Auditing your brand's presence in ChatGPT, Perplexity, and Gemini

You can't log into a dashboard and see your ChatGPT rankings - no such console exists. What you can do is what researchers do when there's no instrument: sample systematically. A structured audit, repeated quarterly, gives you a real (if approximate) read on whether AI assistants know you, recommend you, and describe you accurately.

Building the query set

Start with 20–30 queries in three tiers. Category queries - the money tier: "best linen shirt brands," "where to buy quality knitwear in Scandinavia," "sustainable fashion brands like [bigger competitor]." These test whether you get mentioned unprompted. Brand queries: "what is [yourbrand]," "is [yourbrand] good quality," "[yourbrand] sizing." These test whether you're described accurately. Purchase-intent queries with constraints - fanout-shaped: "linen shirt under €150 that ships to Denmark." These test whether your facts are extractable. Write them the way customers talk to assistants - conversational, constrained - not like keyword-tool entries.

Running it honestly

Run each query across the assistants that matter to your market (ChatGPT, Perplexity, Gemini, Claude cover most ground; Perplexity's visible citations make it especially instructive). Log per response: mentioned or not, position among mentions, sentiment/accuracy of the description, which competitors appeared, and - where citations show - which pages were sourced. Methodological honesty required: responses vary between runs and accounts, so treat single answers as anecdotes and patterns across the set as signal. Fresh sessions, no personalization, same protocol each quarter. Tools are emerging that automate this sampling at scale; they're useful once you've done it manually enough to know what the numbers mean.

Reading the results

Three findings, three responses. Absent from category queries where competitors appear: a retrieval-and-authority gap - the assistants' sources (the review roundups, comparison pieces, and publications being cited) don't include you. The fix is earning presence in cited sources, which is the next lesson entirely. Described inaccurately on brand queries - wrong positioning, outdated facts, hallucinated details: an information-supply gap. The correction is publishing clear, current, consistent facts where systems read them - your own pages first (an accurate, factual About/brand page suddenly has strategic weight), plus the third-party sources being cited. Mentioned but never cited on constraint queries: your pages aren't the extractable source - back to this phase's on-page principles.

The audit's quiet second payoff: the citation log is a map of which sources AI systems trust in your category. That map is the targeting list for everything in the next lesson - and knowing it puts you ahead of competitors still treating AI visibility as unknowable.