TL;DR: We audited our own business exactly the way we tell clients to audit theirs: fresh, signed-out ChatGPT sessions, three samples per question, and a strict rule — a position only counts if a majority of samples name you. On the questions built around what makes us different, we held #1. On the plain questions most buyers actually type, we were not named at all. Both halves of that result taught us something, and this post is the full scoreboard.
Why we ran this audit on ourselves
More buyers now ask ChatGPT or Gemini to recommend a business instead of scrolling ten blue links. The AI names a few businesses and moves on. If it cannot verify your facts, it does not guess — it names a competitor whose facts were easier to verify. We sell exactly that visibility, so the honest test is to point the instrument at ourselves first, in public, with numbers we cannot dress up.
One rule matters before any result: measurements only count from signed-out, guest sessions. A result you see while logged into your own account reflects your own history with the tool, not what a stranger sees. Anyone who shows you a screenshot from their own account is showing you the wrong experiment.
What held: the specific claims
Asked which platform combines an AI receptionist, SEO and social media in one place for Canadian small businesses, ChatGPT named Marvic AI first in three out of three fresh samples. Asked about a $99 AI receptionist with a chatbot, it named us first in two of three. The pattern is not luck. Both questions contain a specific, checkable claim — an exact combination of services, an exact price. Specific claims can be verified against a website, a pricing page and third-party listings, so the AI can repeat them with confidence.
What did not hold: the plain questions
Then we asked the questions buyers actually type: the best AI receptionist for a small business in Toronto, an affordable answering service in Canada, a receptionist that speaks many languages. Marvic AI was not named in any of them. Other companies were — ones whose claims are repeated by directories, review sites and video platforms, not just by their own websites. Our own pages stated our facts well. Almost nobody else on the public web repeated them. To an AI building an answer, an uncorroborated claim is a rumour, and it treats it that way.
What we fixed the same day
We did not file the result and move on. The fixes shipped the same day:
1. The 52-languages claim, stated precisely. Our receptionist answers in 52 languages, but the claim sat in scattered wording. It now reads identically on the receptionist page, the homepage and our machine-readable site summary — the same words an AI can quote.
2. Third-party corroboration. We submitted the same canonical description — services, price, trial — to independent software directories for listing and review, so the facts exist where the AI cross-checks them, not only where we control the page.
3. The Google side of the same problem. Our sitemap now carries 287 URLs, and every page we publish gets a hard 48-hour indexing check: if Google has not confirmed it within two days, we diagnose and fix it rather than assume it.
4. A standing guard. The defended questions are now re-measured every morning in fresh guest sessions. A drop is not a report item — it triggers a same-day strengthening fix for that exact phrasing.
The checklist we would hand any business
If you run this audit on your own business — and you should — these are the five things that decide the result:
1. State your facts identically everywhere: name, services, area served, phone, price signals. Drift reads as doubt.
2. Give every service its own page, and open each page with a direct answer to the question a customer would ask.
3. Write FAQs in your customers’ actual words; question-and-answer text is the easiest content for an AI to lift into a reply.
4. Get the same facts listed by third parties you do not control — directories, reviews, videos. Corroboration is the currency.
5. Measure the way a stranger experiences it: signed out, several samples, majority rule, on a schedule — not once, from your own account, on a good day.
This self-audit is a small version of the work behind Marvic AI’s AEO services — measuring how AI tools see a business, then fixing the facts they cannot yet verify.
The honest bottom line
AI visibility is not a trick and it is not a trophy you win once. It is a position you hold by keeping your facts consistent, corroborated and current — and by measuring like a sceptic. We are publishing our scoreboard, including the questions we lose, because that is the standard we sell. If you want the same audit run on your business, start with a 14-day free trial or book a free presentation — 437-876-0886 · www.marvic.ai.
