How to Run Your First AI Visibility Audit (Without Any Fancy Tools)

The obvious next question is:

“OK – how do we actually audit our AI visibility right now?”

You don’t need a platform to get started. You can run a simple, manual AI visibility audit in a couple of afternoons that will tell you more about your position in AI answers than any SEO report you’ve ever seen.

Here’s a step-by-step guide.

Step 1 – Choose one ICP and one proposition

The fastest way to get stuck is to be too broad. Don’t try to audit your entire business at once. Start narrow:

• Pick one ICP (e.g. “UK mid-market B2B SaaS companies”)
• Pick one proposition (e.g. “Customer success platform”)

You’re going to answer one question:

“When this specific ICP asks AI assistants about this specific problem, how do we show up?”

That’s your scope.

Step 2 – Map the research journey in plain English

Next, sketch out how that ICP actually researches a solution. Keep it simple:

1. Early exploration – “What is this and why should I care?”
2. Problem framing – “What does this mean for us?”
3. Solution discovery – “What options exist?”
4. Selection – “Which vendor can I trust?”

For your chosen ICP and proposition, jot down what they’re worried about at each stage and the internal questions they ask. You don’t need perfection. You just need something that sounds like a real human in your ICP’s shoes.

Step 3 – Turn that journey into real AI questions

Now we create what is essentially your first prompt library. For each journey stage, write 3–5 natural questions your ICP might ask an AI assistant.

Aim for 15–20 questions in total. Don’t sanitise them into neat SEO phrases. Keep them messy and human.

Step 4 – Ask ChatGPT, Perplexity and Gemini

Now take your question list and run it through at least two (ideally three) AI assistants:

• ChatGPT
• Perplexity
• Gemini / Copilot

For each question in each assistant:

1. Paste the question.
2. Capture the full answer (copy/paste into a document or spreadsheet).
3. Note down: Do we appear at all? Who else is mentioned? Does it link or cite any sources? What’s the tone?

Step 5 – Score your visibility with a simple model

This is where we turn anecdote into something you can actually talk about in a meeting.

For each answer where your brand could plausibly appear, score it along a few basic dimensions:

• Brand presence (0–2)
  0 = not mentioned | 1 = mentioned | 2 = clearly recommended or highlighted
• Position in the answer (0–2)
  0 = not mentioned | 1 = buried in a list | 2 = named first or framed as a primary option
• Coverage depth (0–2)
  0 = name only | 1 = one sentence of description | 2 = multiple sentences, use cases, pros/cons
• Evidence & citation (0–2)
  0 = no visible evidence linked to you | 1 = third-party mentions | 2 = your own site or assets cited/linked

You now have 0–8 points per answer. Add a simple column for competitor presence (yes/no) for each main competitor.

Step 6 – Roll it up by theme and journey stage

Once you’ve scored all the answers, step back and look for patterns:

• By journey stage: Are you more visible in early “what is this?” questions than in “who should we choose?” questions?
• By theme: How do you perform in your core topics vs adjacent topics?
• By engine: Do you do better in Perplexity than in ChatGPT?

“Overall, we appear in 35% of relevant prompts, and are recommended in 10%.”

You’ve just built your first AI Visibility Scorecard.

Step 7 – Diagnose what’s behind the numbers

The numbers tell you what is happening. Now you need to figure out why.

• When we’re visible and strong: What pages or assets are being cited?
• When we’re visible but weak: Are we just being name-checked? Does the description feel vague?
• When we’re invisible: Do we actually have content that would plausibly answer the question?

You’ll almost always find a mix of hard gaps (no content), soft gaps (weak content), and discoverability issues.

Step 8 – Turn the audit into a 90-day roadmap

A visibility audit is only useful if it leads to action. From your findings, choose:

• 2–3 priority themes where commercial value is high and visibility is weak
• 3–5 specific questions where you want to move from invisible → mentioned

For each, define 1–2 flagship assets to create or overhaul (deep guide, case study, comparison) and supporting improvements (clearer headings, explicit ICP language).

Step 9 – Decide how you’ll measure change

Before you execute the plan, agree on:

• When you’ll re-run the same questions (e.g. in 3 months)
• How you’ll compare scores
• Which commercial metrics you’ll keep an eye on (AI-tagged traffic, engagement)

On the next run, you want to be able to say: “We increased our appearance rate on priority prompts from 30% to 55%.”

Where a platform like Odyssiant fits

Everything above can be done with a spreadsheet, a few afternoons of disciplined work, and a willingness to look at how AI really describes your brand.

Where it starts to hurt is when you want to scale to hundreds of prompts per ICP, track multiple assistants at once, and re-run the audit quarterly without losing consistency.

That’s the gap Odyssiant is designed to fill: turning one-off AI visibility audits into a repeatable, measurable part of your marketing stack.

But you don’t need to wait for any platform to begin.