The weekly LLM visibility tracker

AI answers shift every run. Nenn asks your questions on 3 AI engines each week and shows the trend.

Start 7-day trialTry the test run

Card needed. Cancel before day 7 and nothing is charged.

Visibility

Elektro Lindqvist · 3 AI engines · Hamburg · Week 45

Questions by topic
QuestionAI enginesMentioned
Electricians in Hamburg7 questions38.1%
Who is a good emergency electrician in Hamburg?English · Germany66.7%
Which electrician in Hamburg answers at night?English · Germany33.3%
Who installs a wallbox in Hamburg?English · Germany33.3%
Electrician near Altona with fair prices?English · Germany33.3%

Mentioned in

38.1%+9.5%

Position

#3
  1. Kestrow Elektrik47.6%+4.7%
  2. Funkenhof42.9%-4.7%
  3. Elektro Lindqvist38.1%+9.5%
  4. Strandvolt28.6%-4.7%
  5. Ohmhaus Nord23.8%-4.8%

What is an LLM visibility tracker?

An LLM visibility tracker is a tool that asks AI models your customers' questions on a schedule and records whether your company is named. Nenn (nenn.io) asks them again every week on ChatGPT, Perplexity and Google AI Overviews.

Why one check is not enough

One check can mislead. In our test, one question asked 10 times gave 9 different top-5 orders.

A tracker keeps the questions, the engines and the place the same each week. Then a change in the number comes from the answers, not from the method.

It counts how many answers name you, and your place when they do. A share of many answers is easier to trust than one answer.

What does LLM mean, in plain words?

An LLM (large language model) is the AI model behind ChatGPT, Perplexity and Google's AI answers.

Run 1

For a law office, the best website builder is A

  • A
  • C
  • G

Run 2

For a law office, the best website builder is C

  • A
  • C
  • G

Run 1 picks A, run 2 picks C for the same sentence.

How one word gets picked. Not real numbers.

  • One word at a time

    It writes an answer word by word, picking from several likely words.

  • A little chance in each pick

    Chance plays a small part, so two runs can pick different words.

  • Other pages on each run

    With web search on, it may read different pages each time.

So the same question can bring back a new list or a new #1.

Where the chance comes from

How much chance plays a part is a setting. In the OpenAI API it is called temperature.

Web search adds more change. In our test, 105 of 115 runs searched the web before answering, and the pages found differ from run to run.

Each run in our test also started a fresh session, with no chat before it.

Mr. Nenn, the Nenn mascot, thinks about how AI picks a word

How much do AI answers change from run to run?

More than most people expect. In our test, every buying question came back in at least 3 different top-5 orders.

One question, 10 runs, places 1 to 5
#1#2#3#4#5
Run 1ABCDE
Run 2CFGAB
Run 3ACBDH
Run 4GCAFB
Run 5ABCDH
Run 6AGCBD
Run 7CFADB
Run 8CABGD
Run 9CGBAD
Run 10ABCDE

9different orders in 10 runs

One question, 10 runs. Letters stand for brands. Our test, 01.11.2026, 10 runs.

  • Answers we read

    606 buying questions, 10 runs each, on one OpenAI model

  • Every question moved

    6 of 6questions came back in 3 or more different orders

  • Steady only at the top

    4 of 6kept the same #1 in all 10 runs, not the places below

See all 6 questions and the method

In the bakery question, the same 5 brands came back every time. Only their order moved.

Four questions kept the same #1 in all 10 runs, but not the places below.

The law-office question named 8 brands, and only 3 were in every answer.

The 10 runs of each question started within one minute. So one check that names you says little about the next answer.

Same question, 10 runs: how the top 5 changed
Question (by category)Different top-5 ordersMost common #1Brands namedNamed in all 10 runs
Website builder for a law office95 of 1083
Project tool for 20 people510 of 1083
Website builder for a bakery57 of 1055
CRM for 10 people410 of 1084
AI assistant for a 20-person company410 of 1064
Accounting software for 10 people310 of 1064

OpenAI model via Codex CLI with web search, not the ChatGPT app, 01-02.11.2026, 13 software questions. 10 runs per question. The 3 e-mail tool questions had only 5 runs each and are not shown. One brand written in different ways counts once. Product names are left out on purpose.

How did we get these numbers?

From our own test: 13 software questions, 115 runs on an OpenAI model, 1-2 November 2026.

See the setup, limits and sources
Model
An OpenAI model with web search, through the Codex CLI. Not the ChatGPT app.
When
1 and 2 November 2026. The 10 runs of each question started within one minute.
What we asked
13 software questions in English, 115 runs. 9 questions asked for a top 5, best first.
How
Each run started a fresh session with the same short instruction. We counted the brands in each top 5 and their order. Spellings of one brand count once.
Limits
One engine, software questions only, English only. Nenn's weekly checks use their own setup. This was a separate test.

Sources

  1. Nenn, Nenn test runs: 115 runs, 13 software questions, checked . OpenAI model via Codex CLI with web search, not the ChatGPT app, 01-02.11.2026, 13 software questions.
  2. Nenn, Nenn app: engines, weekly schedule, plan limits, trial, checked
  3. OpenAI, OpenAI API reference: temperature, checked

What should a tracker keep the same every week?

The method. When questions, engines and place stay fixed, a change in the number comes from the answers.

  • The same wording

    A new wording counts as a new question.

  • Each engine on its own

    One engine can move while the others stay still.

  • One country and language

    Answers depend on where a question comes from.

  • A share of answers

    Count how many answers name you. One answer can land either way.

  • Your place and your neighbors

    Note your place and who stands next to you. Fifth is not first.

  • Every answer, with its date

    Keep the full text, so you can read why a number moved.

What we compareWeek 44kept the same / changedWeek 45
QuestionWho is a good emergency electrician in Hamburg?kept the sameWho is a good emergency electrician in Hamburg?
PlaceHamburgkept the sameHamburg
Engines3kept the same3
Named firstKestrow ElektrikchangedFunkenhof
Your place4changed2
Elektro Lindqvist, two weeks side by side.
How Nenn keeps each point the same

Wording: Nenn asks a new wording from the next weekly check, as a new question.

Engines: each engine has its own column, so one engine moving does not hide the others.

Place: Nenn asks from the country and language you picked.

Share: your weekly number is the share of answers that name you.

Your place: average position counts only the answers that name you.

Every answer: each answer opens in full, with your name and the other names marked.

How does Nenn sample AI answers?

Once a week, on a fixed slot, Nenn asks all your questions on all 3 engines. It saves every answer.

Up to 90 answers a week on Start

How Nenn counts your weekly AI number

Mentioned in

Elektro Lindqvist · 3 AI engines · Hamburg · Last 4 weeks

Invented company and questions, one week of answers in the Nenn app.
CompanyWeek 42Week 43Week 44Week 45
Kestrow Elektrik38.1%42.9%42.9%47.6%
Funkenhof47.6%42.9%47.6%42.9%
Elektro Lindqvist19.0%23.8%28.6%38.1%
Strandvolt33.3%28.6%33.3%28.6%
Ohmhaus Nord28.6%28.6%28.6%23.8%

Mentioned in

38.1%+9.5%

Position

#3

Ranking, week 45

  1. 1.Kestrow Elektrik47.6%+4.7%
  2. 2.Funkenhof42.9%-4.7%
  3. 3.Elektro LindqvistYours38.1%+9.5%
  4. 4.Strandvolt28.6%-4.7%
  5. 5.Ohmhaus Nord23.8%-4.8%

Mentioned in, by engine

  • ChatGPT42.9%
  • Perplexity28.6%
  • Google AI42.9%

The answers behind every number

Read every answer in full

Questions this week

ChatGPTHamburg, Germany · Week 45

Who is a good emergency electrician in Hamburg?

For emergencies, Kestrow Elektrik and Elektro Lindqvist answer at night. Funkenhof has fixed call-out times.

Names in this answer

  1. #1Kestrow Elektrik
  2. #2Elektro LindqvistYours
  3. #3Funkenhof

This question on each engine

  • ChatGPTNamed, #2
  • PerplexityNot named
  • Google AIPartly named

Who AI names instead

Who AI names instead, and which sites it cited

Visibility

Mentioned in38.1%+9.5%
Position#3 of 5

Question: Who is a good emergency electrician in Hamburg?

BrandMentioned in
  1. 1.Kestrow Elektrik47.6%named by ChatGPTnamed by Perplexitynamed by Google AI Overviews
  2. 2.Funkenhof42.9%named by ChatGPTnot named by Perplexitynamed by Google AI Overviews
  3. 3.Elektro LindqvistOwned38.1%named by ChatGPTnamed by Perplexitynamed by Google AI Overviews
  4. 4.Strandvolt28.6%not named by ChatGPTnamed by Perplexitynot named by Google AI Overviews
  5. 5.Ohmhaus Nord23.8%not named by ChatGPTnot named by Perplexitynamed by Google AI Overviews
AI names Kestrow Elektrik, Funkenhof before Elektro Lindqvist. Elektro Lindqvist is mentioned in 38.1% of answers, position 3 of 5.

Invented company and questions in the Nenn app.

How the weekly count works

Once a week, on a fixed slot, Nenn asks every question once on each of the 3 engines.

Start tracks 30 questions, Growth 100 and Pro 250. On 3 engines, that makes up to 90, 300 or 750 answers a week.

The page says up to, because an answer is left out of the count when Google cannot be reached.

The trend line starts after the 2nd weekly check. Each week after that adds one more point.

Mr. Nenn, the Nenn mascot, smiling over the Visibility screen

Which AI engines does Nenn track?

Three: ChatGPT, Perplexity and Google AI Overviews. Claude and Gemini are not tracked.

Who is a good emergency electrician in Hamburg?

  1. ChatGPTWeekly
  2. PerplexityWeekly
  3. Google AIWeekly, since 27.10.2026
  • Claude: not tracked
  • Gemini: not tracked
The question is invented.
AI engines and their status in Nenn
AI engineStatus in Nenn
ChatGPTTracked every week
PerplexityTracked every week
Google AI OverviewsTracked every week, since 27.10.2026
ClaudeNot tracked
GeminiNot tracked

If Google skips the AI Overview, Nenn logs that answer as not named. Do many of your buyers use Claude or Gemini? Then keep that gap in mind.

Which AI engines Nenn checks ->

Can't I just ask AI myself each week?

You can, and it is a good first step. A tracker adds the repeat, the record and the count.

  • By hand

    1chat answer you read yourself

  • With Nenn

    90answers Nenn checks each week

Elektro Lindqvist, 30 questions on 3 engines.

  • Your chat knows you

    Your history, your settings and maybe your city shape what you see.

  • One run on one engine

    A hand check shows one answer from one AI engine.

  • Every question, every week

    A weekly sample logs each name and place on 3 engines. With 30 questions, that is 90 answers.

Check AI answers by hand ->

How do you start tracking with Nenn?

Enter your website. First answers come in minutes, then the weekly checks run on their own.

  • 1. Add your website

    Nenn reads it and asks you to confirm which company you are.

  • 2. Check the questions

    Nenn writes the questions your customers ask. Change any of them.

  • 3. Say where customers are

    Pick your city, your country or online.

  • 4. Read the first answers

    Some show during sign-up. An e-mail tells you when the rest are in. The trend line starts after the 2nd weekly check.

Mr. Nenn, the Nenn mascot, gives a thumbs up

Questions about LLM visibility tracking

Start 7-day trialCard needed. Cancel before day 7 and nothing is charged.

Do you use my personal ChatGPT account?

No. You never log in to ChatGPT or connect any AI account. Nenn asks each engine with its own access, from a clean start: no chat history, no saved memory. It asks from the country and language you picked.

How many weeks until the trend means something?

The line starts after the 2nd weekly check. Two points only show a direction. Each week after that adds up to 90 answers on Start: 30 questions on 3 engines.

Why not check every day?

Answers change even within one minute. In our test, the 10 runs of each question started within one minute. The top-5 order still changed in every question. So one answer a day is no surer than one a week. What helps is many questions, asked the same way every week.

Is an LLM tracker the same as a rank tracker?

Not quite. A rank tracker records your place in Google's list of links. An AI answer is a short text that may not name you at all. So an LLM tracker counts how often you are named, and your average place when you are.

Is LLM brand monitoring the same as social listening?

No. Social listening collects what people post about you. LLM brand monitoring reads what AI answers to your buyers' questions. The two can disagree.

What does average position mean?

It is your place in the answers that name you. Answers that leave you out do not count toward it. So read it next to the share of answers that name you.

What if Google shows no AI Overview for a question?

Then Nenn logs that answer as not named. If Google cannot be reached at all, that answer is left out of the count.

What does the 7-day trial include?

1 website, 2 checks and 3 articles. You add a card at checkout. Cancel before day 7 and nothing is charged. After the trial, plans are monthly, and you cancel in the app.

Start tracking your AI answers

Nenn asks your first questions during sign-up. Every week after that adds one more point to your line.

Start 7-day trialSee plans and prices
Card needed. Cancel before day 7 and nothing is charged.
Visibility card in the Nenn app: mentioned-in chart per day and a position leaderboard