What is a GEO score and how is it calculated?
Your GEO score — short for Generative Engine Optimization score — is a single number from 0 to 100 that captures how visible your business is when people ask AI assistants for recommendations. Think of it as your "search rank," but for ChatGPT, Claude, and Gemini instead of Google.
The higher your score, the more often AI names you, the higher it places you, and the more positively it talks about you when a potential customer asks a relevant question.
Why it matters
Buyers increasingly skip the search engine and ask an AI assistant directly: "Who's the most reliable used-car dealer in Athens?" The model replies with a short list — usually one to three names. If you're not on it, you're invisible, and traditional analytics can't see it happening. The GEO score makes that invisible channel measurable.
What goes into the score
For every scan, GEOpulse runs a set of real prompts your customers would ask across the AI models included in your plan, and records every answer. Five signals are then combined:
Each signal is scored 0–100 on its own, then blended at the weight shown.
- Mention rate — 35%. Out of every prompt we run, in what share does the AI name your business at all? This is the core signal: everything else only matters once you're in the answer.
- Position — 25%. When you are named, how high up? Being first carries far more weight than being last. This leans on your benchmark prompts — the head-to-head "who is the best…" questions — which make up 70% of the position score, with your remaining prompts making up the other 30%. Not being named at all in a benchmark prompt scores zero for that prompt, which is deliberate: those are the questions that decide a shortlist.
- Sentiment — 15%. How the AI talks about you. A confident recommendation scores higher than a lukewarm or hedged mention. Sentiment is multiplied by your mention rate, so it only counts where you're actually named — glowing praise in the single answer that mentions you doesn't outweigh silence everywhere else.
- Coverage — 15%. Across how many of the AI models you show up in. Appearing in all three is worth more than dominating one, because you don't get to choose which assistant your customer uses. Coverage is also scaled by mention rate.
- Citations — 10%. Whether the AI links to your own site when it answers. A link to your Google Business listing, a directory or a review site doesn't count — this is specifically about your own domain. Citations are only measurable when the scan ran with live web search, and they only carry their full weight once enough answers were grounded for the proportion to mean anything — with a handful of eligible answers, one link would swing the result by twenty points or more. Whatever weight citations don't carry is shared across the other four signals, never counted as a zero against you.
Scores are averaged across every AI model on your plan, so a strong showing on one can't hide a blind spot on another. You'll also see the per-model breakdown on your dashboard.
Reading your score
As a rough guide:
Low visibility — AI rarely mentions you. Big opportunity to climb.
Moderate — you show up sometimes, but competitors often rank above you.
Strong — AI consistently names you, high up, in a positive light.
Why your score didn't jump after you fixed things
The number on your dashboard is deliberately not a single scan's result. Every scan produces a raw score; what you see is a weighted average of your last four:
On your very first scan there's no history to blend, so your score is simply that scan's result.
This is why a single great scan can't spike your score and a single bad one can't crater it. AI answers vary a little run to run, and a score that swung ten points on that noise would be useless for deciding what to do next.
The trade-off is that a real improvement shows up over two or three scans rather than all at once. If you fixed something significant and the number moved less than you expected, check the individual signals on your dashboard — mention rate is usually the first thing to move, and it moves before the overall score catches up.
Why your score changes
AI models update constantly, and so does the web they learn from. New reviews, fresh content, schema markup, and press coverage can all lift your score over time — while a competitor's push can pull your relative position down. That's why GEOpulse rescans on a schedule and tracks your score as a trend, not a one-off number.
Want to raise your score? Head to Recommendations in your dashboard — each item is impact-scored and tied to the exact prompts you're currently losing.
Was this article helpful?