How to Test If You're Ranking With AI (Before You Spend a Cent Fixing It)
This article is part 1 of a 11 in a Series of article that will help your organisation rank with AI. Read the overview article “SEO Got You Found. AIO Gets You Recommended" for a broad understanding.
Open ChatGPT, Perplexity, Claude and Gemini. Ask each one the question your best customer would ask about your category and then read what comes back. That's the test, and it costs nothing but ten minutes. Most organisations skip this step entirely. They jump straight to schema markup or an llms.txt file without ever checking whether they show up at all, and importantly, not knowing who's showing up in their place.
Why do this first
You can't fix a visibility problem you haven't measured. Every technical or content change downstream of this - restructuring pages, adding schema, publishing on LinkedIn - is a guess unless you know your starting point. This is the baseline the rest of the series of articles builds on.
The three-question test
The category question. "What's the best [your category] for [your audience]?"
The problem question. Ask the question a prospect asks before they know your category is the solution.
The comparison question. "How does [your brand] compare to [a competitor]?" This reveals whether the AI has any independent read on you at all, or whether it's guessing.
Run all three across ChatGPT, Perplexity, Claude and Gemini separately. Don't average them, treat each platform as its own test because the answers will genuinely differ.
What to record
Keep it to a spreadsheet with five columns: platform, question asked, did you appear (yes/no), where you appeared in the answer (named directly / mentioned in passing / not at all), and what the AI said about you. That last column matters most; an AI that mentions you inaccurately is arguably a bigger problem than one that doesn't mention you at all.
Download a tracking template here
What a "bad" result actually tells you
If you don't appear it usually points to one of three causes:
your content isn't structured for extraction,
your off-site presence (reviews, directories, editorial mentions) is thin, or
you're simply too new to the space for the model to have picked up a confident signal yet.
The rest of this article series addresses each of these, one at a time.
Do this quarterly, not once
AI models update, retrain, and re-crawl on their own schedules. A result from today tells you almost nothing about six months from now. Set a recurring reminder, run the same three questions, and track the trend rather than any single snapshot. The goal isn't a perfect score on one run. It's a visibility trend that moves in the right direction over time.
