Why doesn't ChatGPT recommend my company?

First, check that it is really happening

AI engines are inconsistent. In one study, 600 volunteers ran the same 12 buying prompts 2,961 times and the brands named changed from run to run, often within the same hour^[2]. So one ChatGPT answer that leaves you out is not a verdict. Before you change anything, run each of your buyers' questions ten or more times, on more than one engine, and count how often you are named against each competitor. If you are absent in nine runs out of ten while a competitor is named in seven, you have a real gap. If it is four against five, you have noise.

That count is the whole measurement. Everything below assumes you have it.

The five reasons ChatGPT names your competitors and not you

1. You are not on the pages it reads

This is the cause nine times out of ten. When ChatGPT answers "best X for Y", it retrieves a handful of pages and writes the answer from them. In our own category those pages are Reddit threads (in 31 percent of answers), third-party "best agencies" lists, review directories and a few publishers^[3]. A single list page took 562 citations in 30 days^[3]. If that page lists your competitor and not you, the engine copies the competitor's name into thousands of answers and never sees yours. Your own site barely features: in our ChatGPT audit of 200 citations, owned pages were 9 of them, review sites 70, vendor pages 48 and Reddit 42^[3].

2. You are on them, but as a passing mention

A Reddit comment that says "we tried X, it was fine" and a review page with two ratings are both technically presence. Engines weigh what the source says, not just that it names you. Competitors who are described as "the one most people recommend for Y" in several independent places get recommended; a brand that appears once in a thread about something else does not.

3. The engine cannot read your own site cleanly

Owned pages are a small share of citations, but they are the share you control, and ChatGPT pulls 44 percent of its citations from the first third of a page^[4]. A home page that opens with a slogan and no plain statement of what you do, for whom, gives the engine nothing to quote. Google's own guidance for its AI features says the same: there is no special markup to add, the page has to answer the question in readable text^[5].

4. The engine confuses you with someone else

Similar names, a new domain, or a better-known company in a neighbouring category can send the engine to the wrong entity entirely. Ask ChatGPT "what is [your brand]" and "is [your brand] legitimate" and read what it says about you and which pages it cites. If it describes a different company, the fix is an entity problem: consistent name, address and description across your site, your business register entry, LinkedIn and the review directories, so every source agrees on who you are.

5. The question you care about is not the question buyers ask

Brands often test "best enterprise-grade X platform" when buyers type "what should a 10-person team use for X". Different phrasing retrieves different pages. Build your test set from the way customers describe the problem in support tickets, sales calls and Reddit, not from your positioning document.

How to find out which one is yours, in about an hour

1. Write down 10 to 15 questions a buyer would ask before choosing in your category. Include comparison and "is it worth it" phrasings, not only "best X".
2. Run each one on ChatGPT, Gemini and Google AI Mode. Note who is named and open every source the answer cites.
3. List the cited URLs. Mark which name a competitor, which name you, and which name nobody. The ones that name a competitor and not you are the gap.
4. Sort that list by how often each URL came up. The top five are where the answer is actually written.

That is the method behind our free audit, run on five engines with more repetitions. It comes back with the mention map, the competitor contrast and the source list, so the hour is optional: request the audit and read the result instead.

What moves an AI recommendation

Three things, in order of how much they move the answer in our tracking.

• Get onto the cited pages. A specific pitch to the editor of the "best X" list that the engines copy from. A helpful, disclosed answer in the Reddit thread they cite, from an account with real history there. A completed profile and honest reviews on the directory that comes up for your category. One placement on a page cited daily is worth more than fifty posts nobody retrieves.
• Make your own pages quotable. The question answered in the first paragraph, in plain words. Comparison tables. A definition the engine can lift. The GEO research paper that gave the field its name found that adding citations, quotations and statistics to a page raised its visibility in generative answers by up to 40 percent in its benchmark^[6].
• Fix what the engines get wrong. Where an answer says something outdated or hostile about you, trace it to the source page and correct it there. The engine repeats its sources; it does not reconsider on its own.

What does not move it: adding a chatbot, writing an "AI optimised" blog post nobody links to, or stuffing the word ChatGPT into your meta description. The engine is reading other people's pages about you. Change those.

How long it takes

Engines re-read and re-weigh sources on their own schedules, so a new placement takes weeks to show up in answers. In our campaigns, measurable movement in share of voice usually takes 6 to 10 weeks, and the placements keep working after that because the pages stay cited. The timeline in detail.

What Redaccel does about it

We run the measurement above for you first, free, on five engines. The audit shows which engines name you, which name your competitors, and the exact pages behind each answer. Then we work only those pages: Reddit threads, lists, review sites, comparison pages, YouTube, articles and your own site, in the order the citation counts say. Progress is reported as share of voice on the same questions every month. If it is not moving by month three, the report says so.

Frequently asked