AI visibility methodology — how TopMention measures AEO
The fixed prompt panel, multi-model scoring, technical readiness, and what we deliberately do not count as AI visibility.
If you cannot explain your measurement, you cannot lead the category. This page is the owner for how TopMention defines and scores AI visibility for Shopify brands.
What we are measuring
AI visibility here means: on a fixed set of shopper questions, how often do ChatGPT, Claude, Perplexity, and Gemini name your brand — and whether your store is technically ready for those systems to fetch and understand your pages.
We split the problem into three layers:
- Mention share of voice (SoV) — do answers name you?
- Technical AI readiness — can crawlers fetch you, and is product data machine-readable?
- Citability posture — do you publish extractable facts and earn third-party proof?
Deep definitions: share of voice, prompt panel, technical AI readiness, citability.
The fixed prompt panel (non-negotiable)
We do not refresh prompts every week to chase vanity lifts. A panel is a sample. Change the sample and you invalidate the trend.
A healthy Shopify panel mixes:
- best-of discovery prompts
- comparison tradeoff prompts
- problem-solution pain prompts
- shopping where-to-buy prompts
Brand-check prompts (Is {brand} legit?) are useful in private audits but are not indexed as public ask pages — they collapse into near-duplicates.
Browse public long-tail examples in /ask. Category applications: /guides.
Multi-model scoring
We score per model, then blend. Winning ChatGPT while losing Perplexity often means a citation gap. Losing everywhere usually means access, entity, or category authority problems.
A mention counts when the answer names your brand or a configured alias. Position is noted for prioritization; SoV is primarily mention rate.
Technical checks that belong in the same scorecard
Mentions without fetchable HTML are fragile. Our audits cover crawler access, llms.txt, Product/Organization schema, and Shopping-oriented signals (offers, GTIN where relevant). See AI visibility audit for Shopify and ChatGPT Shopping feed requirements.
What we deliberately do not count
- A single anecdotal ChatGPT screenshot
- Ranking #1 on Google for a blog post about ChatGPT
- Traffic from AI referral logs alone (useful, not the same as SoV)
- “Brand sentiment” without a fixed question set
Operating cadence we recommend
- Baseline scan on your category panel
- Ship a focused technical + content sprint
- Re-measure weekly on the same prompts
- Layer citation/PR work for stubborn gaps
Done-for-you path: Fix Sprint. DIY start: free scan.