Free AI Visibility Checker

What a Visibility Check Actually Measures

The useful question is not "am I in the training data". It is "when a buyer in my category asks for help, does my name come out". Those are different, and only the second one costs you revenue. A check answers the second by generating prompts in the shape buyers use, sending them to each model, and parsing what comes back. We generate across eight prompt categories, because a brand can be strong in one and absent in another:

• Best-of questions ("best project management tool for a small team")
• Head-to-head comparisons ("Linear vs Asana")
• Price questions ("how much does it cost")
• Feature questions ("which tool has offline mode")
• Review and trust questions ("is it any good")
• Industry-specific questions, tuned to your vertical
• Trending questions, dated to the current year
• Long-tail questions, the specific ones with buying intent

Why Two Brands With the Same Mention Count Score Differently

Counting mentions is the naive version and it misleads constantly. A brand named first, warmly, in the exact context it sells into is in a completely different position from one named last as an afterthought, even when both were "mentioned once". Our scorer weights six factors rather than counting, and the weights are deliberate: position and sentiment together carry 40%, more than raw mention count at 25%, because how you are named matters more than how often.

• Mention count, 25%: how many prompts named you at all
• Position, 20%: where in the answer, since the first name carries most of the click
• Sentiment, 20%: recommended, listed neutrally, or named as the thing to avoid
• Source quality, 15%: whether the model cited authoritative pages when it named you
• Context match, 10%: whether you were named for what you actually do
• Competitor gap, 10%: your visibility relative to the rivals in the same answers

Platforms Are Not Worth the Same

Averaging your scores across assistants flatters whichever one you happen to do well on. Reach differs by an order of magnitude, so a per-platform score has to be weighted before it is combined. We apply a multiplier per platform, with ChatGPT weighted highest at 1.2 and smaller assistants below 1.0. The practical consequence: a strong Perplexity score does not rescue a weak ChatGPT score, and any tool that tells you otherwise is averaging away the thing you needed to know.

The Hard Part Is Detection, Not the API Call

Sending a prompt to a model is trivial. Deciding whether the answer mentioned you is not, and it is where most checkers quietly fail. Brand names are frequently ordinary English words, so a naive substring search reports a mention every time the model uses the word normally. Names get pluralised, possessive, hyphenated, or abbreviated. A competitor with a similar name gets counted as you. Detection has to handle all of that before position and sentiment mean anything, because a false positive does not just add noise, it tells you that you are fine when you are invisible.

What You Do With the Result

A number on its own changes nothing. The output that matters is the gap list: the specific prompts where a competitor was named and you were not, and the sources the model leaned on when it named them. That converts directly into work. If comparison prompts consistently pick your rival, you are missing comparison content. If review prompts skip you, you are missing third-party reviews. If you are named but for the wrong use case, your positioning is reaching the model in a distorted form and that is a messaging problem, not a link-building one.