Why is searching for yourself not enough?
The most common mistake when auditing presence in answer engines is asking “what is my brand?” That question almost always returns something, because the assistant looks up the name and finds your site. What it measures is not your visibility: it measures that you exist.
The question that matters is the one asked by someone who does not know you yet. Nobody asks ChatGPT about a company whose name they already have; they ask about a problem. That is where it gets decided whether you are the recommendation or someone else is.
Which questions should you ask?
Write between ten and fifteen questions that meet three conditions: they use business language, not product language; they describe a problem, not a solution; and they do not contain your brand. Cover four categories:
- Vendor: “Which companies implement HubSpot in Colombia?”
- Problem: “My sales team does not use the CRM. What do I do?”
- Comparison: “For integrating systems, am I better off with an agency or a technical partner?”
- Decision: “What should I check before hiring a digital audit?”
The vendor questions measure whether you get recommended. The other three measure whether you get cited as a source, which is how the first one gets built.
How do you run the test so the result is worth something?
Three conditions, and all three matter:
- A fresh session, with no history. If the assistant remembers your previous conversations, it will show you a result contaminated by your own context.
- Several engines. ChatGPT, Perplexity, and Gemini build their answers differently: Perplexity cites sources explicitly and is the quickest to reveal where an answer came from.
- A written record. A sheet with the question, the engine, the date, the brands mentioned in order, and the sources cited. Without a record there is no comparison three months from now, and the comparison is the only data worth having.
How do you read your brand not showing up?
An absence is not one result. It has three distinct causes, and each one is fixed differently:
It cannot find you. Your site is not crawlable by AI crawlers, or the relevant content sits behind JavaScript those crawlers do not run. Quick to check: fetch your page with an AI user agent and see whether the text is there.
It finds you but does not understand you. Your site exists but does not state clearly what you are, where you operate, and who you serve. That is what structured data and a consistent entity architecture fix.
It understands you but does not pick you. The engine knows who you are but prefers sources with stronger trust signals: content with verifiable data, presence in third-party directories, consistency held over time. It is the slowest to fix and the only one that technical changes will not solve.
What does the second test, the one that names your brand, reveal?
After the anonymous block, ask directly: “What is your brand?”, “What does it do?”, “Where does it operate?” Here you are not measuring visibility, you are measuring comprehension.
The typical failures are revealing: the assistant confuses you with another company with a similar name, describes services you no longer offer, puts you in the wrong country, or treats your legal name and your trading name as two separate organizations. That last one is surprisingly common among companies trading under a name other than their legal one, and it is fixed by declaring both as the same entity in the site's structured data.
How often is it worth repeating?
Quarterly. Less often and you cannot tell an improvement from a fluctuation in the model; more often and you will drive yourself mad chasing noise, because the same engine's answers vary between sessions without anything changing on your side.
What you measure is not “did I appear?” but the trend: in what share of your questions you show up, where inside the answer, and whether the engine cites your site or merely mentions you. Being cited as a source is worth far more than being named in passing.
And if the result is that you appear nowhere?
It is the most common result and it is not a catastrophe: it is the normal starting point for almost any company that has not deliberately worked on its presence in answer engines. We treat that diagnosis as the first deliverable of an AEO strategy, because knowing which of the three causes applies to you completely determines what comes next.
What does not work is trying to solve it by producing more content without first knowing whether the problem was crawling, comprehension, or trust. Those are three different jobs.
Telling which of the three applies to you is exactly what we do in an AI Discovery, because that answer decides whether the work ahead is technical, structured data, or editorial.