Why the AI tools name a rival and not you: what 1,685 citations from one account showed

When an AI tool answers a buyer question with a rival's name and not yours, it is usually not choosing between you. It is repeating a list it read. We can see this in the citation data our tracker stores with every answer. This post walks through one account's citations, explains why roundup pages carry so much weight, and ends with a checklist for getting onto the lists that decide your category.

The pattern across the platform

Two numbers set the scene, both measured on the AEOTrack platform as of September 2026:

• 52 of 100 businesses score 0 on their first check. No answer to any tracked question named them.
• In 35 of 43 companies measured (81%), a competitor was named in an answer where the company was not.

So the usual first result is not "the engines have never heard of your category". It is "the engines answered, and they named someone else". The question is where those names came from.

One account, 34 questions, 1,685 citations

Every answer we read arrives with the sources the engine used, because every engine runs with web search on. For one tracked company we counted the citations behind its 34 questions during the 27 to 29 September 2026 reviews:

Measure	Count
Tracked questions	34
Citations across all stored answers	1,685
Distinct URLs cited	640
Distinct domains cited	380
Citations per question (derived from the two rows above)	about 50
Of the ten most-cited sources, "best tools" roundup lists	4

Two things stand out. First, the spread: 380 domains for 34 questions means the engines are not reading one or two authorities. Second, the concentration at the top. Four of the ten pages cited most often were roundup lists with a title like "the best tools for X". A roundup is a ready-made answer. It already has the shape the engine wants to return: a ranked list, a sentence per product, sometimes a price. When the engine's own search finds that page, the shortlist is largely settled before your homepage is read.

This is why a well-built site can still score 0. The free audit grades homepage readiness from A to F, and a good grade next to a visibility score of 0 is common enough that our FAQ answers it. Readiness gets you read. It does not get you listed.

Why lists beat homepages

Three reasons, all visible in stored answers.

1. The list matches the question

A buyer question is a comparison: "best motor trade insurance for small businesses", "top taxi insurance providers in the UK". Your homepage answers "what is Acme". A roundup answers the comparison directly, so it is retrieved for it.

2. The list is corroborated

Engines consult several sources per answer. If four of them agree on the same three names, those names are safe to repeat. A single page that names you is one vote against a room.

3. The list carries rank

Our per-answer score gives 45% of the weight to prominence, which we read from the position in the list that names you (see how the score is calculated). Engines often keep the roundup's order. Being third on the page that gets cited is worth more than being everywhere else.

How to see your own version of this

Open any question where a rival was named and not you. The answer view lists the sources the engine read. The Sources page groups those across every question by domain and counts how often each one appears. Your version of the table above is there in about a minute. If you have not tracked anything yet, the free audit asks two engines one question about your category and shows who they named and the pages the names were pulled from.

The checklist: getting onto the lists

1. Take the questions you lose. Start with the ones where the answer names rivals. A question where nobody is named is advice-only and is left out of the score; it will not help you here.
2. Rank the cited domains. Sort the Sources page by the number of answers citing each domain. The top ten decide your category. Ignore the tail for now.
3. Mark each one: listed, missing, or below the fold. Read the page as a buyer. Are you on it? Being named last on a list is a different task from not being there.
4. Find out how the page is maintained. Most roundups are either editorial (a writer updates them) or directory-style (vendors submit a listing). Editorial pages have an author and a contact. Directory pages have a submission form and criteria. Do the one that applies.
5. Give the page what it needs to list you. A one-paragraph description in plain words, your price, the category you belong in, and one fact that separates you from the names already there. Most listings are missing because nobody sent this.
6. Write the comparison you are missing from. If no page compares you to the names being cited, publish one on your own site. Name the rivals, be specific, and keep each claim in a sentence that survives being lifted out.
7. Re-check after the page changes, not before. The page has to be updated, crawled and then retrieved for the question. Expect weeks. Then run a re-check on that question and look for the citation first: engines start reading a page as a source before they repeat what it says.
8. Do not buy placements the engines do not read. A mention on a domain that never appears in your sources does nothing for your answers, whatever it does for traffic.

What to expect

The first change is usually one engine on one question. In the tracker that is one row turning from "3 rivals named, not you" to "Named". Citations move first, mentions second, and the score moves with its ± range, so read a move inside the range as normal variation. We publish the monthly counts of answers read, cited and named on the Named Report page; the first month is pending.

The related post How to earn citations from the sources ChatGPT already trusts covers the outreach in more detail, and Cited but not mentioned covers the half-won case where the engines read you and still name someone else.

Check your own category

Type your domain into the free audit. We ask two engines one question your buyers ask and show you who they named: you, a competitor, or nobody, with the sources. Free, no account, about a minute. Then track five engines every week on a 14-day trial, no card.