Free audit
Guide

AI rank tracker: why there is no rank in AI answers, and what you can track

Short answer: an AI rank tracker cannot track a rank, because AI answers don't have one. Ask ChatGPT the same question twice and you usually get a different list. In a SparkToro study (January 2026, 2,961 responses) there was "a <1 in 100 chance" of getting the same list twice from ChatGPT or Google's AI, and "more like 1 in 1,000" for the same order (SparkToro). What you can measure honestly: in how many answers you were named, named first or not named, who was named instead, and which sources were cited, across repeated asks on each assistant.

Why "rank" does not exist in AI answers

A Google Maps search has a ranked list. You can be 4th for "plumber Austin" from a given point, and you will usually still be 4th an hour later. An AI answer is written fresh each time. The assistant searches, reads some pages and writes a reply. The number of businesses it names, their order and the sources it cites can all change.

OpenAI says plainly that "Placement is not guaranteed" in ChatGPT search (OpenAI Help Center, checked November 11, 2026). Rand Fishkin, who ran the SparkToro study, put it more bluntly: "any tool that gives a 'ranking position in AI' is full of baloney." His view is that "visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric."

The engines also disagree with each other. In a small June 2026 test of 36 answers for UK trades, ChatGPT, Gemini and Perplexity never all named the same top business, and two of the three agreed in only 1 of 12 questions (Whito). A single "AI rank" averaged across assistants would hide that.

What you can measure

Measure What it means How much to trust it
Named Share of answers that mention your business The most useful number, if the sample is big enough
Named first Share of answers where you came before every other business Useful, but order is less stable than inclusion
Not named Answers that named others and not you Read these. They show who wins the question
Competitors named Which businesses appear instead, and how often Tells you who to compare yourself with
Sources cited The websites each answer cites The most actionable: these are places you can get listed
Average spot Your average place in the answers that named you The weakest number. Order changes most

Each number is only as good as the answers behind it. A good tracker shows the count next to every percentage ("named in 9 of 36 answers"), not a bare score.

The sample-size problem, in numbers

This is a worked example with made-up counts, to show how much one answer moves the result.

Answers checked You were named in Share named One answer changes it by Plausible range (95%)
12 3 25% 8.3 points about 9% to 53%
36 9 25% 2.8 points about 14% to 41%
120 30 25% 0.8 points about 18% to 33%

The ranges use a standard method for proportions (the Wilson interval). They assume each answer is an independent draw, which real answers are not quite, so read them as rough. The lesson holds: with 12 answers, going from 3 to 5 mentions can be noise. Over several months with the same questions, a steady rise is more believable than any one jump.

What changes the result besides you

Factor What the vendor or a study says
Location ChatGPT "may use an approximate location based on your IP address to provide relevant local results" (OpenAI). Local Falcon reports that proximity "is a less predictable ranking factor in AI results than in traditional local rankings" (Local Falcon, October 2026)
Memory and account OpenAI says that if memory is on, ChatGPT "may use relevant saved memories when rewriting a search query"
Personalization In Ask Maps, Google says "Your results are personalized based on things like places you've searched for or saved in Maps" (Google)
Wording Customers phrase the same need very differently. SparkToro measured an average semantic similarity of 0.081 between human-written prompts
Model and date Vendors change models and search providers over time. Results from different months are not strictly comparable

So no tracker sees exactly what your customer saw. It asks from outside and gives you a sample.

How to track honestly

  1. Fix the questions. Use the same 3 to 5 customer questions every time. Changing the wording changes the answers. Our list of AI prompts customers ask local businesses is a starting point.
  2. Repeat each question. At least 3 times on each assistant, in a clean session.
  3. Track each assistant separately. ChatGPT, Gemini, Perplexity and Google AI Mode give different answers. Do not blend them into one number without also showing each.
  4. Count, don't rank. Report "named in X of Y answers" and "named first in X of Y".
  5. Save the full answers. A percentage without the answer text cannot be checked.
  6. Log the sources. They are the part you can act on.
  7. Date everything. Note the day, location and assistant for each run.
  8. Compare over months. One check is a snapshot. Three or more monthly checks show a trend.

Questions to ask any AI tracking tool

  • Which assistants does it check, and which Google surface: AI Mode or AI Overviews? They are different products.
  • How many times does it ask each question?
  • Does it show the number of answers behind each percentage?
  • Can you read the full answers?
  • Does it list the cited sources and tell you whether you are on them?
  • Does it claim to know why an assistant chose a business? If so, ask for the vendor documentation. None of the vendors publishes it.

For how the brand-level tools compare on these points, see AI visibility trackers for local businesses.

Where Pack Leader fits

AI Search asks ChatGPT, Gemini, Perplexity and Google AI Mode your questions 3 times on each assistant, so one question gives 12 answers. It shows Mentioned, Named first, Average spot and Sites used, each with the number of answers behind it, split by assistant. You can read every answer in full, and the Sources view marks sites that list a competitor where we did not find you. Saved questions can run every month, so you compare like with like.

What it does not do: it does not check Google AI Overviews, it does not see what an assistant tells one particular person, and it does not know how any assistant chooses. Treat Average spot as the least stable number. For a first look, the AI visibility checker explains the quick check, and how to run and read a quick AI check walks through the screens. To act on the results, follow the local GEO playbook.

Frequently asked questions

Is there a ChatGPT rank tracker?

Tools can ask ChatGPT your questions and record the answers, but there is no stable rank to track. The same question gives a different list most of the time. What a tool can honestly report is how often you were named across many answers, and which sites were cited.

How many times should I ask the same question?

More than once. Three asks per assistant is a minimum to see variation. SparkToro suggests averaging over many more runs when you need a precise number. Small samples give wide error bars, so compare trends, not single checks.

What is AI search visibility?

The share of AI answers, for questions your customers ask, that name your business. It is a count from a sample of answers on a given day, not a position and not a score that causes anything.

Why does my AI result change every week?

AI answers vary by run, by wording, by assistant, by location and over time. A change from one check to the next can be noise. Look at the trend over several checks with the same questions and the same number of runs.

Does an AI visibility tool see what ChatGPT tells my customers?

No tool sees what an assistant tells one particular person. Answers can depend on location, memory and account. A tracker asks from outside and shows a sample.

Keep reading