A do-it-yourself audit: which questions to ask, how to ask them cleanly, what counts as "visible", and the mistakes that make self-testing misleading.
Muhammet İLBAŞ 7 min readBefore you buy any tool - ours included - you should run the free version of this audit yourself. It takes about ten minutes, costs nothing, and the result usually settles the "do we need to care about AI visibility?" debate one way or the other.
The biggest mistake in self-testing is asking about yourself: "What is [your brand]?" Almost every model will answer that politely - it proves nothing. Your buyers don't ask about you; they ask about their problem. Pull five to ten questions from the places where buyers say what they actually want:
Good prompts look like "best [category] for [audience]", "[competitor] alternatives", "is [category tool] worth it for a small team?" When you're ready to build a longer list, our guide to prompt research for AI visibility covers more sources and how to sort them by intent.
Two rules, both about contamination:
For each question, record three things: are you mentioned at all, at what position (first recommendation vs. an "also consider"), and is the description accurate. A mention that misstates your pricing or pitches you for the wrong use case is its own problem - we wrote a separate playbook for correcting AI errors.
ChatGPT alone is not the test. The major engines disagree more than you'd expect - same question, different shortlists. At minimum, run your set through ChatGPT, Claude, Gemini and Perplexity and score each separately. Perplexity earns special attention because it shows its sources - those citations are a map of which pages you need to win.
This audit gives you a snapshot. Its weaknesses: one run is one sample (models vary their answers), ten questions is a narrow window, and a snapshot can't show drift - the competitor who overtakes you in next month's answers. That's the part you can't keep up by hand: rerunning the same set across every engine and noticing what shifted since last week.
Type your brand name or domain - we'll line up the questions buyers ask about it.
Run the manual version today - it costs nothing and usually lights a fire. When the spreadsheet gets old, Zene's audit reruns your whole question set on a schedule and turns the gaps into an action list.
Ask it five to ten real buyer questions in a fresh, signed-out session and record whether you're named, where you land, and whether the description is accurate. The whole audit takes about ten minutes and costs nothing. Repeat the same set across Claude, Gemini, Perplexity and Grok, because the engines disagree more than you'd expect.
Because almost every model will answer that politely, which proves nothing. Your buyers don't ask about you - they ask about their problem, so test the questions they'd actually type, like "best [category] for [audience]" or "[competitor] alternatives". Asking about yourself only flatters you.
Yes - use a fresh session per question and turn memory off or stay signed out. Models carry conversation context, so once your brand has been mentioned every later answer is biased toward it, and ChatGPT's memory personalises results to you. You want the answer a stranger gets, not the answer you get.
GA4's new AI Assistant channel misses some assistants and app clicks land in Direct. The custom channel group, the regex, and what GA4 still can't see.
Muhammet İLBAŞ 9 minKeyword tools can't show what buyers ask ChatGPT. Mine sales calls, tickets, Search Console and Reddit, sort by intent, and pick the prompts worth tracking.
Muhammet İLBAŞ 10 minHow SaaS products get named when buyers ask ChatGPT, Claude or Perplexity for the best software: category clarity, comparison pages, G2, crawlable pricing.
Muhammet İLBAŞ 10 minPut this guide into practice - get your free visibility score in minutes.
