How to Manually Audit Your AI Visibility in Under an Hour

Most teams that haven’t started a GEO program yet assume the first step is buying a monitoring tool. It isn’t. The first step is spending 45 minutes finding out what AI tools currently say about your brand, using nothing but the free versions of ChatGPT, Perplexity, and a couple of other tools you probably already have access to.

This is worth doing before any tool purchase or agency conversation. It tells you whether you have a visibility problem at all, and if you do, roughly where it sits.

What you need before you start

• A list of 10-15 prompts your actual buyers would type when researching your category. Not your brand name, the problem they’re solving.
• A spreadsheet or even a simple doc to log responses.
• Access to ChatGPT, Perplexity, Google (for AI Overviews), and ideally Claude or Copilot.

That’s the entire toolkit. No account setup, no API keys, no paid plan required for a first pass.

Step 1: Build your prompt list (10 minutes)

Write down the questions someone would ask before they know your brand exists. If you sell project management software for agencies, that’s prompts like “best project management tool for creative agencies” or “how do agencies track client work without spreadsheets,” not “tell me about [your brand name].”

A good starting mix:

1. Category prompts. “Best [category] for [audience]”
2. Problem prompts. “How do I [solve the problem your product solves]”
3. Comparison prompts. “[Your product] vs [competitor]”
4. Direct brand prompts. “What is [your brand]” and “Is [your brand] good for [use case]”

Keep the list to prompts you’d genuinely expect a real buyer to use. A prompt list built around vanity keywords wastes the audit.

Step 2: Run each prompt across tools (25 minutes)

Run every prompt in the list through each tool and log three things for each response: whether your brand appears, whether you’re cited as a source (linked or named) versus just mentioned, and whether anything stated about you is inaccurate or outdated.

Tool	What to check	Where the gaps usually show up
ChatGPT	Does it name your brand unprompted in category/problem prompts	Category prompts, especially without web browsing enabled
Perplexity	Are you cited as a linked source, not just named	Comparison and “best of” prompts
Google AI Overviews	Does the overview pull from a page you control	Problem and category prompts with commercial intent
Claude / Copilot	Consistency with what the other tools say	Direct brand prompts, accuracy checks

The distinction between being mentioned and being cited matters here. A tool naming your brand in passing is a weaker signal than a tool linking to a specific page as its source. Log both differently.

Step 3: Look for patterns, not just individual answers

One bad answer to one prompt isn’t a crisis. A pattern across several related prompts is the actual signal. After logging your results, look for:

• Total absence. Your brand doesn’t come up at all on category or problem prompts where it reasonably should.
• Mentioned but never cited. You’re named, but the tool never links to or sources anything from your site, which usually means competitors’ content is doing the explaining work that should be yours.
• Outdated or wrong information. Old pricing, a discontinued feature, a former product name. This is often the easiest issue to fix and the most damaging to leave alone.
• Inconsistency across tools. ChatGPT says one thing, Perplexity says another. This usually points to conflicting information across the sources these models draw from, not a single bad page.

Step 4: Turn the findings into a short list

An hour-long audit should end with a short, specific list, not a long report. Aim for three to five findings you can act on:

• Which prompts you’re completely absent from
• Which pages, if any, are already getting cited (worth reinforcing, not rebuilding)
• The most glaring factual error to correct first
• Whether the gap looks like a content problem, a third-party presence problem, or both

This maps closely to what a full GEO content audit does at greater depth. The manual version here is meant to tell you whether that deeper work is worth prioritizing, not to replace it.

What this audit won’t tell you

A 45-minute manual check has real limits. It’s a single snapshot, not a trend line, and answers to the same prompt can shift between sessions. It also won’t tell you volume (how often people are actually asking these prompts) or give you historical comparison. For ongoing tracking, you eventually need something more systematic than a spreadsheet you update by hand.

None of that makes the manual version pointless. It’s the fastest way to find out if there’s a problem worth solving, and it costs nothing but time. Run it again in three months using the same prompt list, and you have a real before-and-after instead of a guess.