Before the score: check the run
Open the run detail first. The run determines what was measured:- Preset. A
smokerun (2 prompts) and astandardrun (150 prompts) measure different things. Do not draw conclusions from a smoke run. - Engines. If a key engine was not included in the preset or failed, its absence is a measurement gap, not zero visibility.
- Prompts. If the active prompt set is small or skewed to one intent, certain visibility dimensions are undersampled.
- Date. A run from three weeks ago measures a different state than yesterday’s.
Reading the GEO score
The headline score matters less than the components and trend.Share of voice: the competitive signal
Share of voice (weight 0.35) is the most informative single component. It answers: compared to the tracked competitors, how much of the answer-engine conversation belongs to this brand? A competitor with high share of voice may not have better content. They may have more content in the retrieval index, older domains with more backlinks, or better coverage in the engines’ training data. Use the per-engine breakdown to see which engines favour which competitors. A low share of voice with high visibility means the brand is mentioned frequently but so are competitors. The market is noisy. Improvement requires differentiation or more authoritative content.Which recommendations to act on first
The platform ranks recommendations by evidence weight. The highest-impact recommendations are typically: Technical SEO issues. If a crawler cannot reach the brand’s content, no prompt-driven measurement will improve. Fix robots.txt, sitemaps, and canonical URLs first. Content gaps. Engines frequently surface a category but omit the brand. This indicates the brand’s content is not in the retrieval index or is not authoritative enough to be cited. Create or improve content for the question categories that competitors dominate. Structured data. Schema.org markup improves retrieval relevance. Product, FAQ, and organization schema are the most directly impactful for GEO. Citation weakness. If the brand is mentioned but not cited, the engine is using the brand’s name from some source but not attributing it to brand-controlled content. This usually means third-party sites are the engine’s information source rather than the brand’s own pages.What to expect in 30 days
GEO changes more slowly than SEO in some dimensions and faster in others:- Retrieval index updates. New or improved content can appear in engine retrieval indexes within days to weeks. This depends on the engine’s crawler schedule.
- Model updates. Base-model training data changes only with new model versions. A knowledge-cutoff gap cannot be closed by content changes — it requires the engine provider to retrain.
- Competitor changes. A competitor’s content update or PR event can shift share of voice within days.
- Engine configuration changes. An engine changing its grounding strategy, retrieval source, or system prompt can change visibility for every brand in a category overnight. The platform detects these as unexplained shifts.
Common mistakes
Over-reading a small sample. A score from a 20-prompt run is directional, not definitive. Then value on each component tells you how many observations support it.
Ignoring the per-engine view. A high aggregate score driven entirely by one engine is fragile. If that engine changes its configuration, the score may drop.
Attributing causation to correlation. A score increase after publishing a blog post is not proof the post caused the increase. Other factors — competitor changes, engine updates, seasonal search variance — may have contributed.
Looking only at the headline. The components tell a more useful story than the combined score. A brand with visibility 50 but share of voice 10 is present but not competitive.
See GEO Score for the exact formula and Visibility metrics for the metric definitions.