GEO Metrics

1. Why GEO requires different metrics

Traditional SEO KPIs such as CTR, average SERP position, and impressions assume that users see a ranked list of links and decide whether to click. Generative answers break that assumption.

This new environment has three defining features:

1. Multi-source aggregation: A single answer can cite 3–10 sources, so the question “Where do I rank?” loses its meaning.
2. Zero-click behavior: Users can get the answer without visiting your page, which makes traditional CTR unmeasurable.
3. Separation of mentions and citations: A brand can appear in an answer as a mention without receiving a link or citation.

Measurement therefore shifts from asking whether the user reached your page to asking whether the AI used your content to compose its answer. See Generative Engine Optimization and Zero-Click Search for the broader context.

Measurement choices guide investment decisions. A consistent set of metrics also makes GEO performance easier to connect to business value, as explained in GEO ROI Models.

2. The ten metrics at a glance

The table below provides a quick comparison of the 10 core GEO KPIs.

A note on naming and origin: There is no authoritative, industry-standard set of GEO KPIs. These 10 metrics are GEO Wiki’s synthesis of terminology used by Profound, Otterly, Ahrefs, BrightEdge, and Similarweb. The academic literature (Aggarwal et al. 2024) defines only three metrics under different names: Word Count, Position-Adjusted Word Count, and Subjective Impression. That paper is not the source of this 10-metric taxonomy. Other published lists, such as Search Engine Land’s “8 GEO metrics,” choose a different set. The terminology here follows commercial tools so that readers can apply it to the products they use.

Visibility Score, Citation Rate, and Answer Inclusion Rate are often confused. Visibility Score (§3.1) is the vendor’s headline metric. It asks whether the brand appeared at all, as either a mention or a citation, and is often a composite. Citation Rate (§3.2) counts only answers that explicitly cite your domain. Answer Inclusion Rate (§3.7) indicates whether the brand appears in an answer for each query. For many vendors, “Visibility Score” is Answer Inclusion Rate under a different name, sometimes with position weighting. Always check the formula.

3. The ten core metrics in detail

3.1 Visibility Score

Definition: Visibility Score is the headline metric used to show whether a brand appears in AI answers at all. It most often represents the fraction of tracked prompts or answers in which the brand appears as either a mention or a citation. Some vendors report a composite index based on appearance and position, and sometimes sentiment, rather than a raw percentage.

Formula for the common percentage form:

Visibility Score = prompts_where_brand_appears / total_tracked_prompts × 100%

Otterly publishes the following composite formula for its “Brand Visibility Index” (KPI page):

Brand Visibility Index = 10 + ((5 − avgPosition) / 4) × 90

Unit: A percentage in the raw form or a 0–100 index in the composite form.

Use: This metric summarizes the brand’s presence across all tracked AI answers in a single number for leadership reporting.

Vendor variations:

• Profound (Answer Engine Insights) presents “Visibility score and share of voice metrics” as headline KPIs but does not publish the Visibility Score formula.
• Otterly publishes a composite “Brand Visibility Index” based on Brand Coverage and Average Position, using the formula above.
• Semrush, Quattr, and others offer proprietary “AI Visibility Score” metrics that normalize several factors. There is no universal formula (industry overview).

SEO equivalent: Approximately the headline search visibility or visibility index reported by SEO suites such as Sistrix and Semrush.

Key distinction: Visibility Score includes mentions and therefore reflects any appearance, while Citation Rate (§3.2) requires an explicit citation. The two are not interchangeable. When “Visibility Score” is simply an appearance rate, it measures the same construct as Answer Inclusion Rate (§3.7) under a different label. Always read the vendor’s formula before comparing results.

Pitfalls:

• There is no standardized formula, so raw percentages, composite indices, and position-weighted versions are not comparable.
• Composite indices such as Otterly’s combine several inputs into one number. Examine the inputs before deciding what action to take.

3.2 Citation Rate

Definition: Citation Rate is the fraction of AI answers about a topic that cite your domain through explicit source attribution.

Formula:

Citation Rate = cited_answers / total_answers_about_topic × 100%

Unit: A percentage, typically 0%–30%. A rate above 30% usually means the topic is too narrow.

Use: This metric establishes a citation visibility baseline for one brand within one topic. It answers the question, “What fraction of AI answers about this topic cite my domain?”

Vendor variations:

• Otterly measures this as “Domain Coverage”: prompts that cite my domain / all prompts in selected time window (KPI definitions page). Otterly’s separately named “Domain Citation” is not this rate; it is an absolute count (see §3.6).
• Ahrefs Brand Radar defines “Citations” as “the AI results that cite the entity at least once as a source” (see Brand Radar help). This matches the Citation Rate definition above.
• Profound’s headline “Visibility Score” is not this metric because it includes mentions (see §3.1).

SEO equivalent: Approximately CTR, although the analogy is imperfect because the denominator is answers about a topic rather than search impressions.

Pitfalls:

• The denominator is highly sensitive to the choice of topic or query set. The same brand can show Citation Rates that differ by 5–10 times across query sets.
• The source of the sample and the time window must be public for the result to be reproducible.

See AI Citation Tracking for the operational process and Citation vs Mention for the distinction between the two outcomes.

3.3 Citation Share

Definition: Citation Share is your share of all citations within a fixed competitor set.

Formula:

Citation Share = your_citations / total_citations_in_competitor_set × 100%

Unit: A percentage.

Use: This metric measures competitive visibility. It answers the question, “Within this niche, what share of AI citations am I receiving?”

Vendor variations:

• Commercial naming is inconsistent. Ahrefs’ “AI Share of Voice” is similar to Citation Share but uses different weighting (see §3.4).
• Profound includes this concept in its “Share of Voice” rather than reporting a separate Citation Share.

SEO equivalent: Approximately backlink share, meaning your share of backlinks within a keyword set. Both are relative measures.

Key distinction from Citation Rate: Citation Rate uses all answers as its denominator and measures absolute visibility. Citation Share uses all citations as its denominator and measures relative competitive position. An increase in Citation Rate does not imply an increase in Citation Share because the entire category may be expanding.

3.4 Share of Voice (SOV)

Definition: Share of Voice is your share of brand appearances across AI answers for a topic and competitor set, including both citations and unlinked mentions.

Formula published by Otterly (KPI definitions page):

SOV = number of my brand mentions / total number of all brand mentions × 100%

Unit: A percentage.

Use: This metric measures how often the brand appears in category discussions, whether cited or merely mentioned. It answers the question, “When AI discusses this category, what fraction of those conversations include my brand?”

Vendor variations:

Ahrefs’ impression weighting is a meaningful methodological choice. Its SOV reflects potential exposure rather than raw mention count, so the same mention receives more weight in a topic with high search volume. This makes the metric a more accurate proxy for the commercial value of AI visibility.

SEO equivalent: Approximately the PR industry’s Share of Voice, narrowed from all media to AI answers.

Key distinction from Citation Share: SOV includes unlinked mentions, while Citation Share counts only citations with explicit attribution. See Citation vs Mention and Brand Mentions.

3.5 Average Position: the most contested metric

Average Position is the most ambiguous KPI in the GEO toolkit. Values from different vendors cannot be compared unless each vendor’s definition is clear.

Three competing definitions in active use:

Use A. Citation Order by default. It has the most stable meaning, matches the structure returned by platforms, offers the greatest cross-platform comparability because most major engines expose citation order, and provides the basis for First-Cite Rate (§3.8).

Every report must still explicitly state which definition it uses, whether A, B, or C. Without that qualification, values from different tools are not comparable.

Formula under recommended definition A:

Average Position = mean(citation_rank) for answers where the brand
                   appears as a cited source

Unit: 1.0–N.0, where a lower value is better and N is typically 8–10.

Use: This metric compares authority. The difference between the first and third citation positions is meaningful.

Perplexity API behavior: According to Perplexity’s Chat Completions API, the citations field returns an array of URLs, and inline [1][2] annotations correspond to that array’s order. The public documentation does not explicitly state whether the array is ranked by relevance, so verify that assumption before treating position as a quality signal.

Vendor evidence:

• Otterly’s “Avg. Brand Position” formula is sum of positions of the brand mentions across prompts / number of prompts where the brand appeared. Its documentation does not state explicitly whether “position” means A, B, or C. Given its use of response-text sampling, the metric is closer to B.
• Profound, Ahrefs, and BrightEdge do not publish a precise definition of Average Position.

SEO equivalent: Approximately Average SERP Position, although the GEO version is a much weaker signal because AI answers are not paginated rankings.

Pitfall: A “position 1” measured under definition A and a “position 1” measured under definition B describe entirely different things. Combining them produces meaningless comparisons.

See GEO Glossary for the term-level definition.

3.6 Mention Frequency

Definition: Mention Frequency is the total number of brand appearances in a sample of AI answers over a specified time window. It is an absolute, unnormalized measure.

Formula:

Mention Frequency = count(mentions) / time_window

Unit: Mentions per month.

Use: This metric confirms brand presence at an early stage and monitors long-term trends.

Vendor variations: Nearly every GEO tool reports this metric under one name or another. The main difference is sampling, including which queries and engines the tool samples and how often.

SEO equivalent: Approximately brand search volume, although the data comes from AI answers rather than search boxes.

Pitfalls:

• Absolute values are not comparable across brands because large brands naturally accumulate more mentions.
• Query-set bias can easily distort the result. A set of 100 carefully selected queries does not represent an entire domain.

Academic note: Aggarwal’s “Word Count” metric is the closest academic analog, but it measures normalized word count in cited sentences rather than raw mentions.

3.7 Answer Inclusion Rate

Definition: Answer Inclusion Rate is the fraction of queries in a target set for which the AI answer mentions or cites your brand at least once.

Formula:

AIR = queries_where_brand_appears / total_queries × 100%

Unit: A percentage.

Use: This metric measures coverage at the query level, which provides a more detailed view than SOV. It answers the question, “Of the N queries I track, how many produce an AI answer that includes my brand?”

Vendor variations:

• Otterly calls this “Brand Coverage”: prompts that mention my brand / all prompts in selected time window (KPI page).
• Most other vendors include it within SOV rather than reporting it separately.

SEO equivalent: Approximately keyword coverage rate, which is the number of keywords ranking in the top N divided by the total number of tracked keywords.

Key distinction from SOV: AIR is a binary measure of whether the brand appears. SOV measures the brand’s share of appearances.

3.8 First-Cite Rate

Definition: First-Cite Rate is the fraction of answers that list your domain as the first-cited source among all answers that cite you.

Formula:

First-Cite Rate = first_cited_answers / cited_answers × 100%

Unit: A percentage.

Use: This metric measures authority. Being cited first implies that the AI prefers your source.

Vendor variations:

• This metric requires a platform to expose citation order. The Perplexity API exposes citation order directly (see §3.5), while ChatGPT and Claude expose it only in some response modes.
• Few commercial vendors report this as a standalone KPI. It is usually included in Average Position under definition A.

SEO equivalent: Approximately the rate of ranking first in a SERP. The signal is weaker in an AI answer because the difference between the first and second citation positions is smaller than the difference between the first and second traditional search results.

Pitfall: The metric is noisy when the underlying citation count is small. A First-Cite Rate that rises from 20% to 40% from one week to the next may be statistically meaningless if you were cited only 5 times.

3.9 Brand Sentiment

Definition: Brand Sentiment measures the net emotional tone of an AI engine’s description of your brand, typically as the balance of positive and negative references.

Formula published by Otterly (KPI definitions page):

Brand Sentiment = (positive_mentions − negative_mentions) / total_mentions × 100

Unit: A score from −100 to +100, or a positive, neutral, or negative label.

Use: This metric measures the quality of brand perception rather than presence alone. A brand can have high Visibility but negative Sentiment, as in “X is overpriced.” Sentiment captures what volume metrics miss.

Vendor variations:

• Otterly reports “Brand Sentiment” using the formula above.
• Profound (Answer Engine Insights) offers “Sentiment & Keyword Insights” to show how AI describes the brand, but its formula is not public.
• Sentiment is generated by the model, so it varies with both the engine and the way the prompt is framed. Report both variables.

SEO equivalent: Approximately the brand sentiment used in PR and social listening, narrowed from all media to AI answers.

Pitfalls:

• The result is highly sensitive to prompt framing. A neutral query and a “problems with X” query can produce opposite sentiment for the same brand.
• Small mention counts can make changes in sentiment statistically meaningless, as with First-Cite Rate.

3.10 Source Diversity Score

Definition: Source Diversity Score measures how many distinct AI engines have cited your content relative to the total number of engines you track.

Formula:

Source Diversity Score = distinct_engines_citing_you / total_engines_tested

Unit: A percentage or an absolute count, depending on the tracking scope.

Use: This metric measures dependence on a single platform, such as appearing primarily on Perplexity.

Vendor variations:

• The set of tested engines differs by vendor. Profound covers 9 or more engines, including ChatGPT, Perplexity, Claude, Copilot, Google AIO, Gemini, Grok, Amazon Rufus, Meta AI, and DeepSeek. Ahrefs, Otterly, and Similarweb each cover 6.
• The engine set is itself a variable and must be reported explicitly.

SEO equivalent: There is no direct equivalent. In traditional SEO, Google’s dominance made engine diversity largely irrelevant. GEO spans multiple platforms, which makes this metric essential.

See Multilingual GEO for differences between Chinese and English engine coverage, and Generative Engine for an overview of the engines themselves.

4. Vendor definition matrix

The table below summarizes how 5 major Western vendors handle the 10 metrics. Only Otterly and Ahrefs publish complete formulas; the other vendors provide marketing descriptions without full formulas.

Key observations:

• Published formulas: Otterly publishes full formulas for 7 or more metrics, including Brand Visibility Index and Brand Sentiment. Ahrefs publishes formulas for 4 metrics and provides a methodology article.
• Marketing descriptions: Profound, BrightEdge, and Similarweb provide marketing descriptions but do not publish full formulas.
• Methodological differences: Similarweb is the only vendor that measures actual referral traffic from AI bots through a panel, while the others sample AI responses. Ahrefs is the only vendor that applies impression weighting, and BrightEdge primarily covers Google AI Overviews.

No vendor is recommended on the strength of this matrix alone.

5. Mapping to traditional SEO metrics

The table below maps the 10 GEO KPIs to their nearest SEO equivalents for practitioners who already work with SEO metrics.

See SEO vs GEO for a fuller comparison.

6. Choosing metrics by GEO maturity stage

Each GEO maturity stage calls for a different set of metrics. The mapping below follows the 5-level GEO Maturity Model.

See the metric snapshots section of the GEO Audit playbook for the operational steps.

7. Common pitfalls and confusions

Check for these seven common errors before publishing a report or interpreting vendor data:

1. “Citation rate” is used for two different measures in casual usage. It can mean the absolute Citation Rate or the relative Citation Share. Always state the English metric name explicitly.
2. Average Position cannot be compared across tools without an A, B, or C definition tag. This is the most frequent error.
3. Sample bias can change results by 5–10 times. Publish the sampling method and the query set used.
4. Time-window bias can make 30-day and 7-day results differ substantially. AI answers change quickly, so always report the time window.
5. Chinese and English query results cannot be added together. AI engines use very different source pools for each language.
6. SOV and Citation Share count different outcomes. Including or excluding mentions can change results by a factor of 2–3. Otterly's SOV, Ahrefs's AI SOV, and BrightEdge's SOV are not equivalent.
7. Visibility Score is not standardized. Some vendors use it for a raw appearance rate, which is Answer Inclusion Rate. Others use a composite of appearance and position, such as Otterly’s Brand Visibility Index. It is not Citation Rate. Before comparing tools, check whether the Visibility measure counts mentions or only citations.

8. Further reading

• For related concepts, see Citation vs Mention, Brand Mentions, and GEO Glossary.
• For practical guidance, see AI Citation Tracking to collect these metrics, GEO Audit for the end-to-end audit process, and GEO Maturity Model for the metrics used at each stage.
• For the business context, see GEO ROI Models, which connects metrics to business outcomes.
• For the academic reference point, see Aggarwal et al. (2024): GEO: Generative Engine Optimization, a summary of the KDD ‘24 paper and its key data. The paper defines 3 metrics, not this taxonomy.

References

Academic:

• Aggarwal, P. et al. (2024). GEO: Generative Engine Optimization. KDD ‘24. arXiv:2311.09735 and ACM DL

Vendor documentation (as of 2026-05):

• Otterly.ai: Brand Report KPI Definitions
• Ahrefs: Brand Radar Help and Brand Radar Methodology
• Profound: Answer Engine Insights
• BrightEdge: Share of Voice in 2026
• Similarweb: GenAI Intelligence

API reference:

• Perplexity: Chat Completions API Reference

Industry overview:

• TigerTracks: The Definition of AI Visibility Score (2026)