Tracking Awareness

Every AI visibility tool — including the first version of Spyglasses — was built to answer one question: "Does my brand show up when AI is asked about it?" That measures the decision stage of the buyer journey, where brands are already in play. It misses the awareness stage: the moments when a buyer has identified a problem but hasn't formed brand preferences yet, when they ask AI to help them understand the category, the trade-offs, or the path forward.

Awareness-stage visibility is upstream of share of voice. If your framing of the problem doesn't show up in those early answers, your brand has a much harder time entering the conversation later — even if your downstream optimization is excellent. Tracking awareness is how Spyglasses measures whether your narrative is shaping the buyer's mental model before they go looking for vendors.

What You'll Learn

• Why awareness-stage AI visibility is a different question from share of voice
• The difference between Share of Influence (SoI) and Share of Voice (SoV)
• How Spyglasses computes SoI from your key messages and awareness-stage prompts
• The three steps to set up awareness tracking for a property
• How to read the AI Visibility Funnel on the Historical Metrics dashboard
• Common pitfalls and how to tune your coverage

Why Awareness Matters

Picture a buyer who has just realized they have a problem. They don't know which brands solve it. They don't know what the category is called. They open ChatGPT and ask:

"What should I consider when choosing a data loss prevention tool?" "How do companies handle workflow management at scale?" "What are the main approaches to decarbonizing industrial heat?"

These are Category Entry Point (CEP) questions. The AI's answer to them rarely names specific brands by default — instead, it defines the category, lists criteria buyers should care about, and shapes the lens the buyer will use later when they do start asking about specific vendors.

If your differentiators show up in those framing answers — "low-cost storage", "24/7 zero-carbon heat", "the only platform with SOC 2 and HIPAA out of the box" — buyers come to the vendor evaluation stage already primed to look for what you do. If your differentiators don't show up, you're spending decision-stage effort fighting an uphill battle.

Traditional AI visibility tools score these awareness moments as failures because no brand is mentioned. Spyglasses scores them as opportunities — and gives you a metric (Share of Influence) that measures how often your narrative is winning them.

Two Different Metrics: SoI vs. SoV

Spyglasses tracks two visibility metrics side by side, each answering a different question at a different point in the buyer journey:

They are deliberately separate. Combining them would hide the gap between "AI is using my framing of the problem" and "AI is naming my brand as a solution." You can win one and lose the other — and the diagnostic actions for each are completely different.

How Share of Influence Is Computed

The math has two layers — the matcher, and the rollup.

The matcher. Every awareness-stage AI response runs through a semantic matcher. For each positive key message you've defined, the matcher embeds the message and every sentence in the response, then takes the cosine similarity. A sentence is a "match" when its similarity to a key message is at or above 0.50 — a threshold calibrated so the sentence has to express the same idea, not just share the topic. Sentences talking about cost in general don't match a key message about "lowest-cost energy storage"; sentences that actually claim something is the lowest-cost storage option do.

The rollup. For each positive key message, Spyglasses computes the percentage of awareness-stage AI responses (across every platform and every awareness-stage prompt) that contained at least one matching sentence. Share of Influence is the average of those per-message rates. Three positive key messages matching in 30%, 50%, and 70% of awareness answers gives an SoI of 50%.

Two things to internalize:

• Average across messages, not pooled — if you had three positive key messages, each is weighted equally. Adding a fourth positive key message that almost never matches will drag SoI down, even though it doesn't change the underlying signal for the other three. That's intentional; SoI rewards consistent differentiator coverage, not raw volume.
• Positive only — negative and neutral key messages aren't part of SoI. They're tracked separately (you can see their mention rates in the per-message detail), but they're noise relative to the narrative you're trying to build.

Setting Up Awareness Tracking

Three concrete steps for a new property:

1. Define your positive key messages

Open Key Messages in the property sidebar and add three to seven positive-sentiment messages — the differentiators you want AI to repeat when buyers ask about your category. Be specific and ownable; "the only X with Y and Z out of the box" matches better than "we care about customers". See Key Messages → What Goes in a Key Message for the guidance Spyglasses uses internally to evaluate message quality.

This is the most important step. The matcher is only as good as the messages it's matching against. If your positive key messages are vague or aspirational, SoI will look low even on good AI answers — because there's no clear semantic target for the matcher to find.

2. Author awareness-stage prompts

Open Prompts and make sure your library includes prompts at the awareness stage. Spyglasses recognizes three query types as awareness-stage:

• category_entry_point — "What should I consider when choosing a [category]?"
• jobs_to_be_done — "I need to [outcome] — what approaches can help?"
• buyers_journey_awareness — "What is [problem] and how do companies solve it?"

The fastest way to populate them is the Generate Prompts dialog with the Category Entry Points, Jobs to Be Done, or Buyer's Journey frameworks selected — each generates up to 25 prompts in the right shape, drawn from your brand snapshot. The Coverage Matrix tab on the Prompts dashboard surfaces awareness-stage gaps so you can see where you're thin.

A library that's heavy on category_best and solution_comparison but light on awareness-stage prompts will have an artificially low SoI denominator — there just aren't many awareness responses to match against. Fill the awareness column of the Coverage Matrix first; SoI is meaningless without coverage.

3. Run a report or wait for the nightly run

Once positive key messages and awareness-stage prompts are in place, the next AI Visibility Report (and the next nightly prompt run) computes per-(stage × platform) attribute matches and writes them to the AttributeMentionMetric table. The AI Visibility Funnel on the Historical Metrics dashboard reads from there.

Reading the AI Visibility Funnel

The Historical Metrics dashboard's headline section is the AI Visibility Funnel — a three-stage view of AWARENESS → CONSIDERATION → DECISION with a dashed divider between the awareness segment and the consideration/decision pair. The divider is doing real work: it visually flags that awareness uses a different measurement (SoI) from the others (SoV).

For each segment, the card shows:

• The metric (SoI for awareness, SoV for consideration / decision)
• The headline percentage for the selected platform and tag
• A "count" sub-line — for SoI, "avg across N positive key messages"; for SoV, "X of Y answers"

Click any of the three cards to expand a per-stage trend chart below. The chart honors the same platform and tag filters as the headline number.

Filtering by platform

The Platform select at the top of the dashboard re-scopes all three segments. Picking ChatGPT shows SoI and SoV restricted to ChatGPT executions only — useful for understanding where your framing is winning and losing, since the same key message can perform very differently across model families. "All Platforms" reads the rolled-up bucket.

Filtering by tag

When you've added custom tags to discovery prompts, a Tag select appears next to the Platform filter. Selecting a tag re-scopes the funnel to just the prompts carrying that tag. This is how you slice awareness performance by persona (enterprise, smb), geography (us, eu), campaign (q3-launch), or any other facet that matters to your buyer.

Common Pitfalls

• Vague positive key messages. "Industry-leading platform" is a key message every brand could claim and that matches nothing specific. The semantic matcher needs a concrete narrative claim — "the only X with Y and Z out of the box" — to score reliably. If SoI looks low and the per-message detail shows near-zero matches across every message, the messages probably need sharpening before anything else.
• No awareness-stage prompts. SoI's denominator is awareness-stage AI responses. If your prompt library is all consideration and decision queries, SoI is computed against a tiny denominator and the percentage swings wildly. Run the Coverage Matrix and fill the awareness column.
• Treating SoI like SoV. SoI is a narrative-coverage metric, not a competitive-share metric. It can be high while SoV is low (your framing is winning the category, but buyers aren't yet asking AI for your specific brand) — or the reverse (buyers know your brand, but their mental model of the category isn't shaped by your framing). The diagnostic for each is different; the dashboard separates them so you can act on the right one.
• Legacy reports show no awareness data. Reports written before stage-aware tracking was added have null stageBreakdown JSON. They'll show as flat in the funnel until the next AI Visibility Report or nightly run writes a fresh row. Editing aliases or competitors triggers a background recompute that backfills the breakdown for existing rows.

Practical Tips

• Capture positive key messages before the first report. They drive the SoI metric directly, and they also seed the field-level discovery prompts for PERSON reports (see Tracking Individuals). Defining them up front means the first report tests what you actually care about.
• Watch SoI and SoV together, not in isolation. A campaign that moves SoI without moving SoV is doing top-of-funnel work that hasn't yet converted to vendor recognition — track it for a few cycles before deciding whether it's working. A campaign that moves SoV without moving SoI is brand-name optimization layered on someone else's framing.
• Slice by platform to find the gaps. SoI can vary substantially across model families. If ChatGPT shows 60% SoI but Claude shows 15% on the same key messages, the issue isn't the messaging — it's where Claude is sourcing its category framing. The platform filter on the funnel makes the gap visible immediately.
• Revisit messages quarterly. Like all positioning artifacts, key messages drift relative to competitor moves and market evolution. A quarterly review (delete what's stale, sharpen what's vague, add what's newly relevant) keeps SoI a meaningful signal rather than a leaderboard for messages that don't matter anymore.

Related

• Historical Metrics — Where the AI Visibility Funnel lives
• Key Messages — Define the positive narratives that drive SoI
• Prompts — Author awareness-stage prompts and tag them for slicing
• Tracking Products — Apply awareness tracking to a specific product
• Tracking Individuals — Apply awareness tracking to a person