Not "you rank badly" — wrong. ChatGPT tells patients your Cedar Park office takes an insurance it does not. Perplexity says your downtown hotel has no parking. We check what the AI engines actually say about each of your locations against what you told us is true, quote the words back to you, and generate the corrections.
Examples above and below: Fernhollow Dental is a composite, not a client.
Every figure is a frequency across sampled runs, with the run count shown, and every accusation is a verbatim quote against a fact you declared. No screenshots, no vibes, and nothing we cannot stand behind.
One real check, start to finish. Fernhollow Dental is a composite example — not a client.
The questions people really ask about each of your locations, run live across ChatGPT, Claude, Gemini and Perplexity — each sampled three times per engine, because AI answers change run to run.
Every claim an engine makes about a location, checked against the facts you declared: the verbatim quote, the engine, the question and the run count — and "not measured" wherever you have told us nothing. Then, second, how often you are named and recommended, and the pages the engines cite.
For every wrong fact, the correction written for you — listing text, per-location structured data, an FAQ answer in your customers' words. Then we re-check every cycle, so what got fixed stays dated.
Patients ask AI which dentist takes their insurance, who is open Saturday and who can see them today. See what each engine tells them about every location — against what your own pages say.
AI accuracy for dental groups →Travelers ask AI about parking, pets, resort fees and check-in times. See where the engines get your property wrong, quoted, against the facts you publish.
AI accuracy for hotels →We ran the check before contacting you, using only your public website and what four AI assistants say when asked questions your patients ask. We didn't access anything of yours, and we're not asking you to connect an account, install anything, or give us a password. Every finding quotes the engine word for word, so you can ask the same question yourself and see the same thing.
Every finding prints the exact question, the engine and the date it was asked. Type that question into the same assistant and read the answer. AI answers vary from run to run, which is why we ask each question three times per engine and report how many of those answers got the fact wrong, not just one screenshot.
It happens, and it is a good result. The report says so plainly, with the number behind it — how many claims we checked, across which engines — so "no contradictions" reads as the clean bill it is rather than as a blank page.
We write them: the corrected listing text, per-location structured data for your pages, and an FAQ answer in the words people ask. You, or whoever runs your website and listings, publish them — we never need access to your accounts. The next cycle checks whether the engines picked the correction up, and dates it.
Yes. The Single program covers one location for $799 per month, with the same checks as a group, month to month and cancellable from the billing portal.
We keep your business details, the facts you declare and the AI answers we sample, because trends over time are the product. Payments go through Stripe and we never see your card number. We do not sell personal data, and your business name appears only in the questions we put to the AI providers. The details are in our privacy policy.
Anyone can cherry-pick one lucky AI answer. Leolens runs every question multiple times per engine and reports mention rates — "named in 63% of 300 sampled answers" — with run counts on every figure. API-based sampling approximates a neutral, logged-out view; personalized results vary, and we say so. Every report states its engine list, sample size and date, and is stamped with the engine version that produced it.