A measurement framework for ChatGPT, Perplexity, and AI Overview mentions that ties prompt baskets to pipeline instead of composite vendor dashboards.
By Rees Bayba ·
TL;DR
LLM visibility is not a replacement for SEO rankings, and it is not a leaderboard badge. For B2B GTM teams, it is a measurement layer on top of how buyers shortlist vendors: whether answer engines mention you, cite you, and point at URLs you control when someone asks a revenue-relevant question. This framework keeps the work honest: no fabricated benchmarks, no dependency on opaque composite scores.
| Metric | Definition | Why it matters for GTM |
|---|---|---|
| Mention rate | % of basket prompts where your brand is named in the answer | Are you in the shortlist at all? |
| Citation rate | % of prompts where a URL about you (yours or third-party) is cited | Can the buyer verify the claim? |
| Share of model | Your mentions ÷ (your mentions + named peer set) on category prompts | Relative presence vs the agencies you actually lose to |
| Trigger rate | % of branded or category prompts that produce any grounded answer with sources | Is the topic even answerable yet, or is the model hedging? |
| Source position | Rank of your URL among cited links when present | Are you the evidence or a footnote? |
Example basket themes for GTM engineering / outbound
Monthly is enough for most B2B motions; weekly only if you are in an active content ship cycle and need faster feedback. Run the same prompts logged-out or in a fresh context where possible, on ChatGPT (including search-grounded modes you care about), Perplexity, and Google AI Overviews or AI Mode. Record date, surface, model/settings note, mention (Y/N), cited URLs, position, and verbatim snippet. Store it in a sheet or CRM object your GTM engineer owns, the same person who owns outbound experiment logs.
Visibility without meetings is a vanity project. Meetings without a prompt map leave you guessing why AI helped. Run both. For the companion on why brands get omitted and how to earn citations, see the AI answer engine playbook; for category framing, see what GTM engineering is. Ready to baseline your basket and entity gaps with an operator? Book a strategy call.
It is whether answer engines mention and cite your brand on the buyer prompts that lead to pipeline, measured with a fixed prompt basket and metrics like mention rate, citation rate, and share of model versus named peers.
Composite vendor scores often hide the prompts, peers, and methods. If you cannot map a metric to strategy calls or assisted pipeline, it will not prioritize the right pages or off-site work.
Ten to twenty locked prompts is enough for a monthly cadence. Pull them from real sales language, cluster by intent, and version quarterly so trends stay comparable.
At minimum ChatGPT (including the search-grounded mode you care about), Perplexity, and Google AI Overviews or AI Mode. Add others only if your buyers name them in deals.
Treat mention and citation rate like reply rate: directional quality metrics on a defined ICP (here, a prompt ICP). Use them to decide what to ship next, then confirm with strategy-call attribution.
Enter your email and we'll send you this playbook, plus the setup templates as they're published.
We build and run these systems embedded with your team.