Below is the full menu of approaches teams typically consider - each with their pros & cons

1) Any touch attribution in analytics tools

Track sessions where the referrer/source indicates ChatGPT, Perplexity, Copilot, etc. This is the cleanest click-based proof, but it’s incomplete because many AI experiences hide or drop referrer data.

Note: We prefer “any-touch” attribution, since LLMs are having an impact at all stages of the buying journey. First-touch or last-touch both don’t consider various other key stages of the funnel.

Pros

• Most realistic model for AI’s role across the buyer journey.

• Works well when referrers are passed (Perplexity is often more visible than some chat UIs).

• Lets you do normal analytics: landing pages, engagement rate, conversion rate.

Cons

• Some people don’t show up in analytics tools if they have ad blockers or might appear as direct traffic

• Doesn’t capture “zero-click” influence where users see your brand in an AI answer and later search you on Google.

2) Share of Voice in AI answers (prompt-based tracking)

Test a range of prompts and measure how often your brand appears versus competitors. It’s a strong competitive visibility metric, but prompt sets are never truly representative and coverage gaps are real.

Pros

• Competitive narrative is crystal clear: “We show up 18% of the time; Competitor A shows up 41%.”

• Works even in zero-click environments.

• Very useful for strategy: which topics you’re losing, which engines you’re winning.

Cons

• Requires very thoughtful prompt coverage

• Only really works for a certain type of query (eg “Best X for Y” but not branded searches)

3) Branded demand lift

Track whether branded queries rise after you increase AI presence (mentions/citations). This often captures the “LLM → Google → website” behavior pattern, but it’s correlational unless you control for other marketing.

Pros

• Executive-friendly: “AI visibility is increasing branded demand.”

• Captures invisible influence when referrers are missing.

• Pairs naturally with PR, partnerships, and category education.

Cons

• Impacted by other activity (paid spend, launches, seasonality, PR hits).

• Slow feedback loop; you need enough time window to see signal.

• You must segment branded terms carefully (brand vs. product-line vs. competitor comparisons).

4) Brand mention rate in AI answers

Measure how frequently your brand appears in AI responses for a tracked query set. It’s a strong “presence” indicator, but it doesn’t guarantee favorable context or clicks.

Pros

• Directly reflects “are we in the conversation?”

• Easier than citation analysis in some tools (you don’t need perfect URL extraction).

• Helpful early signal before traffic shifts.

Cons

• Mention ≠ recommendation (you might be listed as an alternative, or in a negative comparison).

• Sensitive to prompt sampling issues (same as SOV).

• Doesn’t prove business impact alone.

5) Citation rate (how often AI links or cites your content)

Track how often AI answers cite your domain/URLs. Citations are one of the best proxies for authority in AI outputs, but extraction and normalization can be messy.

Pros

• Strong trust signal: “The model didn’t just name us; it referenced our content.”

• Actionable for content teams (which pages earn citations, which don’t).

• Maps nicely to “topical authority” work (structure, clarity, sourceability).

Cons

• Not every AI surface provides explicit citations.

• URL variants, tracking params, and syndicated content can muddy counts.

• Citations can lag behind content updates.

6) Conversion rate and revenue from AI-referred sessions

For the AI traffic you can see, measure conversion rate and revenue just like any other channel. This is the most “CFO-proof” method but sample sizes are often small early on.

Pros

• Hard outcome metrics; leadership respects them.

• Helps you avoid vanity visibility reporting.

• Lets you judge traffic quality by engine (Perplexity vs ChatGPT vs Copilot, etc.).

Cons

• Small numbers cause noisy week-to-week swings.

• Missing referrer data can make AI look weaker than it is.

• Requires clean GA4 ecommerce/value configuration and consistent conversion events.

7) Page-level deltas (traffic lift to “AI-optimized pages”)

Identify pages you intentionally optimized for AI retrieval and measure lift (sessions, engagement, conversions) over time. This is pragmatic, but attribution is indirect unless you pair it with visibility signals.

Pros

• Very actionable: “We updated these 10 pages; these moved.”

• Lets you build internal case studies (“this format earns citations”).

• Useful when AI referrers are hidden (you’ll still see downstream effects).

Cons

• Impacted by normal SEO, seasonality, link wins, UX changes.

• Doesn’t tell you which AI engine drove the lift without extra signals.

• Can lead to wrong conclusions if you didn’t set a baseline.

8) In-house “AI visibility score” (a blended index)

Combine mentions, citations, SOV, and perhaps traffic into a single index. This can help internal tracking, but it’s the hardest thing to defend because the weighting is subjective.

Pros

• One number for trendlines (execs love trendlines).

• Lets you roll up lots of fragmented signals.

• Useful for internal prioritization if your team trusts it.

Cons

• Least defensible unless you publish the formula and keep it stable.

• Easy to “optimize for the score” instead of real outcomes.

• Management will ask, “So… what does a +12 score mean in dollars?”