Why ChatGPT recommends your competitors instead of you

You ask ChatGPT for the best tools in your category and it names three competitors and not you. It's frustrating, especially if you think your product is better. The good news: the reasons are usually concrete, and most can be checked. This guide walks through the common causes, how to tell which one applies, and what to do about each.

A caveat up front: OpenAI doesn't publish how ChatGPT chooses products, and answers vary between runs and model versions. What follows is based on OpenAI's documentation, our own published data, and general search practice, labeled as such.

First, make sure it's a pattern, not one answer

Before diagnosing anything, check that you're really missing. One answer is a weak signal. In our AI rankings run on 26 October 2026 (100 ChatGPT answers across 20 SaaS categories), 151 of 264 product-and-category pairs (57%) appeared in only one of the category's five answers. Plenty of products show up in one answer and not the next.

Ask 10 to 20 real buyer questions in clean sessions and count how often you appear. Our guide on how to check if ChatGPT recommends your product covers the method. If competitors appear in most answers and you appear in few or none, keep reading.

Reason 1: You're not on the pages ChatGPT reads

When ChatGPT searches the web, it builds its answer from the pages it finds. OpenAI's help center says ChatGPT search rewrites a question into targeted queries for third-party search providers. For "best X software" questions those queries tend to return review articles, roundups, comparison sites and directories. If competitors are on those pages and you aren't, the answer follows the pages.

What our data shows:

• 83 of 100 answers cited at least one source, and together they linked 424 different websites. Only 204 of those sites were cited inline to back a specific claim; the rest appeared only as image or link cards.
• One publisher stood out: techradar.com was cited in 32 answers across 13 of 20 categories. toolradar.com appeared in 14 answers across 9 categories.
• G2 appeared in 3 answers; Reddit and Capterra in none. Don't assume which sources matter. Look at what's cited for your category.

How to tell: ask your buyer questions with search on and list the cited domains. Open each one. If they list your competitors and not you, this is your main problem.

What to do: get honestly included. Pitch editorial roundups with a clear positioning line, a free account to test, and customer proof. Claim and complete directory and review-site profiles. Ask real customers for reviews. Then re-check whether those pages are still being cited.

Reason 2: The answer comes from training data, and incumbents dominate it

Not every answer searches the web. In our data, 17 of 100 answers cited no sources at all, so those products came from what the model already knew. Training data reflects years of articles, so long-established products have an advantage, and recent changes may be missing.

A telling example: an email marketing answer with no cited sources recommended "Sendinblue" and "ConvertKit", names those companies have since replaced (they're now Brevo and Kit).

How to tell: if competitors keep appearing in answers without citations, the model is answering from memory.

What to do: you can't edit training data, but you can make sure current, consistent information about you is widely published: on your site, in directories, in coverage and in reviews. Over time that's what both search-based answers and future models draw on.

Reason 3: The question is broad, and you win on narrow ones

Broad questions ("best CRM") favor the default names. Narrow questions open the field. In our data, the question "Recommend an affordable [X] with good support" put a product outside the category's top 3 in first place in 16 of 20 categories, such as Hesk (help desk), Patriot Accounting (small-business accounting) and Less Annoying CRM (CRM for startups).

How to tell: test questions that match your real strengths: a price point, an audience, an integration, a region. If you appear there but not in broad questions, that's normal for a challenger.

What to do: make those strengths explicit and findable. Publish "best [category] for [audience]" and use-case pages, and make sure third-party profiles mention the same specifics. Then track the narrow questions as seriously as the broad ones.

Reason 4: ChatGPT doesn't clearly understand what you are

If your site, profiles and coverage describe you inconsistently, as a "platform" here and a "suite" there, or under an old name, it's harder for any system to match you to a category question.

How to tell: ask ChatGPT directly: "What is [your product]? Who is it for? What does it cost?" Wrong or vague answers point to an entity problem.

What to do:

• Use one product name and one category label everywhere.
• Write one plain positioning sentence ("[Product] is [category] for [audience] that [difference]") and reuse it on your homepage, meta description and every profile.
• If you renamed, say "formerly [old name]" clearly and update old profiles.
• Add Organization and SoftwareApplication structured data with sameAs links to official profiles, matching what's on the page.

Reason 5: ChatGPT can't read your site

OpenAI's documentation says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers (apart from navigational links). Robots rules, CDN bot protection or firewalls can block it without anyone noticing.

How to tell: check robots.txt for rules affecting OAI-SearchBot or all user agents, check your CDN or WAF logs for blocked requests, and make sure key pages serve their main text in the HTML.

What to do: allow OAI-SearchBot, allowlist OpenAI's published IP ranges if your firewall needs it, and verify your site in Bing Webmaster Tools and Google Search Console with a clean sitemap. OpenAI names Bing as one of its search providers in its documentation.

Reason 6: Competitors publish the pages that answer the question

Buyers ask comparative questions: "X vs Y", "alternatives to Y", "best X for agencies". If a competitor has a clear, useful page for each of those and you don't, the search results, and so the answer, lean their way. Vendor sites do get cited. In our data, Gusto's own site was cited in 3 of 5 HR-software answers, and Trello's, Pipedrive's and Heap's in 2 of 5 in their categories.

What to do: publish fair comparison pages (including when the competitor is the better fit), "alternatives" pages that list several real options, and use-case pages. Put direct answers near the top and keep facts like pricing in visible text.

Reason 7: You're named, just not first

Sometimes the problem isn't absence but position. In live chat software, Crisp and Tidio were each named in 4 of 5 answers in our data but never first, while Tawk.to was named in 3 and came first in all 3. In product analytics, PostHog appeared in 3 of 5 answers, behind Mixpanel and Amplitude (5 of 5 each), but was named first every time it appeared.

What to do: track position as its own metric. Look at the answers where a competitor comes first and read their cited sources. The framing on those pages ("best overall", "best for startups") often explains the order.

A quick diagnostic checklist

1. Am I missing in most answers, or just one? *(Measure first.)*
2. Which domains are cited, and am I on them? *(Reason 1)*
3. Do competitor-heavy answers have no citations? *(Reason 2)*
4. Do I appear on narrow questions but not broad ones? *(Reason 3)*
5. Does ChatGPT describe my product correctly? *(Reason 4)*
6. Can OAI-SearchBot crawl my site? *(Reason 5)*
7. Do competitors have comparison and "best for" pages that I don't? *(Reason 6)*
8. When I'm named, where am I in the list? *(Reason 7)*

What not to do

• Don't try to trick the model with hidden text or instructions aimed at AI crawlers. Hidden text breaks search engine spam policies and isn't how answers are built.
• Don't buy or fake reviews. It breaks review platforms' rules and undermines the signals you need.
• Don't judge progress week to week. Look at four to eight weeks of the same questions.

The short version

ChatGPT usually recommends competitors because they're on the pages it reads, they're established in its training data, the question is broad, or it understands them more clearly than you. Most of that can be checked and fixed: get onto the cited sources, make your positioning consistent, unblock crawlers, publish comparison and use-case pages, and track share of answers and position over time.

For engine-specific steps, see our guides for ChatGPT, Perplexity and Gemini.