Why does ChatGPT recommend my competitor instead of me?
In short
Because the pages ChatGPT read while answering name your competitor and not you, or name you less clearly. Find the searches it ran and the pages it cited, then read what those pages say about the rival: usually a clear statement of who it's for, a price, a comparison table, or a place on a roundup you're missing from.
Keats, from PressRoom. Updated Read as markdown
It's a frustrating thing to see: you ask ChatGPT for the best option in your category, and it confidently lists your competitor, sometimes in words lifted from their homepage. It isn't personal and it usually isn't random. It's a reading problem, and it can be diagnosed.
The short version of how the answer was built
For a buying question, ChatGPT normally searches the web before answering. It writes several searches of its own (its fan-out searches), reads a handful of the pages that come back, and builds the answer from what those pages say. The brands it names are, mostly, the brands on those pages.
So "why does it recommend them?" turns into three narrower questions you can answer:
- Which searches did it run?
- Which pages did it read and cite?
- What do those pages say about the rival, and about you?
Step 1: read the searches
The searches are often more specific than the person's question. Asked a general question about AEO tools, ChatGPT in our own sample ran a single search that already contained five product names (the sample, as a CSV). That's a strong tell: if the search itself names your rival, the model already associates the category with them, and the pages it finds will reinforce it.
If you can see the searches (some tools record them; Keats does, for every answer), note the words that recur: the category name it uses, the qualifiers ("for small teams", "cheap", "2026"), any brand names. Those words are the vocabulary your pages need to match.
Step 2: read the cited pages
Open every page the answer cited. For buying questions they tend to fall into three groups:
- Roundups and listicles: "the 10 best X for Y". Whoever is on them gets named.
- The rival's own pages: product pages, pricing, help docs. Assistants cite vendors' pages when they look a product up by name.
- Community and review pages: Reddit threads, review sites, YouTube videos.
Count how many of the cited pages mention you at all. If the answer is zero, there's nothing for ChatGPT to name you from, and no amount of polishing your homepage alone fixes that question.
Step 3: compare what the pages say
Put the rival's cited page beside your closest equivalent and read them as the model would: quickly, looking for the sentence that answers the question. Common differences:
| What the rival's page has | What yours often has instead |
|---|---|
| "X is an exam prep course for working adults" in the first line | A slogan and a hero image |
| A visible price | "Contact us" |
| A comparison table against named alternatives | No mention of alternatives |
| A dated, specific claim (pass rate, number of practice questions) | Adjectives |
| The buyer's words ("for people with full-time jobs") | Your internal category name |
Each row is a fix. None of them is a trick; they're the facts an answer needs, stated where a reader can find them.
Step 4: check you're readable at all
Before content work, check the plumbing. OpenAI says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers". If your robots.txt blocks it, or your key pages render only in JavaScript a crawler can't run, the rival wins by default. Our guide to AI crawlers and robots.txt has the list of user agents to check.
Step 5: fix in the order that works
Once you know why, the order of the work is the one in our guide to getting named in ChatGPT: correct anything wrong it says about you, update the pages it already reads, write the page it's missing, then get onto the third-party pages it cites. For category questions the last step often decides the answer, and our guide to getting onto the lists AI cites covers how.
Don't judge it by one answer
The rival may be named first in today's answer and third in tomorrow's. Before you rebuild your site around one screenshot, ask the question several times on several days and count. A rival named in nine of ten answers is a different problem from one named in four of ten.
Here's the head-to-head view in Keats for Prepco, our made-up demo brand: which questions each rival wins, and from which pages.
What not to do
- Don't attack the competitor. Assistants don't reward it, and it reads badly when a buyer finds it.
- Don't stuff their name into your pages. A fair comparison page that names them is useful; a page repeating their name for its own sake isn't.
- Don't expect a week. Answers change as sources are re-crawled; read the result after three to five weeks, against questions you didn't touch.
Sources
- Keats: eight ChatGPT answers to buyers' questions about AEO tools (CSV), read November 13, 2026
- OpenAI: Overview of OpenAI crawlers (OAI-SearchBot), read November 13, 2026