
Oct 26, 2026
Why ChatGPT recommends your competitor instead of you (and how to fix it)
Your CEO asked ChatGPT for the best tool in your category and read out three names. None was yours. Two of them rank below you on Google. The question lands on the SEO manager's desk with a second one attached: "what are we doing about AI search?" This guide explains why the engine picked your competitor, gives you the audit to find your own cause in an afternoon, the seven fixes in the order they pay off, and a plan for this week.
Disclosure: GetMint publishes this guide and makes one of the tools mentioned. We name it where it does the job described; the audit and the GA4 steps work without it.
TL;DR
ChatGPT names brands from what the model learned in training and from pages it retrieves live when it searches. On category prompts, your competitor usually wins on retrieval: more third-party pages name it.
Five common causes: no third-party pages naming you in the category, inconsistent brand facts from site to site, product pages with no comparison or pricing context, pages the retriever can't read, and no reviews where the engine reads them.
Each engine searches a different index, so a competitor can win on ChatGPT and lose on Claude. Work on the engine your buyers use first.
Diagnose by reading answers where your competitor wins: open the sources, find which of the five causes applies and which signals they have that you lack. That tells you which fix comes first. Your share of mention is a separate number, and it takes a tool that samples every prompt repeatedly.
In GA4, ChatGPT, Gemini and Copilot visits now land in the native AI Assistant channel (since May 2026). Perplexity still arrives as Referral, and Claude may too. The full setup is in our GA4 guide.
What does it mean when ChatGPT recommends your competitor?
ChatGPT builds a brand recommendation from two layers. The first is model memory: what the model absorbed in training, which favors brands written about often and consistently before the training cutoff. The second is live retrieval: when the question calls for it, ChatGPT rewrites the prompt into one or more searches, sends them to its search providers (Bing among them), opens a few pages, and names the brands those pages name (OpenAI). The source links under the answer show you which pages did the naming.
The same category question yields different names with a different qualifier. "Best CRM" and "best CRM for a 10-person agency on a budget" pull different pages and different brands. You are competing in a set of results per intent, per buyer stage and per qualifier, and your competitor may win the budget prompts and lose the enterprise ones.
Why it is a revenue question and not a PR one: 85% of B2B buyers purchase from their "day one" list, the vendors they had in mind before searching (Bain & Company, 2025), and a growing part of that list now gets built in a chat window. A brand absent from the answer is absent from the list. Search Console won't show the loss, because no search occurred on Google.
Why does ChatGPT recommend a competitor instead of you?
If your competitor keeps getting recommended by ChatGPT instead of you, the reason sits in the sources under the answer. Open three answers where they win, click the source links, and check which of the five causes below each source points to.
No third-party corroboration. The engine trusts a review site, an analyst page or a comparison article ahead of a vendor. Your competitor is on G2, Capterra, two "best X tools" roundups and a Reddit thread. You're on your own site.
Entity inconsistency. Your company is "Acme", "Acme Software" and "Acme.io" on different pages, with three different one-line descriptions and two different founding years. The model is unsure what you are, so it picks the brand it is sure about.
No comparison context on your pages. Your product pages say what you do. The competitor's pages say who they're for, what they cost, and how they compare with you. The engine quotes the page with the answer to the buyer's question.
Pages the retriever can't read. Text rendered by JavaScript after load (most AI crawlers do not execute it, per Vercel's crawler analysis), cookie walls that hide the content, a robots.txt rule blocking OAI-SearchBot. The page exists for humans and stays invisible to the retriever.
No reviews where the engine reads them. Star ratings on your site count for little. Reviews on the platforms the engine cites for your category count for a lot.
How ChatGPT weighs trust signals and brand authority
When AI answers about a brand, most of the pages it cites are third-party. In PR software company Muck Rack's May 2026 study of 25 million AI citations, earned media (any page the brand neither owns nor pays for) made up 84% (Muck Rack), and a University of Toronto team found the same bias against brand-owned pages in controlled tests (arXiv, 2025). In practice the sources under recommendation answers are third-party category pages and roundups, review platforms, expert quotes and case studies published off your domain, press coverage, awards and analyst mentions, and founder or executive profiles. Your own site confirms what those pages say.
Generic positioning fails here in a specific way. "The number one platform for modern teams" gives the engine nothing to match against the buyer's qualifier. "CRM for agencies under 20 people, from $29 a seat" matches the prompt word for word.
Common reasons your content reads as less reliable
Claims with no number and no date. "Trusted by thousands" gives the engine less to quote than "412 agencies as of June 2026" (an illustrative example).
No author, no date, no update history on the page.
Pricing kept behind "contact us" when the competitor publishes it.
The same paragraph on ten pages, so the retriever finds no one page worth quoting.
Marketing superlatives stacked in the first 100 words, where the engine looks for the definition.
How each engine picks its sources: ChatGPT, Perplexity, Gemini and Claude
The engines agree less than most teams assume. Each one searches a different index, so the same prompt often returns different sources, and different brands, on each.
ChatGPT | Perplexity | Gemini | Claude | |
|---|---|---|---|---|
Search backend | Third-party providers, Bing among them, plus partner content; the prompt is rewritten into searches (OpenAI) | Its own index; searches on every question | Google Search, with query fan-out across subtopics (Google) | Brave Search (Xponent21) |
Sources it reaches for | Wikipedia is its top cited domain, then Reddit (Profound) | Reddit is its top cited domain, and it over-indexes on review sites such as G2 (Profound) | Pages that already rank in Google Search, plus forums and video | Pages that already rank in Brave results, often tech review sites and directories |
What to work on first | Roundup placements, Wikipedia and Wikidata, community threads | Reddit discussion and recent reviews | Your Google rankings, Reddit and video | Your visibility in Brave results, review sites and directories |
One content plan won't cover the four. Pick the engine your buyers use most (the GA4 channel split below tells you) and fix its ecosystem first. Google also states that no special schema markup is needed to appear in AI Overviews or AI Mode (Google), so treat structured data as hygiene and put the effort into sources.
How to audit your brand's visibility on AI engines
The audit is a diagnosis: it tells you why your competitor wins, not by how much. It takes an afternoon.
Build the prompt set. Twenty-five prompts in your category, written as buyers talk, split five ways: category ("best X"), qualifier ("best X for small teams"), comparison ("you vs competitor"), problem ("how do I solve Y"), and pricing ("how much does X cost"). Freeze it: it is also the prompt set your tool will track later.
Read the answers on four engines. ChatGPT, Perplexity, Gemini and Claude. Ask each prompt a few times, since AI answers change run to run: the same question asked twice on the same day can name different brands and cite different pages. For each answer, note who gets named, who gets recommended first, and why.
Open the sources. Click the source links under the answers your competitor wins. For each page, ask which of the five causes it points to: a roundup that names them and not you, a description of you that contradicts your site, a comparison page you lack, a review platform where you have no reviews.
Check six signals, for you and one competitor, present or absent:
Signal | You | Competitor | Where to check |
|---|---|---|---|
Named on a third-party roundup cited in the answers | The source links | ||
Consistent name and description on the top 10 sources | Google the brand, read the snippets | ||
Comparison and pricing context on the product page | The page | ||
Schema markup present and valid | Schema Markup Validator, Rich Results Test | ||
Text rendered without JavaScript | View source, or a text-only fetch | ||
Reviews on the platform cited for the category | The source links again |
The signals your competitor has and you lack are the fix list, in order. This is a checklist, not a score to track: a row that reads "present" for them and "absent" for you on the pages that keep winning is no longer a mystery.
Note the engine differences. A rival that wins every answer you read on one engine and rarely on another points to which ecosystem from the table above to work on first. For the full competitive benchmarking method, see how to compare your brand vs competitors in AI answers.
Get the number from a tool. The audit does not produce a share of mention.
Why a hand count misleads. AI answers are sampled, not fixed. In a SparkToro and Gumshoe study, 600 volunteers ran 12 identical prompts nearly 3,000 times on ChatGPT, Claude and Google's AI, and the odds of getting the same list of brands twice were under 1 in 100 (SparkToro, January 2026). What holds steady is how often a brand appears across many runs, and measuring that takes volume. On 75 answers (25 prompts run three times), a 50% mention rate carries a margin of about ±11 points at 95% confidence; on 15 answers, about ±25 points. A hand count can't tell a real move from chance. Read answers by hand to understand them; count them with a tool that runs every prompt many times.
A share of mention you can report, per engine and against one competitor, comes from the same frozen prompt set run repeatedly, on each engine, over several weeks.
How to fix it: seven steps to earn ChatGPT citations
The order matters: fix the sources before the site, because third-party pages carry most of the weight. Then make the site quotable. Then track the same frozen prompt set with a tool that runs every prompt repeatedly, so you know which fix moved what.
Get on the pages already cited. Pull the source list from the audit. For each roundup or comparison page naming your competitor and not you, contact the author with a factual one-paragraph entry. Two placements move more than a year of blog posts. GetMint's AI Visibility Gap view builds this list for you: the prompts you lose and the page cited instead.
Write one canonical company description and use it everywhere: site footer, LinkedIn, G2, Crunchbase, press boilerplate. Name, category, who it's for, one number.
Add comparison copy to the product pages. Who the product suits, who it doesn't, what it costs, and a section naming the two competitors the engine keeps recommending, with an honest difference.
Publish answer pages. One page per buyer question from the prompt set, with the answer in the first two sentences, a number, a date and a source line. Format and structure are covered in how to create content that gets cited by ChatGPT.
Ship the machine layer. Organization, Product and Article schema with a Person author. Author bios with credentials. Server-rendered text. In robots.txt, allow OAI-SearchBot: it decides whether your pages appear in ChatGPT search answers. GPTBot only governs training, so blocking it does not remove you from answers (OpenAI crawler docs).
Collect reviews where the engine reads. Ten recent reviews on the platform cited for your category beat two hundred on a widget on your site.
Publish one original number a quarter. A benchmark from your own data, dated and sourced, is the page every engine quotes, because nobody else has the number.
Create a surround-sound content plan
Editorial mentions, expert roundup inclusions and third-party review coverage act as corroboration. The engine treats three independent pages saying the same thing about you as more reliable than one page of yours saying it louder. Brands with strong SEO and weak AI recommendations nearly always fail here: the domain ranks, yet nobody else vouches for the brand.
Don't manufacture it. Bought reviews, spun mentions on link farms and AI-written testimonials get detected by the platforms hosting them and, once removed, leave a hole where the corroboration was. Two real placements a quarter outperform twenty fake ones.
Strengthen entity signals on trusted sources
Claim and complete the profiles the engines read: Wikipedia if you qualify, Wikidata, Crunchbase, LinkedIn company page, G2 or Capterra or the review platform for your category, Google Business Profile for local. Same name, same description, same founding year, same category on each. The engine assembles its picture of your company from all of them together.
Publish original data and first-hand knowledge
Engines quote numbers, and they quote the first page to publish a number. A survey of your customers, a benchmark from your product's usage data, a pricing index for your category. Date it, source it, put the figure in the first paragraph, and update it on a schedule with the date changed.
Measure it in GA4
Since May 2026, GA4's Default Channel Group has a native AI Assistant channel that catches visits from ChatGPT, Gemini and Copilot, among others (Google). Perplexity is not in Google's channel definition and still lands in Referral; Claude is named only in Google's announcement, so check your Referral report. Clicks from AI apps that strip the referrer land in Direct. A custom channel group placed above both AI Assistant and Referral puts every assistant on one line. Then read which landing pages AI visitors hit (those are the pages the engines already trust) and where the AI channel sits in your key event paths.
The full setup, with the regex, the channel ordering, the Direct leakage and where answer tracking fits next to the AI line, is in our guide on how to track ChatGPT and AI assistant traffic in GA4.
Where to start this week
For the SEO manager whose CEO asks what the team is doing about AI search, the plan fits in three moves:
Run the audit this week. By the end of the afternoon, the CEO's question has an answer: which pages make the engine pick your competitor, and which signals you lack.
Set up the GA4 AI view next, so you can show the traffic you already get from AI assistants.
Put the frozen prompt set in a tool when the CEO wants a number: your share of mention against one competitor, per engine, from repeated runs over several weeks.
What you'll do yourself: the audit, the canonical description, the schema, the GA4 channel, the comparison copy. What usually needs help: the placements on third-party pages, which take relationships and time, and the measurement at scale, which takes a tool.
What to look for in a tool:
Prompt-level mentions kept separate from citations.
The competitor page named on each prompt you lose.
Daily or weekly refresh, with repeated runs.
A per-market and per-engine split.
Clear pricing and commitment terms.
GetMint is one option. GetMint names the page the engine cited instead of you, prompt by prompt and market by market, keeps mentions apart from citations, and turns each gap into a Content Studio brief, in one platform. Its AI Search Visibility module covers the monitoring, and the AI Visibility Gap view is the placement list for step 1 of the fixes. GetMint has two plans, both sized on a call to your brands, markets and goals. Advanced, on a 12-month commitment, covers 300+ prompts, 5 topics, 3 AI models, 10 briefs and articles a month, unlimited users and 3 countries. Enterprise & Agency adds custom prompts, topics and engines, API access, a dedicated AI search strategist and role-based access. Details on the pricing page. Per-engine pages: ChatGPT tracker, Perplexity tracker.
Peec AI, Profound, Otterly and Semrush's AI Visibility Toolkit also track brand mentions in AI answers. Our comparison of the best AI visibility tools sets them side by side with prices.
To track the audit's prompt set with repeated runs on every engine, book a GetMint demo.
FAQ
Why does ChatGPT recommend my competitors instead of my business?
Because more pages the engine retrieves name them. When ChatGPT searches, it opens a few pages and recommends the brands those pages name. If your competitor appears on the roundups, review sites and forum threads under the answer and you don't, it wins. Click the source links on three answers to see which pages decide.
Does ChatGPT rely more on reviews, brand authority or website content?
Third-party pages carry the most weight. When AI talks about a brand, the large majority of cited pages are third-party: roundups, review platforms, press and forums. Reviews count when they sit on the platforms the engine cites for your category. Your own site mostly confirms facts, so keep it consistent and quotable.
Does classic SEO still matter if AI search is growing?
Yes, as the floor. Retrieval starts from search results, so backlinks, domain authority and page speed still decide whether the engine finds the page at all. They no longer decide whether it names you. Ranking gets you retrieved; corroboration by independent sources and pages that answer the buyer's exact qualifier get you recommended. The good news: writing for the buyer question, answering it in the first two sentences, sourcing the numbers and getting cited off-site helps Google and the engines at once.
Does GA4 track ChatGPT traffic automatically?
Yes, since May 2026. GA4's Default Channel Group now has an AI Assistant channel that catches ChatGPT, Gemini and Copilot visits. Perplexity is not in Google's channel definition and still lands in Referral, Claude may too, and app clicks without a referrer land in Direct. A custom channel group fixes the first gap; our GA4 guide shows how.
How long until ChatGPT starts recommending us?
Expect a quarter. A new third-party placement usually takes a few weeks to start appearing under answers, once the engines have crawled and retrieved it; expect it to come and go day to day before it settles. The first sign of progress is your page cited under an answer that doesn't name you yet. Recommendations follow once enough independent pages corroborate you.
About the author: Jean-Noël Escande is AI Search Strategist at GetMint. He came to AI search from data analytics in commercial real estate, at CoworkIntel and The Instant Group, and now helps brands shape how ChatGPT, Perplexity and Gemini cite and surface them, from performance monitoring and schema to content built for AI retrieval. LinkedIn
BUILT FOR ENTERPRISE BRANDS
Your competitors are already visible in AI. Are you?
See what ChatGPT, Gemini and Perplexity say about your brand. Set up in 5 minutes.

