To track whether AI assistants recommend your competitors, build a fixed set of 15–30 buyer questions, run each one several times per cycle in ChatGPT, Google AI Mode, Gemini, Copilot and Perplexity, and measure your share of voice: the percentage of runs in which you're named, compared with each competitor. Add the free first-party reports from Google Search Console, Bing Webmaster Tools and GA4, and review the trend monthly. A single check tells you almost nothing, because AI answers change from run to run.

This post covers setting up the tracking. Once you know who's winning, find out why AI recommends your competitors and close the gap.

Why One Check Isn't Enough

  • Answers rarely repeat. In SparkToro's research (2,961 runs across 12 prompts, each repeated 60–100 times per platform), ChatGPT and Google's AI returned the same list of brands less than 1% of the time, and the same order less than 0.1% of the time (Search Engine Journal, Jan 2026). How often a brand appeared was far more stable than its position.
  • Local answers name a few businesses, and the names shift. ChatGPT named 4.1 local businesses per answer on average, AI Mode 3.5 and AI Overviews 2.5. Repeat runs of the same question overlapped only 20–33% (BrightLocal, Sep 2026).
  • Each engine is different. The engines overlapped only about 20% with each other in BrightLocal's study, so winning in ChatGPT doesn't mean you're winning in Google.

So track how often you appear, not where you "rank."

Step 1: Build Your Prompt Set

Write 15–30 questions your customers actually ask, across three types:

  • Discovery: "best emergency plumber in Brampton," "top fractional CFO for SaaS in Toronto."
  • Comparison: "[Your business] vs [competitor]," "which is better for [need], X or Y?"
  • Problem-led: "who can fix a furnace that won't ignite in Mississauga tonight?"

Keep the wording fixed from cycle to cycle so the results are comparable. Pick 3–5 competitors to track by name.

Step 2: Set a Sampling Routine

  • Runs: ask each question several times per cycle (three to five is a practical start). No study has set a definitive number; the SparkToro authors list it as an open question.
  • Cadence: every two to four weeks.
  • Conditions: use a clean, logged-out or fresh session and a consistent location. For local businesses, location changes results a lot: two points in the same town overlapped only 36–47% in BrightLocal's study.
  • What to log: date, engine, question, run number, every business named, and the sources cited.

Step 3: Calculate Share of Voice

Share of voice = runs where a business is named ÷ total runs, per engine. Calculate it for you and each competitor. If you're named in 9 of 30 ChatGPT runs and your main competitor in 21, your share is 30% to their 70%. Watch the trend over months, not single cycles.

Step 4: Add the Free First-Party Reports

ReportWhat it showsLimits
Search Console: Generative AI reportImpressions from AI Overviews and AI Mode (Google, Jun 2026)Impressions only; your own site only
Bing Webmaster Tools: AI PerformanceCopilot and Bing AI citations, cited pages, and "grounding queries" (Bing, Feb 2026)Public preview; no competitor data
GA4: AI Assistant channelVisits from ChatGPT, Gemini, Claude, Copilot and othersClicks from Google's AI Overviews count as Organic Search; some AI app clicks arrive as Direct
ChatGPT's utm_source=chatgpt.comChatGPT adds this tag to links it sends (OpenAI)Only on clicks

None of these show competitor data. Your prompt set is the only competitive view, which is why you need both.

Step 5: Manual, Tool or Automated?

  • Manual (spreadsheet): fine for 15–20 questions on three engines. A few hours a month.
  • AI visibility tracking tools: several SEO and local platforms now track prompts across ChatGPT, Gemini, AI Mode and Perplexity and report share of voice and cited sources. They save time and usually let you set a location.
  • Automated (n8n or Python): run your prompt set on a schedule through the official APIs (for example, OpenAI's API with web search, or Perplexity's Sonar API) and log results to a sheet or database. API answers aren't identical to what a logged-in user sees in the ChatGPT app, so treat them as an approximation. Don't scrape the ChatGPT app or Google's results: OpenAI's terms prohibit extracting output outside its API (OpenAI), and Google's terms prohibit automated access that ignores its machine-readable instructions (Google).

For the automation side, see continuous SEO auditing with n8n.

What to Track, and How Often

What to trackWhereHow oftenWhy
Share of voice vs 3–5 competitors on your prompt setSpreadsheet, n8n, or a tracking toolSeveral runs per question, every 2–4 weeksOnly a percentage shows the trend
Results by engineSame prompt set, tagged by engineEach cycleEngines overlap only about 20%
Results by location (local businesses)Location-set sessions or a tool with location settingsMonthlyResults change across a city
Google AI impressionsSearch Console Generative AI reportWeeklyFirst-party Google data
Copilot citations and grounding queriesBing Webmaster Tools AI PerformanceWeeklyShows which of your pages get cited
AI referral visits and leadsGA4 AI Assistant channel; "How did you hear about us?"MonthlyTies visibility to revenue
Sources cited for competitorsYour log or a tool's source reportMonthlyFeeds the competitor analysis

Rankr Client Example

A Toronto plumbing company used a Rankr AI audit to find schema errors and gaps in its listings and citations. Within 45 days, ChatGPT began listing them as a top option for their core plumbing questions. Tracking share of voice across repeated runs is how we confirm a change like that is real and not a one-off answer.

Frequently Asked Questions

How do I check whether ChatGPT recommends my competitors instead of me?

Write 15–30 questions your customers ask, run each several times per cycle in ChatGPT and the other assistants, and log every business named. Because answers rarely repeat, report how often each business appears rather than a single result.

How many times should I run each prompt?

No study has set a definitive number; the SparkToro researchers call it an open question. Since repeat local runs overlapped only 20–33% in BrightLocal's study, run each question several times per cycle and judge the trend over months.

What metric should I track?

Share of voice: the percentage of runs in which your business is named, compared with each competitor, per engine. Position isn't meaningful when answers reshuffle on every run.

Does Google or Bing show AI visibility data for free?

Yes, for your own site. Search Console's Generative AI report shows impressions from AI Overviews and AI Mode, and Bing Webmaster Tools' AI Performance report shows Copilot citations and the queries behind them. Neither shows competitors.

Can I see ChatGPT traffic in Google Analytics?

Yes. GA4 has an AI Assistant channel for visits from ChatGPT, Gemini, Claude and others, and ChatGPT tags links with utm_source=chatgpt.com. Clicks from Google's AI Overviews still count as Organic Search.

Which AI engines should a local business track?

ChatGPT and Google's AI Mode and AI Overviews first, then Gemini, Copilot and Perplexity. 45% of consumers have used AI tools for local recommendations, and the engines overlap so little that each needs tracking separately.

Get a Baseline

The $99 AI SEO Audit gives you a starting share-of-voice baseline: which businesses ChatGPT, Gemini and Google AI name for your services, which sources they cite, and your AI visibility score, with a 90-day roadmap. Or book a strategy call.

Related reading: Why AI recommends your competitors (and how to close the gap) · Why your AI search traffic is under 1% · How to get AI assistants to recommend your business