Before you spend anything on AI visibility, you should know what AI says about you today. Not a vague impression from one chat, but a baseline you can repeat and compare against in three months.
This is the method behind our paid audit, written out so you can run a simpler version yourself. You need an afternoon, a spreadsheet and access to a few AI tools. That's it.
Step 1: write twenty questions
The questions matter more than anything else. They should be the questions your buyers actually ask, in the words they'd use, not the words on your website.
Aim for twenty, spread across five types:
- Category questions (5): "What's the best [category] for [type of company]?" "Which [category] tools do people recommend?"
- Comparison questions (4): "[Your brand] vs [competitor]: which is better for [use case]?" "What are the alternatives to [competitor]?"
- Problem questions (4): questions that start from the pain, not the product. "How do I [solve the problem you solve]?"
- Direct questions (4): "What is [your brand]?" "Is [your brand] any good?" "Who founded [your brand]?"
- Reputation questions (3): "What do people complain about with [your brand]?" "Is [your brand] trustworthy?"
The best source for these is your sales and support teams. Ask them what prospects ask in the first call. Write those questions down, word for word where you can.
Step 2: pick five models
Use the tools your buyers are likely to use. A sensible set covers the major general assistants plus at least one search-focused tool: for example ChatGPT, Claude, Gemini, Perplexity and Microsoft Copilot. If your buyers use Google heavily, note what its AI answers say too.
For each tool, write down how you used it: which version, whether web search was on, whether you were signed in. Answers vary with these settings, and you'll want to repeat them exactly next time.
Step 3: set up the spreadsheet
One row per question per model, so 100 rows. Columns:
- Question and question type
- Model
- Named? Did the answer mention you at all? (yes / no)
- Position: first, second, third, or later
- Accurate? Is what it says about you correct?
- Tone: positive, neutral, hedging or negative
- Competitors named: who else appeared
- Sources: every source the model cited or linked
- Notes: anything odd, out of date or wrong
Step 4: ask, carefully
Use a fresh conversation for every question, so earlier answers don't influence later ones. Turn off memory or personalisation where you can, or use an account that hasn't discussed your company before.
Answers vary between runs, so for the questions that matter most, especially the category and comparison ones, ask two or three times and note whether the result changes. One answer is a snapshot; three is a pattern.
Paste the full answer somewhere you can find it again. You'll want to reread them.
Step 5: score it
Three numbers tell you most of what you need:
- Share of answers naming you: how many of the 100 answers mention you.
- Models naming you on category questions: out of five, how many name you unprompted when asked about your category. This is the number that matters most, because it's the one closest to a buyer's first question.
- Accuracy: how many answers that mention you get the facts right.
Do the same counts for your two closest competitors. The gap between you and them is more informative than any absolute number.
Step 6: read the sources
This is the step most people skip, and it's the most useful. Collect every source the models cited and group them: Reddit threads, Wikipedia and other reference pages, review sites, news articles, your own site, competitors' sites.
You'll usually find:
- a handful of threads or pages that shape most answers in your category
- that your own website is cited for basic facts but rarely for judgements
- that competitors appear in sources you're absent from
- at least one source that's out of date and still being quoted
That list is your to-do list. The sources models trust are the places where your visibility has to change.
Step 7: repeat it
Save the questions, the settings and the spreadsheet. Run exactly the same baseline again in three months, and then every quarter. Reviews can move in weeks; community presence takes a few months; answers usually shift over four to six months.
Without the baseline, you won't be able to tell whether anything worked. With it, you'll know exactly what changed and which source made it change.
Common mistakes
- Asking only about yourself. Direct questions flatter you. Category questions tell the truth.
- Using your marketing language. Buyers don't ask for "an AI-powered operations platform". Use their words.
- One run per question. Answers vary. Repeat the important ones.
- Ignoring competitors. Your number only means something next to theirs.
- Skipping the sources. The answer tells you where you stand. The sources tell you why.
That's the whole method. It takes an afternoon the first time and less after that. Run it before you decide whether to hire anyone, including us.