New metrics for the AI era: the GEO metrics that actually matter

GEO metrics measure whether AI systems retrieve, select and cite your content.

Most B2B marketers still measure visibility with a dashboard from 2018: Google positions, organic traffic and click-through rate. Those numbers say less and less.

An AI answer appears above position one, resolves the question, and there’s no click to measure. So you can “score well” on your old scorecard while going invisible in AI search.

The verdict: selection rate and citation share are the 2 AI visibility metrics to start with. Rankings and CTR still matter, but they no longer tell you whether AI picks you.

By new GEO metrics we mean measures that track whether your content gets retrieved, selected, cited and correctly understood by ChatGPT, Perplexity, Gemini and Google AI Overviews.

Want the broader framework first? Read the ultimate GEO guide. This article zooms in on the measurement side.

Curious how AI sees you? Run your page through our free GEO check.

Why are the old metrics becoming misleading?

Because they assume a click, and AI answers increasingly satisfy people without one.

A large part of our classic SEO metrics has degraded into a misleading proxy.

Three metrics that lost their meaning

Here’s where each one falls short:

• Keyword rankings ignore AI answers above the first organic position.
• Organic traffic gets cannibalised by answers that need no click.
• Click-through rate assumes the click is the goal.

The traffic effect is measurable. Pew Research saw Google users click a regular result on 8% of visits with an AI summary versus 15% without.

A link inside the summary itself got clicked on just 1% of those visits. In that March 2025 study, 18% of searches showed an AI summary.

What to measure instead

The old metrics weren’t bad: they were right for traditional search.

But the game has changed and our scorecards haven’t.

So we measure what decides visibility in AI search: selection, extraction, salience and presence.

Google doesn’t split this out for you. AI Overviews and AI Mode traffic is reported in the Performance report, within the “Web” search type, mixed with classic results.

What are the four layers of the AI visibility stack?

Foundation, quality, position and health, where each layer builds on the one below.

You can’t be selected if you aren’t retrieved first.

And your position against competitors means little without healthy underlying numbers.

We’ll walk through the most important ones layer by layer.

Watch out: the thresholds below are working thresholds from our own measurements, not published industry benchmarks. No standard exists yet, so compare against your competitors first.

Layer 1: foundation metrics

These are the base measurements everything else is derived from.

Retrieval rate = queries for which your content is retrieved / total relevant queries.

This is the precondition for everything: if you aren’t retrieved, you can’t be chosen either.

Below 30% usually points to content that isn’t indexed, misses search intent, or lacks authority signals.

Above 60% you’re in good shape.

Selection rate = selections / retrievals.

As far as we’re concerned, this is the core indicator. It measures whether the model actually picks your content from the candidate set.

Below 25% is weak, above 50% is strong.

Low selection despite good retrieval almost always points to quality or extraction problems, or a relevance mismatch.

For example, say you track 40 prompts. You’re retrieved on 20 and cited on 8, so your selection rate is 8 / 20 = 40%.

Citation frequency = citations / total tracked queries.

This combines retrieval and selection into one outcome metric. Our guide to citation rate shows how to count it.

Rule of thumb: citation frequency is roughly retrieval rate times selection rate. Here’s that sum as an example:

Retrieved 50% of the time and cited in 40% of those cases? Your citation frequency sits around 20%.

Layer 2: quality metrics

These measure how efficiently you capture value when you’re actually visible.

Citation coverage = extracted characters / total characters on the page.

How much of your content investment turns into AI visibility?

Below 20% is low. It often points to bloated content, or a structure where the extractable part is hard to find.

Extraction depth = your extracted words / total words in the answer.

When you’re cited, how much of the answer comes from you?

Below 10% you’re a supporting source; above 25% you’re the primary voice. For example, 40 words of a 200-word answer gives an extraction depth of 20%.

Primary source rate = primary citations / total citations.

How often are you the first-cited source rather than a footnote?

Primary status drives brand visibility, click likelihood and authority association.

You can see that ranking in a sources panel. Perplexity shows every page behind an answer:

Perplexity’s sources panel for our own test prompt, October 2026: 10 web pages behind one answer.

Ten sources, one answer. Being source 1 or source 9 is the difference this metric captures.

Layer 3: position metrics

This is your competitive standing. Always compare these numbers with competitors, not with absolute thresholds.

Citation share = your citations / total citations (all sources).

This is the AI equivalent of market share.

In competitive markets, 30% or more is a dominant position. Segment it by topic cluster, platform and period.

Tools now show it as a headline number. Ahrefs, for instance, leads its Brand Radar page with AI share of voice:

Ahrefs Brand Radar’s product page, October 2026. AI share of voice is the first metric it names; the figures are Ahrefs’ demo data.

More on the concept in AI share of voice, and on tooling in our overview of AI visibility tools.

Competitive index = your citations / a competitor’s citations.

Where citation share measures your overall position, the competitive index tracks a head-to-head duel.

Above 1 you’re cited more often; below 1 you’re losing. For example, 30 citations against a rival’s 20 gives an index of 1.5.

Our guide to AI competitor benchmarking sets this up.

Visibility score = a weighted summary of several factors.

A workable formula weights citation frequency (0.3), extraction depth (0.2), primary rate (0.2) and citation share (0.3).

Normalise each component to a 0 to 100 scale first.

Layer 4: health metrics

This layer measures sustainability and direction. It’s where many teams stay blind.

Salience score measures how strongly the model associates your brand with your category.

It’s built from mention probability (the chance the model names you unprompted), recommendation confidence and association clarity.

The question: is your brand anchored in the model, independent of what’s retrieved live?

Freshness index = weighted average age of your visible content, weighted by extraction frequency.

Content that gets extracted often weighs more.

Older than 12 months on average is stale; under 6 months is fresh.

Velocity = (citations current period minus citations previous period) / citations previous period.

Apply it to every metric.

For example, 50 citations last month and 60 this month is a velocity of +20%. Negative velocity is an early warning before it becomes a crisis.

How do you measure at platform, query and page level?

By splitting every aggregate, because aggregated numbers hide the variation you need to act on.

Platform level

Calculate citation share and selection rate per platform.

You can dominate on Google AI Overviews and lag behind in ChatGPT.

A platform divergence index (the standard deviation of your platform metrics divided by the average) shows where platform-specific opportunities lie.

Bing already reports AI citations for you. Here’s our own Bing Webmaster Tools AI Performance report:

Our own Bing AI Performance report, October 2026: 60.3K citations and on average 140 cited pages over three months.

That’s citation frequency for one platform, straight from the source.

Query level

Track selection rate and your rank per query.

That instantly shows which queries to defend (you’re number 1), attack (you’re losing) or conquer (you’re absent).

Page and section level

Calculate a page visibility score per page, plus a section extraction rate: how often a section is extracted per time the page is retrieved.

That tells you which content structures drive extraction and which ones need rework.

Which five brand-focused GEO metrics complement the stack?

Five indicators that measure how a model understands and trusts your brand, not just query behaviour.

We describe them in depth in the 5 core indicators of AI visibility. In brief:

• Prompt Recall Rate (PRR): how often your brand surfaces in category prompts
• Semantic Accuracy Index (SAI): how accurately the model conveys your positioning
• Source Diversity Ratio (SDR): unique authoritative sources versus total references
• Trust Alignment Score (TAS): authority and sentiment of your sources combined
• Model Drift Index (MDI): how fast your representation changes across model updates

What the numbers look like

PRR is the generative equivalent of impression share.

As a rough working range, emerging brands often sit at 5 to 10% and mature brands at 25 to 40%.

Our experience: for our own site, Perplexity named Customer Impact in 16 of 125 agency-choice prompts, up from 7. That’s a PRR of about 13%.

For SAI, ask: does the model call you a “trucking app” or a “fleet intelligence platform”? A low SAI signals narrative drift.

For SDR, more spread means you’re harder to train away.

For TAS, a neutral Wikipedia plus positive G2 reviews plus factual press scores high.

A high MDI means an unstable story and a risk of reputational decay in the model’s memory.

How do you turn metrics into a dashboard and alerts?

Put every number in context, because isolated numbers are noise.

Trends over snapshots

Use a 30-day rolling average to smooth out volatility.

Year-on-year comparisons filter out seasonal effects.

Detect break points and tie them to causes: an algorithm update, a competitor’s move or your own optimisation.

Benchmark against the right group

Compare your metrics with 3 groups:

• The category average
• The leader, and whether that gap closes or widens
• A realistic peer group of comparable players

Set up alerts

A threshold alert (“citation share below 15%”) and anomaly detection (“selection rate 2 standard deviations off trend”) catch problems early.

Competitor alerts are just as valuable when a rival suddenly gains ground.

Two dashboards, two audiences

An executive dashboard keeps it simple: visibility score, citation share, selection rate, trend, plus one insight, one risk and one action.

An operational dashboard shows citation share, selection rate and trend per topic cluster. Our GEO dashboard guide and our tips on reporting GEO to leadership show both.

Connect visibility to visits

Citations matter most when they turn into sessions and leads. Here’s chatgpt.com traffic in our own GA4 property:

GA4 Traffic acquisition in our own property, last 90 days, October 2026: 58 sessions from chatgpt.com, with one day at 5.

That’s small next to the citation count, which is exactly why clicks alone undercount AI visibility. More in AI traffic attribution and measuring AI Overviews traffic.

Link both to your GEO KPIs, so the dashboard ends in leads.

Your first month, step by step

You don’t need the full stack on day one. A simple start:

1. Week 1: pick 25 to 50 prompts per topic cluster
2. Week 2: run them on each platform and log mentions and citations
3. Week 3: calculate selection rate and citation share
4. Week 4: rerun, compare and set your first alert

After that, add one layer of the stack per month.

Velocity only means something from your second measurement onward.

Frequently asked questions

What is the difference between retrieval rate and selection rate?

Retrieval rate measures how often your content is retrieved for relevant queries, so whether you make it into the candidate set at all. Selection rate measures how often the model then actually picks and uses your content.

High retrieval but low selection points to a quality or extraction problem: you’re found, but not found good enough to be cited.

Which metric is most important to start with?

Start with selection rate and citation share. Selection rate tells you whether your content is strong enough to be chosen. Citation share puts that into competitive context, the AI equivalent of market share.

Then add velocity to keep an eye on the direction.

Do I need expensive tooling to measure this?

Not to get started. Many of these metrics need discipline rather than infrastructure: a fixed prompt library of 25 to 50 representative queries, run periodically through the relevant AI platforms.

Count manually whether you’re mentioned, cited and how prominently. Specialised tooling helps later for scale, but the framework is independent of the tool. Our guide to measuring AI visibility without tools has the routine, and our AI visibility checker gives a quick first read.

How often should I measure these metrics?

AI answers are probabilistic and change with model updates, so a one-off measurement says little.

A SparkToro study of 2,961 prompts found ChatGPT and Google’s AI features returned the same brand list less than 1% of the time. How often you appear across many runs is far more stable than your position in one answer.

Measure monthly if you can, with rolling averages and velocity. Set alerts for sharp deviations, so you don’t notice a collapse 3 months later. Why answers vary is explained in the probabilistic nature of AI answers.

Need help?

Want to translate this into execution? See how we approach AI visibility, or contact us for a first baseline of your GEO metrics.