One number to steer your AI visibility
Visibility, authority and sentiment compressed into a 0-100 index, averaged across every engine. Track it weekly, benchmark it against competitors, prove progress to your clients.
- Cancel in one click
- Results in minutes
- 10 AI engines covered
73 / 100
A number you can actually defend
The GEO Score is the average of your per-engine scores, smoothed across iterations to absorb the randomness of LLMs. It is the single metric you report up, and the one your team optimizes against.
Per-engine, then averaged
Each engine is scored in isolation, then combined. A great ChatGPT score and a weak Perplexity one no longer hide behind a single blurry number.
Three weighted pillars
The score rests on one explicit formula: visibility multiplied by 0.4, plus authority multiplied by 0.3, plus sentiment multiplied by 0.3. Each pillar is computed on its own zero to one hundred scale before weighting, so you can always see which one is dragging. Visibility is the share of answers containing at least one mention of your brand across the fifty queries configured in the project: mentioned in thirty answers out of fifty and your raw visibility is sixty. Detection is not an exact search for your name, NLP also catches implicit mentions, paraphrases, references to your URL and mentions of specific products. Authority grades how you are cited, across four positions with published weights: primary recommendation counts 100 percent, a place in the top three counts 75, a contextual mention among other options 40, an incidental mention 15. Your authority score is the weighted average of those positions across all your mentions. Sentiment sorts each mention into positive, neutral expert, negative or not relevant. Positive and neutral expert lift the score, negative penalises it, not relevant is discarded.
Non-determinism handled
Large language models are not deterministic: the same question can cite you on one call and omit you on the next identical call. Citeme answers that with three mechanisms. Each query is executed several times on each model, and the score is computed on the average of those iterations rather than on a single execution. The displayed score is then smoothed over time, a blend of the current run and previous audits with a decreasing weight, which removes the cliff edges caused by model variance. Finally the platform publishes significance thresholds so you know when to act: around five points of movement is normal fluctuation and calls for nothing, ten points is a notable variation worth checking against your recent content changes, twenty points or more is a real change and usually means a competitor gained or lost a major semantic advantage. The trade-off is explicit. Smoothing buys stability at the cost of reactivity, and the documentation recommends running an audit at least every two weeks to keep the picture accurate.
Four readable tiers
The zero to one hundred scale is cut into four tiers, and each one carries a different priority rather than just a label. Zero to twenty is invisible: the models do not know your brand or judge it irrelevant to your queries, the typical position of a company that has never worked on GEO, and the absolute priority is enriching factual content and adding structured data. Twenty to fifty is emerging: the models know you but do not recommend you systematically, you surface on some models and not others, or only on brand queries, which is the moment to close the semantic gaps. Fifty to eighty is visible: you are cited regularly and often well placed, and the job shifts to moving from one option among several to the sector reference, through thought leadership, social authority and full semantic coverage. Eighty to one hundred is reference: you are cited first and recommended proactively, and the work becomes defending the position. The tier describes the blended score, not any single engine, so a strong average can still hide one model where you are absent.
Benchmarked by default
Your score sits next to your competitors on the same prompts. A 73 means nothing in a vacuum and everything next to a market leader at 68.
Trend over time
The dashboard charts your score week after week, with the delta, so a single optimization can be tied to a measurable movement.
Where the 100 points come from
Being mentioned is not enough. The GEO Score weights where you appear and how. A primary recommendation is worth far more than a passing mention at the bottom of a list, and a glowing description beats a lukewarm one.
That is why the score blends three signals: visibility for frequency, authority for position, and sentiment for tone. Together they answer the only question that matters: when AI talks about your market, are you the answer?
- Visibility, 40%, how often you are cited
- Authority, 30%, your rank in the recommendation
- Sentiment, 30%, positive, neutral or negative tone
73 / 100
0-100
single index
3
weighted pillars
4
readable tiers
weekly
trend tracking
How it works
- 1
The audit runs
Hundreds of AI answers are collected and parsed across the engines.
- 2
Pillars are computed
Visibility, authority and sentiment are scored separately per engine.
- 3
The score lands
The weighted, smoothed index appears on your dashboard with its trend.
Our articles on this topic
All articlesSearchable AI review 2026
What Searchable does, what its own search footprint says about it, and where it fits against the rest of the category. Our score: 3.5 out of 5.
Alexandre TeilletMarketingGetMint review: our analysis of the French GEO platform
Real pricing, the engines actually tracked, a 4 million euro round and measured authority. Our score: 3.8 out of 5.
Alexandre TeilletMarketingBest Shopify apps for GEO in 2026: get your store cited by AI
The Shopify apps that help with GEO, sorted by job: what Shopify already does, AI visibility tracking, structured data, llms.txt and product content. Prices and ratings from the Shopify App Store on October 21, 2026.
Alexandre TeilletGEOGood to know
A mention buried at the bottom of a list is not worth a primary recommendation. The score weights position and tone, not just frequency, so it reflects real influence.
All features
See it on your own market
Test your prompts on the ten AI engines: 7 days for €1, then cancel anytime.