Tracking AI Citations Over Time

Most agency dashboards still run on metrics built for a search world that no longer exists. Organic position, click-through rate, time-on-page, bounce rate, page authority. Each one is real, each one still measures something, and none of them captures whether the brand is winning or losing in the surface where roughly half of all category buying decisions are now mediated. AI Overviews trigger on more than forty percent of local business queries. ChatGPT's daily user base passed 75 million in 2025. Perplexity, Gemini, and AI Mode each carry their own user populations. The classic dashboard cannot see any of this.

Vector 11 is the work that closes the measurement gap. The framework is three KPIs, the cadence is weekly to quarterly depending on the question being answered, and the deliverable is a dashboard that translates GEO work into the executive-level summary clients can actually use to evaluate the engagement. The framework is also, at the data-collection layer, the same thing as the Vector 1 diagnostic, repeated on a schedule. Diagnose is the first instance of measure; Measure is the recurring instance of diagnose.

The Three KPIs That Define GEO Progress

Three metrics, in coordination, produce a defensible measurement of AI search visibility. Each measures a different surface; each tells the brand a different story; each routes to a different remediation action when the metric moves the wrong direction.

The three together produce the executive-summary table. Brand Visibility climbing while Citation Rate stagnates means the engines know about the brand but are citing competitor content; the remediation is Vector 4 (Embed) and Vector 5 (Cite). Citation Rate climbing while AI Share of Voice plateaus means the brand is gaining ground in absolute terms but the category is also growing; the remediation is Vector 7 (Distribute) and Vector 9 (Cluster). Both metrics climbing with Share of Voice rising means the GEO program is winning; the remediation is to keep doing what is working and resist the temptation to redirect.

Brand Visibility: How Often the Brand Surfaces

Brand Visibility is the simplest of the three KPIs to compute and the most useful for spotting acute problems. The standing prompt set, established during Vector 1 diagnostic and refined through Vector 3 (Resonate), runs against ChatGPT, Perplexity, Gemini, and AI Overviews on the measurement cadence. For each prompt, the answer is logged for whether it mentions the brand. The aggregate (mentions divided by total prompts) is the Brand Visibility score, expressed as a percentage.

Two refinements matter for credibility. First, every prompt runs a minimum of three times across separate sessions because of the non-determinism reality from Vector 1; a prompt that mentions the brand on one of three runs has a 33% mention rate, not a binary yes. Second, the test set has to be stable across measurement periods; substituting prompts midstream invalidates trend comparison. The right discipline is to lock the test set quarterly, refresh it deliberately at quarter boundaries, and document the changes when prompt-set composition shifts.

Citation Rate: How Often AI Links Back to Your Content

Citation Rate is the higher-value metric of the three because citations produce the referral traffic. A mention is brand-awareness; a citation is a click candidate. The mechanic of the click conversion gap matters here. Industry tracking shows ChatGPT referral converting at roughly 15.9 percent versus Google organic's 1.76 percent. The order-of-magnitude difference is the single most important fact in the GEO measurement framework: an AI citation is worth approximately ten organic clicks in conversion terms.

Computing Citation Rate is operationally similar to Brand Visibility. For each prompt in the standing test set, log whether any URL on the brand's domain appears in the engine's source citations. The aggregate is the Citation Rate. Tools like Otterly, Profound, and HubSpot AEO Grader automate this collection across ChatGPT, Perplexity, AI Mode, and AI Overviews simultaneously; manual collection works for smaller test sets but consumes labour at scale.

AI Share of Voice: The Competitive Lens

AI Share of Voice is the metric that turns the first two KPIs into a competitive narrative. The computation is straightforward: total brand citations across the standing test set divided by total citations across the brand plus the named competitor set, expressed as a percentage. A 35% AI Share of Voice in a five-competitor category means the brand is roughly the leading source in the engines' citation network; a 10% Share of Voice in the same category means four competitors are being cited more often.

Matt Griffin notes in client briefings that "the conversation with the client always shifts when AI Share of Voice enters the dashboard. Brand Visibility and Citation Rate measure progress in absolute terms, which is useful but easy to dismiss as 'we are doing better but maybe everyone is.' Share of Voice forces the comparison: are you winning or losing relative to the competitors who are sitting on the same buying-intent prompts? The answer is auditable, and the remediation is targeted." The competitive lens also produces the most useful executive summary slide, because it answers the strategic question rather than the tactical one.

The Mattress Miracle Measurement Story

Mattress Miracle is the case where the Vector 11 framework has been fully exercised over the longest measurement window on record for a Formative Digital engagement. The story is worth telling in detail because the numbers are auditable and the methodology is the same one we apply to every client.

The methodology that produced these numbers is the entire 12 Vectors program. Vector 1 diagnosed the baseline (negligible AI surface presence at engagement start, organic rank concentrated in the long tail of low-intent queries). Vector 2 anchored the entity at the Wikidata and schema layer. Vector 3 produced the prompt inventory the content programme runs against. Vector 4 wrote the answer blocks. Vector 5 layered the citations. Vector 6 wrapped the schema graphs. Vector 7 earned the distribution. Vector 8 maintained the freshness. Vector 9 built the topical clusters around mattress retail, sleep health, mattress materials, and the local Brantford-Hamilton-Cambridge geographic surface. Vector 10 localized the work for the regional buying intent. And Vector 11 measured the entire stack across 18+ months of continuous tracking.

The Brad Wickens quote captures the executive-summary version: "In 40 years of advertising I've never seen anything like this. It's a completely new business." Forty years of industry experience is the relevant context; the testimonial does not say "this is good marketing," it says the methodology produced a structurally different business outcome. Results depend on industry, competition, and existing digital presence; the methodology is what is repeatable, and Mattress Miracle is the case where the full instrumentation has run long enough for the longitudinal data to be defensible.

The value of telling the Vector 11 story with these specific numbers is that the measurement framework itself is what made the gain legible. Without the Brand Visibility, Citation Rate, and AI Share of Voice tracking running continuously, the engagement would still have produced the traffic gain, but the story would not have been auditable, the optimizations would not have been targeted, and the velocity event of 25,000 keywords in one month would have looked like luck rather than methodology. Vector 11 is what turns the work into evidence.

Tools: Manual, DIY, and Enterprise

The tooling layer for Vector 11 has matured fast. Three options matter for most service businesses:

The breakeven for moving from Manual to Mid-tier is roughly one measurement cycle per month consuming more than four hours of skilled labour. Below that, manual is faster. The breakeven for moving from Mid-tier to Enterprise is multi-location complexity, compliance requirements, or prompt-set size beyond what mid-tier tools handle efficiently. Most brands run on Manual for the first six to twelve months, graduate to Mid-tier when the measurement cycle gets too long, and consider Enterprise only when the program scales beyond a single brand or geography.

If your existing dashboards still show only organic position and click-through rate, the dashboard rebuild is the highest-leverage measurement change available. A Vector 11 dashboard build typically takes about two weeks and produces the standing measurement framework all the other vectors are tracked against from that point forward.

From Measure to Iterate: The Vector 11 Handoff

Vector 11 is the measurement stage; Vector 12 is the iteration stage. The handoff is the recognition that measurement only matters if it changes the next quarter's work. A dashboard that shows visibility numbers but does not feed back into editorial planning, distribution priority, or schema attention is decorative rather than operational. Vector 12 closes the loop by reading the Vector 11 deltas and deciding which vector gets the next quarter's marginal effort.

The pattern that works across most engagements is a quarterly review cadence. The Vector 11 dashboard runs continuously; once a quarter, the FD team plus the client review the trends, identify the prompts and engines where the brand is gaining ground or losing ground, and decide which vectors get the next ninety days of priority work. The decision is data-driven (the dashboard provides the evidence) but not data-determined (the decision still requires judgment about strategic priority, competitive context, and resource constraints). Vector 12 is where that judgment gets installed as a repeatable operating discipline.

Frequently Asked Questions