AI visibility monitoring for agencies
Customers ask ChatGPT which clinic, bank or tool to choose, and go where it points. Polyxe asks five AI engines your customers’ questions every week and measures how often they name you.
Free demo: 10 questions · Claude · report in about 10 minutes. Runs in Telegram, nothing to install.
Every report opens with them. Each number comes with a plain-word grade, so nobody has to ask whether 31% is a lot.
We measure frequency over a series of answers, not one lucky screenshot.
You tick the topics — the services you compete for. We build up to 100 customer questions for them from our phrasing library (“which clinic to choose”, “where to get a cheap ultrasound”).
Every week we ask the AI engines each question several times — answers vary run to run, so we measure how often you're mentioned.
You get share of voice (SoV — how often it's you they name) by topic, question type and engine, a comparison with competitors and the exact sources the web-search engine cites — where to publish.
Five AI engines, three repeats each. A filled cell means the brand was named in that answer. Engines answer differently every time, so the report counts how often you are named, not whether it happened once.
namednot named
Engines: Claude, ChatGPT, Perplexity, DeepSeek and Gemini. Which ones run depends on the plan: Solo — two, Start — three, Pro and Agency — all five.
No "#1 in ChatGPT" guarantees — nobody can promise that honestly. We measure, you fix, the next scan shows growth.
The same report lands in Telegram after every scan, and goes to your client as a PDF.
How often each engine names you — and how often it names anyone of you and your competitors (if almost nobody, that is the engine, not your weakness). Perplexity searches the web and answers from the pages it finds — publications affect it directly; the others answer from memory.
Share of answers mentioning the brand, per scan topic; “general questions” — about the business as a whole, not tied to a service.
green — 50% and up, amber — 20–49%, red — under 20%
What the engines praise and what raises doubts, in the phrases they actually use.
What gets praised
What raises doubts
Engines with web search cite real pages. The report sorts them into directories and review sites where you can get listed, company sites, and portals you cannot influence.
42% Directories, review sites, media36% Company sites22% Portals
Directories, review sites, media — where you can get listed
Biggest potential: reddit.com — 22 answers cite it without the brand (you are in 0%).
Per topic: who leads, which kinds of pages the engines cite, and a verdict on whether to go in now or later.
Veneers 4%
Fragmented, sources are open — go in first.
Emergency dentist 21%
CityDent 24 74% — near-monopoly; sources are open, but not the first priority.
Share of Voice week by week, with the margin of error. A move inside the margin is labelled as noise, not as growth.
You sell the result to your client. The tool stays behind the scenes.
After every scan the bot sends the summary with buttons to every screen. Forward it to your client as is, or send the PDF.
The report ships under your logo and agency name. Hand it to the client as your own.
Every brand is scanned once a week, on its own day and hour. The report arrives by itself.
If Share of Voice falls by more than your threshold and more than the margin, the report arrives marked as an alert.
Questions, topics and the client PDF use the language your client’s customers search in. The interface stays in yours.
Ask about the numbers in plain words, or order a publication brief for a weak topic: titles, where to publish, outline, facts to include.
The engines name players you did not list. The report surfaces them, and you add the relevant ones to the comparison with one tap.
No dashboard to learn and no logins to hand out. Brands, scans, reports and payment are in one chat.
Prices in USDT per month. Available now: the Pro plan — weekly auto-scan, all AI engines, PDF with your own logo, breakdown, sources and sentiment. The other plans are in development and coming soon.
In development — coming soon.
In development — coming soon.
In development — coming soon.
1 brand, 1 topic, 10 questions × Claude, 1 repeat — a real scan, report in ~10 minutes. Free; repeat after 30 days.
Run the demo30 days free — tap "Discuss a pilot", we reply within a day.
Discuss a pilotHow to pay: Plan → Pay → USDT (TRC20), or we'll arrange any convenient way together.
No auto-billing — a plan runs for a month and renews only when you pay. To cancel, simply don't renew; history stays.
Assistant is in every plan: “About the service” — for everyone; questions with brand context — from Solo (100 turns/month), “Topics” — from Start (300), “Publication brief” via Claude — from Pro (1000 turns and 10 briefs); Agency — 3000 turns and 30 briefs. Add-on “+1000 turns” — 15 USDT per 30 days, switched on after payment via support.
Short answers. The assistant inside the bot answers the rest.
The demo is a real scan, only shorter: 10 customer questions, one engine, a report in about 10 minutes.
Get a free demo scan