Before you pay anyone to measure your AI visibility — including us — you should know how to do it yourself. The manual version is tedious enough that you'll eventually want a tool. But you'll also know what the tool is measuring, why it matters, and whether the number you get is plausible.

Here is the 10-minute version. Three steps. No accounts, no signups, no scripts. Just you, a notepad, and three AI tools open in browser tabs.

Step 1: Pick your 5 reference queries (2 minutes)

The queries you pick are 80% of what determines whether your audit is useful.

The mistake everyone makes the first time is picking queries that contain their own brand name. "Tell me about [my company]" will of course return information about your company. That tells you what AI tools already know. It does not tell you whether you appear when buyers actually search.

The right queries are the ones a buyer would ask BEFORE they've heard of you. Three categories:

Category queries. "What's the best [product type] for [customer type]?" — "What's the best CRM for a small SaaS company?", "Recommend an accounting tool for freelance designers."

Comparison queries. "[Competitor A] vs [Competitor B]", "Alternatives to [largest competitor]". If buyers compare you against a known leader, that's where you want to appear.

Use-case queries. "How do I [problem your product solves]?" — "How do I track which AI tools mention my brand?". If your product is the answer to the problem, the question should surface it.

Write down five queries. Mix the three categories. Keep them in the language and style your actual buyers use — not the keywords your marketing team writes for Google.

Step 2: Run them on 3 platforms (5 minutes)

Three is the right number for a first audit. More platforms gives diminishing returns at this stage. Pick:

  • ChatGPT — largest user base, most diverse query types
  • Perplexity — strongest at citing sources, will show you exactly which pages it's pulling from
  • Claude — strong on reasoning queries, often gives the most thorough answer

Open each in a separate tab. For each of your 5 queries, ask it on each platform. That's 15 query-platform combinations. About 20 seconds per ask.

For each answer, note three things in your notepad:

  1. Does your brand appear? Yes or no.
  2. In what context? Mentioned positively, listed neutrally, mentioned negatively, mentioned with wrong information.
  3. Which competitors appear? Write them down. This is your competitive landscape from the AI's perspective, which may differ from yours.

Don't agonize over interpretation. This is data collection, not analysis. Just write what you see.

Step 3: Score yourself out of 50 (3 minutes)

Your score is the count of "yes, appeared positively" across all 15 combinations, times some weighting.

The simple version: out of 15 combinations, how many included a positive mention of your brand? Multiply by 3.33 to get a score out of 50.

  • 0 appearances: 0/50. You're invisible. This is the most common result for SMBs.
  • 3-6 appearances: 10-20/50. You're known in some surfaces, missing in others.
  • 7-10 appearances: 23-33/50. You have meaningful presence. Most agencies score here.
  • 11-15 appearances: 37-50/50. You're cited consistently. Rare territory.

Where most people land on a first audit: 0-5/50. Including most well-known brands.

That's not a doom signal. It's a starting point. AI visibility is a recently-formed surface that almost nobody has optimized for yet. The brands that get to 30/50 in 2026 will be category leaders by 2028.

What this audit tells you (and what it doesn't)

The manual audit gives you four things real quickly:

A starting score. A real number you can compare against next month, next quarter, next year.

Your competitive set from the AI's perspective. Often different from your perceived competitors. Companies you've never heard of might be eating your share of AI mentions in your category.

The queries you're weakest on. Knowing which 3 of your 5 queries returned zero mentions tells you exactly where to focus content.

A sanity check on any tool you later evaluate. When a paid tool tells you your score, you can spot-check it against your manual numbers. If they're wildly different, ask why.

What the manual audit does NOT give you:

  • A repeatable measurement that you can run weekly without burning out
  • Coverage across more than 3 platforms
  • Tracking of changes over time (you'd have to do the whole thing again)
  • Specific fix-list recommendations (you see the symptoms, not the causes)
  • Multi-language or multi-region testing
  • Sentiment scoring beyond your own subjective read

This is the trade-off honest about it: manual audits are educational. Tools are scalable. If you only ever want to know once where you stand, do the manual version and stop. If you want to track week-over-week, you'll want a tool.

What to do with your starting score

If you scored 0-5: you need the foundations. Start with The 4-Week AI Visibility Sprint, which walks through the technical fixes that move a 0 toward something measurable.

If you scored 6-15: you have presence on some queries but not others. The work is content depth on the queries where you're weak. One cite-worthy article per month, targeted at a specific weak query, will compound over a quarter.

If you scored 16-30: you're doing better than 90% of your category. The work shifts from foundations to defense — making sure competitors don't eat your share — and offense in adjacent categories. This is where a dedicated tool starts to pay back, because tracking 15 manual combinations weekly is unsustainable.

If you scored 31+: rare, and either you're already optimizing systematically or you're the dominant brand in a small category. Either way, the question shifts from "are we visible?" to "are we cited correctly?". That's a sentiment-and-accuracy problem more than a visibility problem.

A note on what to ignore

If you read other AI visibility advice while doing this audit, you'll encounter a lot of noise. Two specific things to ignore on your first pass:

Ignore tools that promise to "rank you in ChatGPT." There is no ranking. ChatGPT generates an answer per query, not a ranked list. Anyone selling rank-tracking for AI is either confused or hoping you are.

Ignore the conflicting acronyms. AEO, GEO, LLMO, HEO — they all describe roughly the same job. We wrote the acronym guide if you want the full picture. For this audit, all you need is: did your brand appear, yes or no.

The noise gets louder every month. The fundamentals stay the same. Run the manual audit. Get the number. The rest follows.

If you've made it through this whole audit, you have something most agencies and even most product teams don't have: a real, measured starting point for your AI visibility, made with your own eyes. From here, every change you make is testable. That's the most underrated part of this whole field — it's not voodoo, it's just measurement done patiently.

The tool version is faster. The manual version is more educational. Do at least one of them.

If you'd like to compare your manual number against an automated one, run a free audit — same methodology at scale, 50+ queries across 6 platforms, sentiment-scored, with a per-platform breakdown. About 8 minutes, no credit card.