AI answer engines like ChatGPT, Google AI Mode, Perplexity, Copilot, etc. are a discovery channel now. If you write content and don't know whether it's showing up in these answers, you're flying blind where a growing share of your audience is starting their search.
There's a market of paid tools built to track this. Most of them are priced for teams with budget to spare, and most marketers testing the waters don't need a fancy platform. They really just need a habit. A spreadsheet, run consistently, can give you a strong signal for the decision that actually matters: keep publishing on a topic, or reconsider it.
Below is the method I use, plus a free template you can copy.
How to pick your 10–20 prompts
Start from your content pillars: one prompt per core topic you actually publish on. From there, build the list across four categories:
Brand and name recognition. "Who is [you/your client]?" or "What is [your unique value prop]?"
Framework or concept ownership. If you've coined a term or method, test whether the mechanism gets attributed to you/your client. This is the highest-signal category if you have one.
Category and topic authority. The broad questions in your space that have nothing to do with your brand like "how to measure SEO ROI beyond rankings," "how to structure a content hub." This is where most of your list should live.
Practitioner-problem framing. How someone searches before they know the terminology. Not "what is customer churn" but "why are subscribers canceling after month two." This category tests whether your content gets surfaced for the problem, which is usually how you get discovered in the first place.
Cap the list at 10–20 prompts. Since this is a manual exercise, you will want something manageable on a weekly basis.
The cadence
Test the prompts each week, worded exactly the same way each time. You're tracking change over time, not testing how phrasing affects the answer. Pick 2–3 tools to start (ChatGPT and Google AI Mode is an easy starting point) rather than trying to cover every platform at once.
To avoid contextual bias muddying results, make sure to run these prompts via incognito mode or without being signed in.
How to read a result
Mark each prompt Y, N, or Partial.
Y and N are self-explanatory. Partial is when your framing, your language, or your mechanism shows up in the answer, but you're not named or linked. This can be a leading indicator that the content is influencing the answer before the citation catches up. Treat a run of Partials as a reason to keep pushing on a topic, not a reason to drop it.
Taking it further
Y/N/Partial helps you build a trend line, but it doesn't get you the full response. Once the weekly habit is running, a few additions make the tracker more useful for long-term content decisions:
Save the actual response, not just the verdict. Paste the relevant paragraph or screenshot it. Answers reword and decay over time. The citation from six weeks ago won't look the same today, and without the original text you can't tell whether the mention got stronger, weaker, or just different.
Separate citations from mentions. Being described and being cited as a source are different signals. A tool can accurately summarize your framework without ever naming you or linking to your site. Tracking these as two categories can be insightful because they tell you different things about how the content is being used.
Note where you show up in the answer. Named directly in the first paragraph is a different result than buried in a list of five sources at the end. It's a rough signal, but position tends to correlate with how much the tool is leaning on that source.
Using it as a content compass
The real value of this tracker is what it tells you to do next. If a topic on your list sits at zero for months while everything else on the list is moving, that's a strong insight: either you have room to own the gap, or it's a crowded topic where you're unlikely to break through and your time is better spent elsewhere. Run this consistently and the tracker becomes a filter for your content calendar.
Get the template
Free Template: Make a copy, fill in your own 10–20 prompts, and run your first check this week. Building this cadence will help give you insight into your AI performance without paying hundreds each month for a visibility dashboard.