AI visibility for Swiss SMEs
THREE COMPANIES
IN THE ANSWER.
ARE
YOU ONE?
Your customers ask ChatGPT before they open Google. Every week ECHO checks four AI services and, on request, Google’s AI box: whether they name you, who they name instead, what they get wrong, and what to change.
This is what we actually measured
No study, no extrapolation. Valid measurements across brands from several industries, as of 29 September 2026, every answer stored verbatim.
A Swiss B2B manufacturer
That is how often AI names them when someone asks for suppliers. One single competitor appears about three times as often in the very same answers.
A Swiss IT services provider
Two competitors show up in 36 % and 25 % of the same answers. It is a close race, and without measurement nobody sees it.
How often the engines disagree
Across all measurement days with at least three engines. Gemini names the tracked brands in 31 % of supplier answers, ChatGPT in 14 %. Check one engine and you check the wrong one.
How far the measured brands sit apart
One shows up in nearly two answers out of three, another in one out of twenty. Both are measured the same way, across the same engines.
And where the AI got its answer
Across the sources that AI services supplied with their answers in our measurements, these sources sit at the top next to your own pages. That is where it gets decided who the AI knows.
Two different bases appear on this page, and they mean two different things: the evidence set with every valid measurement, and the comparison across every company we measure, which only counts companies with enough measurements in the shared window. As of 29 September 2026, computed from the ECHO database, not lifted from a study. Customer names anonymised, competitors unnamed. A single week is noise, so we sample repeatedly and only read trends across weeks.
Three things ECHO measures beyond mentions
Every block carries the figure behind it and the denominator next to it. The full story sits on its own page.
The AI box above the Google results
Google now writes its own answer above the search results. For your questions, ECHO checks three things: whether that box appears at all, whether it names you, and whether it credits your website as a source.
When the AI gets you wrong
You give us your hard facts, and what you do not do. ECHO compares every measured answer and shows the sentence verbatim when ChatGPT, Claude, Gemini or Perplexity claims otherwise.
German, French, Italian, English
Whether an AI names the same companies in French as in German cannot be assumed, only measured. ECHO can therefore measure in the three national languages and in English, each language with its own questions and its own figures, never mixed.
Where you stand next to other companies
A number on its own says little: being named in 30 per cent of answers is good or bad depending on where everyone else sits. ECHO measures the exact same thing for several companies with the same method and shows you the middle and the range most of them fall into, within your field once enough companies are measured there. You never see an individual company, no league table, no top score.
The comparison appears only once a group reaches a minimum number of companies, and a company counts only from 30 measurement units in the shared window of 28 days. Only questions without your brand name in them are compared: with «what does company X offer?» a mention is close to certain, and a single project full of such questions would pull the middle up for everybody else.
Middle across every company we measure, as of 19.10.
Every measurement wobbles. We write down how much.
Ask the same AI the same thing three times on one day and the result flips in around 10 per cent of cases. Between the four services the spread is a good four times larger: on 44 per cent of our measurement days up to 29 September 2026 they disagree about who deserves a mention. Check one service and you check the wrong one.
So every figure here carries a range. When the range is nearly as wide as the figure itself, the cockpit tells you: that means there are still too few measurements to say anything. We do not read single weeks, we read trends across weeks.
Anyone selling you a decimal place without a range is selling you noise.
The full arithmetic behind it →One screen for how AI talks about you
No SQL. No data team. We build the questions your buyers actually ask, you strike what does not fit. ECHO runs them across four AI services every week, and you see where you stand and what to fix first.
At the end of every week four things sit in front of you: a list of names, the answers word for word, the sources the AI reads, and a task list ordered by estimated impact.
None of it needs interpreting. You read what ChatGPT says about your industry, see who stands there instead of you, and get one job per gap: what to do and who does it, ordered by estimated impact and effort. On Google you could sit in eighth place and still get clicks. An AI answer names three companies. There is no second page.
A list of names
Who appears in the answers when somebody asks for what you sell. Your brand and everyone else, ranked by how often and split by engine.
The answers word for word
Not just a number, but the sentence you appear in, or the one you are missing from. Every answer stored, with date and engine.
The sources the AI reads
Which pages feed the answer, and that yours usually is not one of them. That is where it is decided who the AI knows.
A task list
One job per gap, ordered by estimated impact: what to do and who does it. Not a checklist nobody opens.
Not just the gap. The specialist who closes it.
A chart on its own closes no gap. ECHO hands each gap to the right role, and turns it into a brief you can act on this week.
Tech consultant
robots.txt blocking GPTBot? No price on the page, no visible date? You don’t get the warning, you get the finished fix, ready to paste.
When an AI has two pages it found for the same question in front of it, a peer-reviewed study finds that topic, price and date help decide which one it cites. Source: Vishwakarma, Kumar, Jamidar, SIGIR ’26, arXiv 2605.25517. The authors work at Sprinklr, a provider of customer communication software.
Copywriter
For “What is GEO”, others serve as the source, never you? You get the brief: the H1, the questions to answer, the prices and the date, and which source to beat, written in your voice.
PR advisor
The AI recommends agencies from three directories you’re missing from. You get the list, ranked by how often the AI actually pulls from each source.
CMO as a Service
Competitors own the questions that decide the deal? You get the strategic call: which category to claim, which comparison pages to build, which message wins.
Four phases, one loop
No agency retainer to read a PDF once a quarter. Scroll through the loop ECHO runs every week, each phase pins, and the next slides up over it.
WHEN AI ANSWERS, does your brand echo back?
Gartner expects 25 % less search engine volume in 2026 as AI chatbots take over the query. Without measuring, you will not notice when a competitor takes your place in the answers.
Pricing, without a sales call.
The first measurement and the call that explains it cost nothing, even if nothing comes of it. If you then want to keep measuring, here is what it costs, in CHF. Switch between monthly and annual below.
The first measurement is free
We put twenty questions from your customers’ everyday life to ChatGPT, Claude, Gemini and Perplexity and talk the result through with you. Then you decide whether to keep measuring.
- 40 tracked questions
- 1 website, unlimited users
- All 4 answer engines, every week
- Ready-made actions to close each gap, with the tech check
- Answer archive: 8 weeks
- Support by mail
What the AI gets wrong: from Pro up. You can already store your facts here. Monthly briefing: from Pro up.
Google AI box: on request we switch it on per project. A Google query counts as one normal query.
- 120 tracked questions in total
- shared across up to 2 websites, unlimited users
- All 4 engines, weekly or daily
- Ready-made actions plus the tech check
- What the AI gets wrong: fact watch included
- Monthly briefing for the management team, with print view (beta)
- A competitor’s view and export of the answers as a table
- Full answer archive, no time limit
- Quarterly strategy call
Google AI box: on request we switch it on per project. A Google query counts as one normal query.
Fair use. Daily screenings are included and run with one reading per day, weekly ones with three readings per screening. That is our measurement method.
Request a measurement- 350 tracked questions in total
- shared across up to 5 websites, unlimited users
- All 4 engines, weekly or daily
- Everything in Pro: tech check, fact watch, monthly briefing (beta)
- A competitor’s view and export of the answers as a table
- Full answer archive, no time limit
- Monthly strategy call
Google AI box: on request we switch it on per project. A Google query counts as one normal query.
Fair use. Daily screenings are included and run with one reading per day, weekly ones with three readings per screening. That is our measurement method.
Request a measurementOn every plan: four languages with their own figures, a monthly standing to pass on, an email after every scheduled measurement, AI crawlers on your website, a comparison with other companies once enough companies in your group are measured, unlimited users and questions to the founder. Measurement runs weekly, monthly or on demand, from Pro up also daily.
One entry point: we measure first and talk you through the result, then we set up your access. Request a measurement →
Add-ons & Enterprise
Build the package that fits your situation. Every add-on works on top of any plan.
More questions in tracking when your allowance runs tight. One package at most on Start, as many as you need on Pro and Agency. On a plan with several websites the questions apply to all of them together.
Your data on a dedicated instance hosted in Switzerland.
Additional AI services beyond the four we track as standard.
Everything freely configurable: questions, websites, AI services, hosting and support.
digital@pixels.ch → Request a measurement →All prices in CHF, excl. VAT. Annual billing is 15 % cheaper than paying monthly.
What people ask before their first scan
What is GEO?
How do I measure whether ChatGPT recommends my company?
What does a GEO tool cost?
Is this just SEO with a new name?
Which AI services do you track?
How is a “mention” different from a “citation”?
What do I actually get in the fix queue?
Do I have to fix everything myself?
How long until I see results?
Who builds ECHO?
Find out whether you are in the answer
We build twenty questions your customers really ask, run them once through every service, and explain the result to you in a call. Free, even if nothing comes of it. If AI answers in your field hardly ever name a supplier, we tell you that instead of selling you a subscription.