Share of Answer, live
Our own AI visibility, measured every week.
Why this page exists
We sell AI visibility, so we run our own method on ourselves and publish the result, up or down. Every week, 20 real buyer expressions go to ChatGPT, Claude, Perplexity, Gemini and Google AI Overview, and we count how often NextChannel Studio is named. Day one: 124 answers, zero mentions. Week of September 14: 135 answers, still zero. This page is the climb.
Baseline measured August 31, 2026 · last reading September 18, 2026: 0 of 135 · page created September 1, 2026, rewritten every Monday with the new reading · judged on a 4-week rolling average because AI answers change constantly between runs · written by Thibault Genon-Catalot, founder
The definition
What is Share of Answer?
Share of Answer is the share of AI-generated answers that name a given brand, measured on a fixed set of buyer questions asked repeatedly across engines like ChatGPT, Perplexity and Gemini. It is GEO’s version of market share: not where you rank in a list of links, but how often you are part of the answer itself.
Three rules make the number trustworthy. The questions stay fixed, so readings are comparable over time. Each question is asked several times, because AI answers change between runs. And the result is judged on a 4-week rolling average, never a single reading, because cited sources change constantly from month to month.
The rest of this page is the metric in action: our own Share of Answer, measured every week by the same robot that runs our client audits, published whether it goes up or down.
The baseline
Day one, August 31, 2026: named 0 times out of 124 answers.
Measured on August 31, 2026, the day before the first method page went live: 25 questions, five engines, one pass each. Exactly what a brand with no citation layer should expect, and the honest starting point of every curve we will ever show a client. The questions cover our own buying questions: audits, prices, agencies, lead magnets, ChatGPT apps.
| Engine | Answers recorded | NextChannel named |
|---|---|---|
| ChatGPT | 25/25 | 0 |
| Claude | 25/25 | 0 |
| Perplexity | 25/25 | 0 |
| Gemini | 25/25 | 0 |
| Google AI Overview | 24/25 | 0 |
Month by month
One reading a month, published empty until it is filled.
The same buyer expressions, the same five engines, one reading taken on the last full week of every month. Since October 11 the set is the 20 expressions of our own published audit of September 21. The months ahead are already in the table with the week they will be measured, so you can see the series before it exists. We publish each line whether it goes up, down or nowhere.
| Month | Reading taken | Answers measured | Times named | Share of answer |
|---|---|---|---|---|
| August 2026 | Baseline, August 31 | 124 | 0 | 0% |
| September 2026 | Week of September 14 to 18 | 135 | 0 | 0% |
| October 2026 | Week of October 26 to 30 | not measured yet | – | – |
| November 2026 | Week of November 23 to 27 | not measured yet | – | – |
| December 2026 | Week of December 21 to 25 | not measured yet | – | – |
The latest reading
Where the counter stands right now.
Pulled live from our measurement robot, the same one that runs client audits. One engine is re-measured every weekday, so a full week is one complete reading of every question on every engine. Since September 21, 2026 the set is 20 expressions, the same number every client audit uses, built from what buyers really type into Google. A full week is 100 answers. The date below is the week the reading was taken, and it is the only date that matters here.
Loading the latest reading…
Raw answers are timestamped and stored; ask and we will show them. Judged on a 4-week rolling average, never a single run.
The actions behind the curve
What we did, dated, next to what moved.
A curve without the actions behind it is a horoscope. This log is what lets anyone, including us, connect cause and effect.
Baseline measured: 124 answers, named 0 times. The three first method pages went live the same day, after the baseline ran: /methodology, /geo-pricing, /geo-price-comparison.
Domain moved to nextchannelstudio.com (brand name = domain, for entity clarity). Four more pages shipped: /aeo-vs-geo-vs-seo, /why-chatgpt-ignores-your-brand, /best-geo-agencies, and this page.
The entity work: Clutch company profile created (under review), the free Spot-Check (now the AI Visibility Checker) submitted to G2's Answer Engine Optimization Tools category, and NextChannel Studio got its Wikidata entry (Q141246576) with the official website, inception and EU VAT number as anchors.
/ai-search-statistics published: 32 numbers, every one read at its primary source, nine widely quoted figures rejected as unverifiable and named on the page.
Measurement rebuilt on the five engines we now report: Google AI Overview is read through a search data provider instead of a memory-only run, and the question set moved from 25 to 27. The weekly series restarts from this date.
First two audits delivered to real clients, 615 answers measured between them. The findings on how unstable engine answers are now sit on /ai-visibility-score.
First client reviews published on Trustpilot and Clutch. The Wikidata item was deleted in September for lack of independent coverage; it goes back once press coverage names us.
/ai-visibility-score published, and the founder is now named on the home page: ChatGPT had been answering that NextChannel Studio's founder was not publicly identified. Question 28 added to this measurement.
The question set rebuilt: 20 expressions taken from real Google searches in the United States, the same method and the same number as every client audit. It replaces the 29 questions we had written ourselves. The weekly series restarts from this date.
The weekly measurement now asks the 20 expressions of our published audit of September 21 (the list a buyer sees in the audit itself), so this curve and the audit measure the same thing. The 20 expressions of September 21 to October 10 were a sister list from the brief of September 20; both series stay archived, never mixed. Our first implementation tool, built the same day, works from that audit: pages to build, sources to get listed on, a weekly re-measure that feeds this page.
Free, 2 minutes
Before you buy any audit: run one free check.
5 buyer questions, 2 AI engines, and you see who gets named instead of you. The verdict shows on the page, and a copy lands in your inbox.
The same measurement, on your brand
This is the exact curve the audit starts for you.
The €1,500 audit runs this same measurement on your buyers’ questions: 300 answers, who is named, from which sources, and what to fix first. The day-90 re-measure is a €200 add-on. The method is public at /methodology. The wider market numbers, 32 of them with sources, are on /ai-search-statistics. If a free tool has already handed you a score out of 100, here is what that number measures and what it cannot tell you. Our own study of 615 AI answers, and what it says about measuring any of this once, is at /do-ai-answers-change-every-time.