When an assistant answers a category question without naming anybody

There are three ways to lose an answer and only two of them get discussed. You can be absent while a competitor is named. You can be absent because the app names nobody. And you can be present but described wrongly. The second is the one that gets mistaken for the first, and it needs a different fix. On 2026-10-16, in the scan published on our proof page, the middle row has a worked example. ChatGPT read 6 sites on a question about the completeness of Indian court record search, then wrote its answer without putting a single business in it.

The three outcomes, and why they are not the same problem

Outcome	What it says about the sources	What it asks you to do
A competitor is named and you are not	A source makes a comparative claim, and it is not about you	Read that source and publish the equivalent specifics
Nobody is named	No source read was willing to settle the comparison	Publish the thing that would settle it, with a date
You are named with wrong facts	A source about you exists and is out of date	Correct the source, which is usually faster than either of the above

The middle row is the unusual opportunity, because the slot is empty rather than occupied. Displacing a named competitor means persuading a system to prefer your claim to theirs. Filling an empty slot means being the first to publish a claim in a form something can use.

It is also the row most likely to be recorded wrongly. A log that only has a column for whether you were named will mark this reading identically to one where three rivals were listed, and the two call for completely different work. So the field that has to be in the record is not a yes or no about your own business. It is who was named, which includes nobody as a valid and informative answer.

How often it happens, in our own readings

Often enough to plan for. In the scan on the proof page, one of the two published questions came back from ChatGPT with no business named, which is one of the four readings printed. On a much larger set the pattern shows up as a source mix rather than a count of empty answers: across 1,260 readings on ChatGPT and Gemini on 16 October 2026, 20,143 citations were recorded and 12.3% of them went to official platform documentation. Documentation describes how something works and names no supplier, so a reading that leans on it is a reading pointed away from any business.

Why an app names nobody

Not because it has nothing to say. The answer in this case was substantive and ran without any business in it, which is what happens when the sources read describe how a thing works rather than who does it best. A comprehensiveness question needs an enumeration to compare: which courts, which record types, how current. Where no source publishes that in a comparable form, a careful system writes about the problem and leaves the names out.

Our wider readings make the same point from the other direction. Across 597 usable readings on ChatGPT and Gemini on 15 October 2026, 10,652 citations were recorded, and 9.7% of them went to official platform documentation while 85.6% went to the site of a company or agency selling the thing being recommended. Documentation explains mechanisms and names nobody. When the documentation is what gets read, an answer with no business in it is the natural output.

How to tell which of the three you have

• Read the whole answer, not the name check. An answer with no business in it reads as a near miss and is a different result from one listing three rivals.
• Count the sources it read. A high count with no business named means it looked and found no comparison, which is more encouraging than it sounds.
• Ask the same question on a second app. In this scan Perplexity named two publishers on the same question after reading 8 sites, so one app found something to name and the other did not.
• Check an adjacent question. The same scan had both apps naming this company first on a pricing question, so the retrieval reaches it and the gap is the claim rather than the presence.

What to publish into an empty slot

The specific object the question is asking for, in a form something can quote: a list rather than an adjective, with a date, a scope and a statement of how often it is refreshed. That shape is not a guess. Put the question to ChatGPT and Gemini, as we did across 102 conversations on 15 October 2026, and the same fields come back: every conversation wanted a date, a last verified line and a refresh interval, 98 wanted the engines listed by name instead of lumped together, and 80 wanted the sample size. A page that answers a comparison question and carries those fields is doing the thing the engines said they were looking for.

Then fix the surface as well as the substance, because a page nothing can read is the same as a page that does not exist. The practical sequence for that, including fetching your own page as each named crawler and reading what comes back, is in the workflow standard we publish, which puts that check at the station most teams skip.

Questions about an AI answer that names no business

Is a generic answer better or worse than seeing a competitor named?

Better, in the sense that nobody is occupying the position. It is also a signal that the category has not published what the question needs, which means the first adequate page can do a lot of work.

Does it mean my category is too small to be answered?

Not necessarily. It more often means the comparison is unsettled in public, which is common in categories where the detailed specifics sit behind sales conversations rather than on pages.

Should I write a comparison page naming my competitors?

Publish your own specifics first, with the date you checked them. A comparison written by a seller about rivals is the kind of page these systems discount, and your own enumeration is the part nobody else can write.

How do I know the app really looked?

The source count is the evidence, and it belongs in the record for every answer. 6 sites read with no business named is a materially different reading from no sources read at all.

Who fixes this, and what does it cost?

It is page work rather than technical work in most cases, which is slower than it sounds because the specifics have to be gathered before they can be published. We sell both halves of that, the measurement and the pages, which is the interest to read this page with.

An empty slot does not stay empty

The reason to treat this as the most urgent of the three outcomes is that it is the least stable. A question with no business named in the answer is a question any competent publisher in the category can answer first, and once a source exists that settles the comparison, every later answer tends to lean on it. The window is the period in which nobody has written the thing down, and that window closes without any announcement.