Why this is worth an afternoon
A growing share of buying decisions now starts with a question to a model rather than a search box. Somebody gets named in the answer, and the person asking usually stops there.
The uncomfortable property is that absence is silent. A lost search ranking shows up in your analytics as traffic that fell. An answer that recommended three competitors and not you produces nothing at all: no impression, no click, no line in any report you own. It is the only channel where being invisible looks exactly like being fine.
Step 1. Write the questions a buyer would actually ask
Not your product name, which proves nothing except that you exist. The questions somebody asks before they know who you are.
Ten is plenty, and they should span the three shapes that matter: the category question ("best tool for X"), the constrained one ("best X for a small team on a budget"), and the displacement one ("alternatives to [your competitor]"). That last shape is where deals are won and lost, and it is the one most companies never check.
Step 2. Ask several engines, in a clean session
Use at least four assistants. They read different sources and they disagree, and where they disagree is the useful part: a brand named by one and missed by three has a coverage problem in specific places rather than a general one.
Start a fresh conversation each time and stay logged out where you can, so your own history does not bias the answer into telling you what you want. Ask the question plainly and do not lead it.
Step 3. Record more than whether you appeared
Write down four things per answer. Were you named at all. In what position, since first is worth several times fourth. What claim was attached to you, because a wrong reason to pick you is its own problem. And which sources the answer leaned on, when the engine shows them.
That last column is the actionable one. Answers are assembled from pages: comparison articles, review directories, documentation, listicles somebody else wrote. If the same three sources keep feeding the answers in your category and you are absent from all three, you have found the work.
Step 4. Repeat, because one run is not a ranking
Ask the same question twice and you may get two different lists. Models are not deterministic, they update, and phrasing moves the answer more than most people expect.
So a single run is an anecdote. Ten questions across four engines on three separate days is a measurement, and the number that matters is the share: out of forty answers, how many named you. Do it once and you know where you stand today, which is genuinely more than most of your competitors know about themselves.
What to do with a bad result
The instinct is to rewrite the homepage. That is the least effective available move, because the model was not reading your homepage.
It was reading the pages that compare things: third-party listicles, review platforms, community threads, documentation, and comparison pages including the ones your competitors wrote. Being present and accurate there is what changes the answer. You cannot argue with an engine, but you can be present on the pages it reads, and that is an editorial and outreach job rather than a technical one.
Check the traffic half too. Being named and getting a visit are different things, and the share arriving from assistants is measurable separately from whether you were mentioned.
When the manual check stops being enough
Doing this by hand answers today's question. It does not tell you whether you are gaining or losing, which is the question that actually matters, and nobody keeps up a forty-answer manual routine every week for a year.
That is the part we automate: the same questions put to four engines on a schedule, with the answer text kept so a claim can be checked months later, tracked as a share over time next to your competitors. Every one of those runs costs us money per question, which is worth saying plainly, and your first one costs you nothing. One domain, measured against you, no card and no call. The page describing how it works is here, and the argument for why this is a ranking worth tracking at all is its own note.
Either way, do the manual version this week. An hour spent finding out whether you exist in the answers is the highest return hour available to most companies right now.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.