When someone asks an AI about your category, does your brand come up?
More buying decisions now start with an AI answer than a search results page - and those answers are invisible to you unless you go and ask. A check asks eleven assistants the 12 questions your buyers ask, keeps every page they read, and tells you where you stand, who gets named instead and what to do about it.
You are named only where a buyer would name you
The discovery questions ask about your category, never about you: if an assistant brings you up unprompted, that is real visibility, and naming you would guarantee a mention and measure nothing. The reputation, comparison and complaint questions name you, as a real buyer would.
Your site, graded like a crawler sees it
Some sites accidentally block the fetchers assistants send when a person asks a question (ChatGPT-User, Claude-User, PerplexityBot), making everything they publish invisible in those answers. We check that from the outside in a second, and we tell training crawlers such as GPTBot apart from them.
The full check goes much further
Eleven assistants (ChatGPT, Claude, Grok, Gemini, Perplexity, Mistral, DeepSeek, Kimi, Microsoft Copilot, Google AI Overviews and Google AI Mode), 12 buyer questions each, every page they read, who AI names instead, what to publish and what to fix, and tracking over time.
Every question gets chased
Your first result lands within about an hour. Any question an assistant fails to answer in time is re-asked automatically 15 minutes, 3 hours and 9 hours later, at no extra cost, and we email you when the check is complete. We never count an error message or a cut-off answer as an answer: it is asked again. We never invent an answer to fill a gap: a question that stays unanswered is shown as such, because that is what a buyer asking that assistant would get.
Asked live, never from a cache
Every check asks every question live, at the moment it runs: eight assistants through their own official APIs, with web search switched on, and Google AI Overviews, Google AI Mode and Microsoft Copilot, which offer no API for this, straight from their live results. No answer is reused from a cache or an earlier check.
Where you stand, question by question
A preview of the check's own layout. The sample is a real, unedited check on a real brand.
Lead recommendation in 0 of 118 answers · named in 96
That is 35% of answers, give or take 9 points; named in 81%, give or take 7. Each question is asked once per assistant, so read the range rather than the decimal.
How the answers lean when you are mentioned
switcher.ie fed 6 answers that named Oryndex and never you. bonkers.ie fed 4. Get listed there.
“Veltora is a reliable, transparently priced option for small businesses in Irish towns; installation lead times are the recurring complaint.”
Median page is 14 months old · 2 pages from 2019 still feed 7 answers · your own pages read: home, pricing, help; no reviews page read
The eleven assistants every check asks
Each question is asked once per assistant, live: through the providers' official APIs with web search on, and Google AI Overviews, Google AI Mode and Microsoft Copilot from their live results. Nothing is gated behind a higher plan: every check asks all eleven. Assistant names and marks identify the services asked; no affiliation or endorsement is implied.
Five chapters, every number traceable to an answer
The check reads in the order you would ask the questions yourself. Nothing is a black box: every figure links back to the stored answers and the pages behind them.
What AI says about you
How each assistant describes you in its own words; what changed in the wording since your last check; where you moved; and AI versus real people when you also hold a sentiment report on the brand.
Where you stand
The matrix: question by question, assistant by assistant - first recommendation, named, in passing or absent. Wins and losses by buyer stage, a score per assistant, every answer stored word for word with its date.
Who wins the answer
Share of voice over the same answers, who AI names instead, you versus your main rival on each question, and the citation gap: the pages behind the answers you lost.
What the assistants read
Every page retrieved, how many answers it fed, who owns it (you, rivals, reviews, news), which of your own pages were read, and how old the reading is.
What to do next
Claims of fact to verify, with a correction kit for each; what to publish, one brief per lost question; and an AI-readiness audit of your site with pasteable fixes.
See one on a real brand
A complete, unedited check: every chapter, every answer, every page read. Then decide.
Three steps from a brand name to the receipts
Name your brand, market and rivals
We draft the 12 questions a buyer in your market would ask: discovery, comparison, reputation, trust, value. Edit any of them before you pay. Your set is saved and re-run word for word next time, so checks compare like with like.
Eleven assistants answer, we keep the receipts
Each question goes to ChatGPT, Claude, Grok, Gemini, Perplexity, Mistral, DeepSeek, Kimi, Microsoft Copilot, Google AI Overviews and Google AI Mode through their official APIs with web search on. We store every answer verbatim and every page each assistant retrieved, not only the ones it cited.
Read it within about an hour, then track it
Your first result lands within about an hour. Any question an assistant fails to answer in time is re-asked automatically 15 minutes, 3 hours and 9 hours later, at no extra cost, and we email you when the check is complete. Track the brand weekly, biweekly or monthly and get an alert when you stop being the first recommendation.
We never invent an answer to fill a gap: a question that stays unanswered is shown as such, because that is what a buyer asking that assistant would get.
One check against a monthly tracker
AI visibility trackers sell a subscription to a dashboard. A PollyX check sells the evidence, by the check. Here is what differs.
PollyX AI Visibility check by the check, from €20.99 | Subscription AI trackers typical, as published or reported | |
|---|---|---|
| Assistants asked | Eleven, all included: ChatGPT, Claude, Grok, Gemini, Perplexity, Mistral, DeepSeek, Kimi, Microsoft Copilot, Google AI Overviews and Google AI Mode. | Three on entry plans (typically ChatGPT, Perplexity and Gemini or Google AI Overviews). More as paid add-ons, or on enterprise tiers only. |
| What you pay to start | €20.99 for one check, VAT included. No subscription. | A monthly plan, reported from about $95 a month, billed monthly or annually. |
| Behind every answer | Every page each assistant retrieved while answering, often a hundred or more per question, with publish dates and who owns them - not only the pages it cited. | The sources the answer cited. What was read behind them is an enterprise feature where it exists at all. |
| The questions | 12 buyer questions per check, drafted for your market and rivals, every one editable before you pay, saved and re-run word for word. | 25 to 50 prompts on entry plans, hundreds on higher ones, tracked daily. Volume per prompt as a 1 to 5 scale, not a number. |
| Your own website | AI-crawler readiness audit and pasteable fix kits (llms.txt, JSON-LD) in every check; which of your own pages AI actually read. | Crawl audits on enterprise tiers, or a separate SEO tool. |
| What AI gets wrong | Claims of fact extracted from the answers; mark one wrong or outdated and a correction kit is drafted, with the pages the claim came from. | Fact-check views on some tools and tiers. |
| Tracking over time | On demand, or weekly, biweekly or monthly on a plan from €93 a month. Position-change alerts to Slack, Teams or a webhook. | Daily tracking, included in the subscription. |
| Honesty of the number | A margin of error beside the headline share. An unanswered question is shown as unanswered and re-asked free, never filled in. | A visibility percentage over the week's chats, usually without an error range. |
| Money back | Money back if a check finds nothing or fails. 14-day withdrawal on credit packs and plans. | Fees typically non-refundable. |
| Pairs with | A sentiment report on the same brand (the Full Picture): what real people say, beside what AI says. | AI answers only. |
The right-hand column describes AI visibility subscription tools as published on their own sites or reported by third-party reviews on 28 October 2026. Prices and limits change; check each vendor. Daily tracking across hundreds of prompts is a real strength of theirs: if you need it, that is the instrument. A PollyX check answers what the assistants say and why, with the evidence attached.
Are you still the first recommendation next month?
Your brand checked every week across 11 AI assistants, with what they read, what they get wrong, and what to publish to fix it. One check is a snapshot; a plan is a monthly allowance of checks, one brand every week on Starter or three on Pro, each compared with the last.
Position changes, with alerts
Every tracked check says which assistants stopped or started naming you first, by buyer stage. A material loss goes to Slack, Teams or a webhook.
The trend that actually moves
Mention rates barely budge on established brands. The trend tracks your visibility score, first-recommendation share and tone - the numbers that change.
Same questions, every time
Your saved question set is re-run word for word, so a move is a move. Consistency across checks shows how often each assistant puts you first.
One check, one price
Pay per check. Buy a pack to pay less per check, or a plan only if you want tracking.
Packs, plans, the API plan and refunds are on the pricing page.
Questions about the check
What is an AI Visibility check?
Why does AI visibility matter for my brand?
If I ask an AI assistant myself, will I get the same answer you show?
What if an AI assistant doesn't answer during a check?
How is PollyX tracking different from other AI visibility trackers?
How is the AI Visibility check different from the sentiment report?
Can I add more questions, cities or daily tracking to a check?
More in the full FAQ and the help center.
What do the assistants say about you?
Your brand, your market, your questions. Your first result lands within about an hour.