Pulse index
NyttPulse’s own score, out of 100: each answer is scored on whether your brand is recommended, then on its place. A recommendation is worth more than a mere first mention.
Position tracking
Measure how your brand shows up in the answers of seven AI engines, plus Claude for a surcharge, benchmark yourself against competitors and see which sources the engines rely on.
Why track your position
To compare offers, choose a provider or simply look something up, more and more people ask ChatGPT, Gemini or Perplexity directly, and Google itself now shows generated answers. The answer they read names only a handful of brands: that’s where you need to be visible.
Position tracking shows where your brand appears today: on which questions, in which engines, against which competitors and thanks to which sources.
The technical audit will check that AI engines can access, read and understand your pages.
Article writing and the action plan will help you climb, with tracking measuring the results.
The new reflex
In two years, asking an AI has become a reflex for finding information, comparing and choosing. Even Google now answers with AI. If your brand isn’t in those answers, it drops out of your customers’ journey.
1.2bn
people use ChatGPT every week
2.5bn
monthly users of Google’s AI Overviews
38%
of internet users used AI to get information last week, up from 11% in 2024
57%
saw AI-generated answers in a search engine last week
| Dec 2024 | 300m |
|---|---|
| Feb 2025 | 400m |
| Mar 2025 | 500m |
| Aug 2025 | 700m |
| Oct 2025 | 800m |
| Feb 2026 | 900m |
| Sept 2026 | 1.2bn |
| May 2025 | 400m |
|---|---|
| Jul 2025 | 450m |
| Oct 2025 | 650m |
| Feb 2026 | 750m |
| May 2026 | 900m |
| Jul 2026 | 950m |
| Oct 2026 | 1bn |
NyttPulse projection: the 2024–2026 trend measured by the Reuters Institute, extended for illustration only. This is not a published figure.
In 2027 · ≈ 51%
If the trend continues, around one internet user in two will turn to AI for information in 2027. The brands cited today are building the lead — don’t wait for your competitors to take your place in those answers.
Sources
OpenAI · DevDay 2026
September 2026engadget.com
Google · I/O 2026
May 2026blog.google
Reuters Institute · Generative AI and News Report
October 2026reutersinstitute.politics.ox.ac.uk
CRÉDOC · Baromètre du numérique
February 2026credoc.fr
TechCrunch · ChatGPT
February 2025October 2025February 2026techcrunch.com
Alphabet · résultats trimestriels
October 2026blog.google
In France, 48% of people aged 12 and over use generative AI.4 Figures published by the companies themselves or by independent institutes, as of the date shown. Weekly and monthly measures cannot be compared with each other.
AI vs Google
A classic SEO tool measures where a page ranks in a list of links. In a generated answer, the question changes: is your brand named, where does it rank, and which sources put it there?
| Criterion | Classic Google rank tracking | Tracking in AI answers |
|---|---|---|
| What you look at | A results page: ranked links | A written answer that names brands and cites sources |
| What you track | Keywords | Questions phrased the way your customers ask them |
| What is measured | The position of a URL | Whether the brand is mentioned, its rank among the brands named and the sources used |
| How results are read | One ranking per query | Answers can vary, so measurement is repeated and read as a trend |
The two are complementary: NyttPulse doesn’t replace your SEO tracking — it covers what SEO tracking can’t see.
Metrics
Every metric shows its definition, its basis (the number of answers or questions taken into account) and its scope: brand, market and active filters.
NyttPulse’s own score, out of 100: each answer is scored on whether your brand is recommended, then on its place. A recommendation is worth more than a mere first mention.
The share of answers naming the brand, out of all answers in the run.
The share of answers citing your domain as a source. Being named and being cited as a source are two different things.
When the brand is mentioned, its rank among the brands named in the answer — 1 is best. Answers that don’t mention it don’t count towards the average.
Among the mentions of your brand and your tracked competitors in AI answers, the share that goes to your brand. Each brand counts once per answer; your brand and your tracked competitors make 100%. Brands you do not track are not counted: they are reported separately.
The share of questions where your brand is cited at least once.
Untracked brands: 4, reported separately
Compared with the previous run
Answer sentiment
Every answer that mentions your brand gets a sentiment: positive, neutral, negative or mixed. It is backed by one to three excerpts copied word for word from the answer and checked automatically: an excerpt that can’t be found in the answer, or that doesn’t name your brand, is rejected.
For hands-on local support, several agencies are often mentioned. Your Brand is appreciated for how responsive its customer service is, even if its prices are above average.
Excerpt verified in the answer
Questions analysed
Each question belongs to a brand and a market: country, language, national or local tracking. During the trial, NyttPulse suggests them from your website; you can also add them one by one or import them from an Excel template.
Engines tracked
Each question is asked in a fresh conversation located in your market’s country, with no personal account or history. The collection method is shown for every answer.
Public chat interface (web search enabled where the engine offers it)
Public chat interface (web search enabled where the engine offers it)
Public chat interface (web search enabled where the engine offers it)
Public chat interface (web search enabled where the engine offers it)
Public chat interface (web search enabled where the engine offers it)
Google results for the market’s country and language
Google results for the market’s country and language
Official API with web search, answer labelled “API answer”
Public chat interface (web search enabled where the engine offers it)
You choose which engines to track. The free trial uses ChatGPT and Gemini. Claude is available as a premium engine; Google AI Overviews and Google AI Mode is coming soon.
Competitors and sources
Add them by hand or import them from an Excel file (name, website, aliases). NyttPulse also flags the brands that appear in answers: you choose whether to track or ignore them. 3 tracked competitors are included; more can be added as an option, one at a time, up to 30.
The websites engines rely on are grouped by URL and by domain. The Sources tab separates sources cited in the answer from those merely consulted, with columns, a filter and counters. Some engines, such as Perplexity and Google AI Overviews, present all their sources as cited: the tool points this out.
Frequency and trends
The same question can get different answers from one run to the next. That’s why NyttPulse repeats the measurement at the frequency you choose (daily, twice a week, weekly, twice a month, or monthly) and works with rates and trends rather than a single answer.
Reliability
Summaries and alerts
Two tracking views to make the most of each run without rereading everything.
After each run, NyttPulse gathers the passages of the answers that name your brand and summarises them: key points, their attributes and tone, each linked to the excerpts that support it.
Each user chooses what they receive: channel, frequency, day, time, format, content and language.
MCP connectorComing soon
NyttPulse will offer an MCP (Model Context Protocol) connector as a paid option: from a compatible assistant such as Claude, you’ll query your results in plain language without leaving the conversation.
The connector starts with position tracking. More doors will open gradually as the other modules launch.
Which questions did my brand lose ground on in ChatGPT this week?
NyttPulse · reading tracking measurements
Here are the questions where your position dropped, with the competitors cited instead of you…
Enter your website: NyttPulse prepares your market and questions, then runs a first check on ChatGPT and Gemini.