AI Search Visibility Audit Checklist

Most ecommerce brands have no visibility into whether AI systems cite them. This checklist provides the framework we use in every initial assessment.

Phase 1 — Discovery audit

Query each AI platform with the questions your customers actually ask. Not vanity queries about your brand — real purchase-decision queries.

For each query, record: does the AI mention your brand? Does it mention your competitors? Does it cite your website as a source? Does it recommend your product category? What sources does it cite instead?

Test across ChatGPT, Gemini, Perplexity, and Copilot. Each has different source selection behaviour. A brand can be well-cited in Perplexity and invisible in ChatGPT.

Use 20-30 queries minimum across informational, comparison, and transactional intent. Fewer than that produces unreliable patterns.

Phase 2 — Entity assessment

Evaluate how clearly each AI system understands your brand.

Ask directly: "What is [your brand]?" and "What does [your brand] do?" across each platform. The response tells you how the model represents your entity.

Common failure modes: the AI confuses your brand with a competitor. The AI describes your brand using outdated information. The AI can identify your brand but doesn't associate it with your core expertise. The AI doesn't know your brand at all.

Each failure mode has a different root cause. Confusion indicates inconsistent entity signals across the web. Outdated information indicates a static digital footprint without fresh content signals. Weak expertise association indicates shallow topical coverage. Zero recognition indicates insufficient entity presence in the model's training data and retrieval index.

Phase 3 — Competitive citation mapping

For the 20-30 queries from Phase 1, map which entities get cited. Build a citation frequency table: entity name, number of citations, which platforms, which query types.

This reveals the competitive landscape in AI search — which is often very different from the competitive landscape in organic search. Brands that rank well in Google may be invisible in AI systems, and vice versa.

Phase 4 — Schema and structured data review

Audit your site's structured data for machine-readable entity signals.

Check for: Organisation schema with complete fields (name, URL, logo, description, areaServed, knowsAbout). Service or Product schema on appropriate pages. FAQPage schema that addresses the queries from Phase 1. Consistent NAP (name, address, phone) across all directory listings and schema.

The gap between what your site tells humans and what it tells machines is typically significant. AI systems read structured data first because it's unambiguous.

Phase 5 — Topical depth assessment

Map your existing content against the topic clusters that drive citations in your vertical.

Count the number of pages that address each topic cluster with specific, original content. Compare against the entities that are currently being cited. Identify gaps where competitors have deep coverage and you have none.

The goal isn't to cover everything. It's to identify the 3-5 topic clusters where you can build genuine authority and focus exclusively there.

Phase 6 — Gap prioritisation

Rank every gap by two criteria: commercial value (does this topic influence purchase decisions?) and citation difficulty (how many authoritative competitors already own this space?).

High commercial value, low citation difficulty is your starting point. These are the topics where original, specific content can shift AI citation patterns within 2-3 months.

Low commercial value topics are not worth pursuing regardless of difficulty. This isn't a traffic strategy. It's an authority strategy. Every investment must connect to revenue.

How to use this

Run the full checklist quarterly. Track changes in citation frequency over time. Correlate content publishing with citation shifts. The lag between publication and citation is typically 4-8 weeks — faster than traditional SEO but still requiring patience.

The goal is a clear, data-informed picture of where you stand in AI-mediated discovery. Without this baseline, any optimisation work is guesswork.