Answers / Do it yourself

How to check what AI says about your business (free, 15 minutes)

By Hoss · Published July 13, 2026 · Sources linked at the end

This is the complete method. There's no tool to sign up for, no email gate, and nothing to buy at the end of it — a version of this check is something anyone can run themselves in about fifteen minutes [1]. If you follow the steps below, you'll know more about your AI visibility than most businesses ever find out.

Step 1: write down the questions your customers actually ask

Start with 5–10 questions, phrased the way a real customer would type them — not the way you'd describe your own business. The useful questions are the ones asked right before someone chooses a provider. Examples by business type:

• Med spa: "best med spa for Botox in [your city]" · "is microneedling better than a chemical peel" · "what credentials should a Botox provider have"
• Law firm: "best personal injury lawyer in [your city]" · "DUI lawyer near me" · "how do I choose a divorce attorney"
• B2B software: "best [your category] for a small team" · "alternatives to [the category leader]" · "[you] vs [your main competitor]"
• Local services: "best plumber in [your city]" · "who should I call for [the job you want]"

Mix two kinds: selection questions ("best ___ near me") and research questions ("is X better than Y", "what should I look for in a ___"). Research questions matter because customers ask them earlier, and whoever gets cited there shapes the shortlist later.

Step 2: open a fresh chat in each engine

A fresh chat matters because prior conversation history skews what the engine says — if you've asked it about your own business before, it may remember, and your results won't reflect what a stranger sees. The manual audit process is the same one the tracking tools automate: new chat, run the query, record the result [2].

Test at least these four:

• ChatGPT — with search enabled, so it's answering from the live web rather than training data alone.
• Google Gemini
• Perplexity
• Claude

Don't assume one engine stands in for the rest. A study of 127,198 citations across five engines found only 2.7% of sources were cited by all five [3] — the engines barely read the same web, so being named in one tells you almost nothing about the others.

Step 3: run each question two or three times

AI answers are probabilistic. The same question, in the same engine, five minutes apart, can name different businesses. One run can make you look present when you actually appear one time in five — or absent when you appear four times in five. Manual audit guides converge on the same rule: repeat each query two or three times before you trust the pattern [2].

Step 4: record four things per run

For every run, write down:

1. Are you named? Yes, no, or only when asked about you directly.
2. How are you described? Verbatim if possible. Watch for wrong locations, wrong services, outdated facts.
3. Which competitors appear? Names and order.
4. Which sources are cited? Review sites, directories, Reddit threads, news, your own site.

A tracking sheet that works

One spreadsheet row per run. Seven columns:

The example rows above are illustrative — yours will look different. When you're done you'll have 40–120 rows, and the pattern will be obvious at a glance: how often you're named, who wins instead, and which sources the engines keep reaching for.

How to read your results

You're never named

This is usually one of two gaps. Either the engines can't retrieve you — crawler blocks, unindexed pages, content the engine can't read — or there isn't enough third-party corroboration for the engine to name you with confidence. Businesses that rank well on Google and still don't appear in AI answers almost always have the corroboration gap: not enough said about them anywhere but their own site [4]. The technical checks below sort out which gap is yours.

You're named, but described wrong

Wrong city, wrong services, a founder who left years ago — this is an entity and factual problem, not a visibility problem. It typically traces to fragmented or inconsistent information about you across the web: old directory listings, conflicting addresses, thin or contradictory profiles [5]. The fix is at the sources the engines cite, not in your ad budget.

Competitors are always named — study their footprint

If the same two or three competitors appear run after run, look at the sources column of your sheet: that's where their advantage lives. A competitor who consistently shows up typically has a heavier third-party footprint — verified directory profiles, review platforms, industry publication mentions, press coverage [5]. This matters more than most owners expect: a Semrush study of 150,000+ LLM citations (June 2025), reported by Custom Legal Marketing, found 40.1% pointed to Reddit and 26.3% to Wikipedia [6] — what others say about a business outweighs what the business says about itself.

Five-minute technical self-checks

Three checks anyone can do, no developer required:

1. Is your robots.txt blocking AI crawlers?

Visit yourdomain.com/robots.txt in a browser and look for lines that disallow GPTBot, OAI-SearchBot, or PerplexityBot. Many sites blocked all AI crawlers in 2023–2024 on principle and are now invisible in AI answers as a direct result [7]. Note that GPTBot and OAI-SearchBot are different: GPTBot gathers training data, OAI-SearchBot powers ChatGPT search — blocking one is not blocking the other [7]. Each AI surface has its own crawler: Googlebot feeds AI Overviews, OAI-SearchBot feeds ChatGPT search, PerplexityBot feeds Perplexity [8].

2. Do your pages have structured data?

Right-click a key page, choose "View page source", and search for application/ld+json. If nothing comes up, your pages carry no schema — no machine-readable statement of what your business is, where it is, and what it offers.

3. Do you have an llms.txt?

Visit yourdomain.com/llms.txt. It's an emerging convention — a plain-text map pointing AI systems at your authoritative pages. Absence isn't fatal, but it's a signal of how much machine-readable groundwork has been done. (You can see ours.)

When the free check is enough — and when it isn't

For many businesses, this check is genuinely all you need right now. If you run it and you're consistently named, described accurately, and cited from sources you recognize — you don't have an urgent problem, and nobody should be selling you one. Re-run it quarterly and get back to work.

What this method gives you is a snapshot. What it can't give you:

• Statistical confidence. Ten questions run twice is a signal; it isn't a measurement. Manual methods scale to roughly 50 queries across platforms before the spreadsheet becomes the job [9].
• Change over time. One snapshot can't tell you whether you're gaining or losing ground — that takes the same question set, re-run on a schedule, month after month.
• Cause. The sheet tells you that you're missing; it doesn't tell you which crawler rule, rendering failure, or missing citation is responsible.

If your snapshot looks bad and the revenue at stake justifies it, the next step is systematic measurement — whether you build that discipline in-house or hire it.