Understanding the five score signals

Your GEO score blends five distinct signals. Reading them separately tells you where to focus — a low score for one reason needs a very different fix than another.

Mention rate

The share of prompts where the AI names you at all. This is the foundation: if you're never mentioned, position and sentiment don't apply. A low mention rate usually points to a visibility or coverage problem — the model simply doesn't have enough reason to bring you up.

It's also the signal that moves first. When something you've fixed starts working, mention rate reacts before your overall score does.

Position

When you are mentioned, how high in the list you appear. Being named first is worth far more than being last, because users rarely read past the first couple of suggestions. A healthy mention rate with poor position means you're on the radar but losing the ranking battle to competitors.

Position leans deliberately on your benchmark prompts — the direct "who is the best…" comparisons — which count for 70% of this signal, with your other prompts making up the rest. Those are the questions that actually decide a shortlist, so not being named in one at all counts as a zero rather than simply being skipped.

Sentiment

How the model talks about you — confident and positive, neutral, or hesitant. Sentiment reflects the quality of your mentions. Strong reviews, clear differentiators, and consistent messaging across the web push this up.

Sentiment is scaled by your mention rate, so it can't be gamed by a single glowing answer: being spoken about warmly in the one prompt that names you doesn't outweigh being absent from the rest.

Coverage

How many of the AI models mention you, rather than just how often. Appearing in all three is worth more than dominating one, because you don't choose which assistant your customer opens. Each model has its own training data and its own habits — a business that's well known to one can be invisible to another.

Coverage is also scaled by mention rate, so it measures breadth of real presence rather than a single lucky answer. If your coverage is low while mention rate is healthy, look at which model is missing you and check whether it can reach your site at all.

Citations

Whether the AI links to your own domain when it answers. This is the strictest signal: naming you doesn't count, and neither does a link to your Google Business listing, a directory, or a review site. It has to be your site.

Citations are only measured when the scan ran with live web search — a model answering purely from memory could never link to anyone. If none searched there's nothing to measure at all, and if only a handful did, the proportion is too thin to trust: a single link across four answers would swing it by twenty points. So citations earn their full share only once there are enough grounded answers behind them, and any weight they don't carry is shared across the other four signals rather than counted as a zero against you.

Read them together. Low mention rate is a coverage problem. High mention rate with low position is a competitive problem. Weak sentiment is a reputation problem. Low coverage is a per-model blind spot. Low citations mean AI knows you but isn't sending anyone to your site.