Three years ago you optimized for Google. Two years ago you started optimizing for "answer engines." Today, a marketing director sits at her desk and types her own company's name into ChatGPT — and reads, with a slow and dawning horror, a confident summary of a product that her company stopped selling in 2022.
She does this on a Tuesday. Nobody had told her. Nobody is checking. There is no dashboard, no Slack alert, no quarterly review that surfaces it. The AI is the most-trafficked search interface in human history and it is repeating, with the authority of a chess engine, a lie about her company. And there is no one whose job it is to notice.
Otterly counts whether your brand is mentioned. Profound counts how often. SEMRush counts the prompts. Ahrefs counts the keywords. All of them measure presence. None of them measure truth. Visibility without accuracy is the same as a customer review with the rating stripped off — you know someone said something, but you don't know if it was the thing you wanted them to say.
This is the problem misquoted exists to solve. We ask ChatGPT, Gemini, and Perplexity the same questions a buyer would ask — about your pricing, your products, your founders, your competitors — and we read the answers back. We check whether the models agree. We check whether they're right. We tell you, in plain English and on a single page, where the picture has gone sideways.
We didn't invent this work. Fact-checking has existed since there has been print, and the editorial discipline of a magazine like the FT or The Economist is older than most software companies. What's new is that the thing being fact-checked is a probabilistic text engine, and the audience being misled is, increasingly, every customer you have. The tools should be more like a fact-checking desk and less like a marketing dashboard. That is what we are trying to build.
§ 1 — What we believeThree tenets, in order of priority.
-
01
Accuracy comes before visibility.
Being mentioned correctly is more valuable than being mentioned more often. A confident wrong answer is worse than no answer. We will always weight accuracy above mention-count, and we will always show you the exact words the model used.
-
02
Consensus is the cheap signal. Truth is the expensive one.
If three models agree, the picture is consistent. That doesn't mean it's true — three models can be wrong together, especially when they read the same outdated page. The free and $49 tiers measure consensus; the $199 tier verifies it against your real website. Both numbers matter. We never collapse them.
-
03
Drift, not snapshots.
What a model says about you today is less interesting than what it stops saying tomorrow. Every customer who buys a report keeps a free background watch. If the answer changes, you find out before your customers do.
§ 2 — Why this matters nowThe reading layer changed.
Search did not disappear. The reading layer did. For twenty years your customer typed a question, scanned ten blue links, and made a judgment. That layer of judgment has been pulled out. Where there were ten links there is now one paragraph. That paragraph was generated by a model that read your llms.txt — or didn't — and a Wikipedia entry from 2021, and a Reddit thread arguing about your pricing, and a SEC filing from a similarly-named company in Ohio.
"The model is the new front page. And nobody is fact-checking the front page."
This is not a hypothetical. We have run scans on Fortune 500 companies where three out of three models confidently misstated the year the company was founded. We have run scans on Series B SaaS where the AI quoted, verbatim, a pricing page that was retired six quarters ago. The errors are everywhere, and they are confident, and they are quoted as fact in the place your customers now go to make decisions.
If the AI is the front page, then somebody has to be the fact-checker. Otherwise it's a misprint that runs forever — only the misprint is dynamic, and personal, and rewritten every time a model is retrained.
§ 3 — What we will not doThe discipline of restraint.
An editorial product is defined as much by what it refuses as by what it ships. There are a half-dozen things every other tool in our category does that we will not. We are writing them down so we can be held to them.
- No fake urgency We will not invent countdown timers, "24-hour discounts" that reset on refresh, or "3 people are looking at this" badges. Either the report is worth $199 or it isn't.
- No score inflation Our scores are blunt. If your brand is failing, the number will say so in red. We will not soften the verdict to keep you on the page, and we will not lower the bar to make the average customer feel better.
- No vanity metrics We will not invent a metric for the sake of having one. Every number we publish corresponds to a model's verbatim answer or a verifiable fact about your site. No "AI sentiment index," no "brand momentum quotient," no proprietary scoring you can't audit.
- No system-prompt theater We do not coach the models. We do not feed them role-play, chain-of-thought scaffolding, or hidden context that would inflate their performance. We ask the question your customer would ask, and we report what comes back. The point is to measure reality, not to stage it.
- No upsell hostage-taking The $199 report is complete and satisfying. We are not going to withhold a feature you need so we can sell it back to you in a subscription. The drip campaign exists to show you when something changed — not to fence off the data you already bought.
§ 4 — The promiseWhat a customer can expect.
You will type your domain into a box. Ninety seconds later you will see a number, a worst-quote, and a list of fixes. That is the free tier and it will always be free. If you want more — every question, every model, every verbatim answer — the report is $49 and it will always be $49. If you want the same work but cross-checked against your real site, with a correction playbook, that's $199. If you want all of it on a rolling basis with email when the answer flips, that's $399 a month.
None of those numbers will change quietly. None of them will be raised on you mid-subscription. None of them will be hidden behind a "contact sales" button until you cross some arbitrary line. We are not building a maze. We are building a fact-checking desk.
That is the contract. You can hold us to it.