spending limit with each provider adds another safeguard.

Sources

Related articles: