Why AI models cite competitors instead of you
You ask ChatGPT the question your buyers ask, and it confidently recommends your competitor. Not you. It feels arbitrary — but it almost never is. Models cite competitors for specific, addressable reasons. Here are the common ones, in roughly the order worth checking.
1. Their page answers the question; yours sells
The single biggest pattern. Assistants reward pages that *directly answer* a question in plain language, and quietly demote pages that read like landing-page marketing. If a buyer asks "what's the best X for small teams?" and your competitor has a page that opens with a clear, qualified answer — "For small teams, the main trade-offs are A, B, and C…" — while your page opens with "The #1 platform trusted by industry leaders," the model has an easy quote from them and nothing safe to lift from you.
Fix: publish pages that answer the real question in the first paragraph, in prose a model can quote verbatim. Save the superlatives.
2. Their claims are quotable; yours are vague
Models prefer to ground statements they can attribute with confidence. "Processes refunds in under two minutes" is quotable. "Lightning-fast refunds" is not — it is an adjective, and a careful model won't stake a recommendation on it. Specific, checkable claims get cited; mood words get skipped.
Fix: replace adjectives with facts. Numbers, timeframes, supported integrations, concrete limits.
3. They are corroborated across sources; you are an island
Retrieval-grounded answers lean toward facts that show up in more than one independent place. If three reputable sites describe your competitor's strength the same way and only your own homepage makes your claim, the model trusts the corroborated one. This is the AI-era version of off-site authority.
Fix: earn third-party mentions — reviews, comparisons, directory listings, press. Consistency of facts across them compounds.
4. Your page isn't crawlable or extractable
Sometimes the content is fine but the model can't read it. Key facts rendered only in client-side JavaScript, important claims trapped in images, or no structured data — all of it raises the cost of extracting your facts versus a competitor whose page hands them over cleanly.
Fix: make sure the important content is in the server-rendered HTML, and add JSON-LD (Organization, Product, FAQPage) so the facts are machine-readable on the first pass. See How structured data influences AI discovery.
5. You're mentioned — just not recommended
A subtle but common case: the model *does* name you, but as the runner-up. "You could use [you], but I'd recommend [competitor] because…" You are in consideration; you are losing on a specific reason the model states out loud — price, a missing feature, a use-case fit. That stated reason is a gift. It tells you exactly what to address.
Fix: read the "because." Then close that specific gap with content, evidence, or a corrected fact.
How to find your actual reasons
The reasons above are the menu; you need *your* order. That means looking at the real answers: which questions cite a competitor, which page of theirs got cited, what claim the model rewarded, and what you have (or don't) on the same topic.
Lumear does this systematically — it runs the questions your buyers ask, captures the competitor citations beating you, and points at the specific page on your site that should have been cited instead, with the change that would tip it. The recommendations are ranked by expected impact, so you work the gaps that move visibility most.
The encouraging part: every reason on this list is something you control. Competitors aren't winning because the model likes them. They're winning because, on that question, their page is the better answer. Make yours better and the citation moves.