When AI Recommends Your Competitor Instead of You

The moment you find out an AI system chose someone else

Picture this. A customer texts you a screenshot, or you get curious one slow afternoon and open ChatGPT yourself. You type the question your own store exists to answer. Something like "what's the best portable espresso maker for camping" or "which dog food is best for a puppy with a sensitive stomach," a question squarely inside your category, a question you'd bet your product could win on merit.

The answer comes back fast, confident, and specific. It names a brand. It explains why that brand's option fits the use case. It links out to a page. And the store it names is not yours. It's a competitor's, one you compete with on price, on quality, maybe one you've beaten in every review comparison you've ever seen. But the AI didn't compare you. It never considered you at all.

This is a hypothetical scene, not a specific documented case. But it is becoming a common one, and if you sell online it is worth running the experiment yourself before reading further. Open ChatGPT, Claude, or Perplexity right now and ask the exact comparison question a buyer in your category would ask. See who gets named.

The gut reaction is almost always the same: confusion, then a little bit of anger, then a theory that doesn't hold up. "They must be paying for placement." "Their SEO must be better than mine." "The AI must just be biased toward bigger brands." None of those explanations are how this actually works, and believing them sends you toward fixes that won't move the needle.

The assumption that sends you chasing the wrong fix

Most store owners who watch a competitor get cited instead of them reach for one of two explanations, and both miss what actually happened.

"They must have better SEO." Traditional SEO and AI citation overlap, but only partially. A site with strong domain authority, years of backlinks, and a page-one Google ranking can still lose an AI citation to a smaller competitor, because the AI system is not scoring the whole domain. It's evaluating individual pages against one specific question, and a smaller site with the right page wins that evaluation constantly.

"They must be paying for placement." There is no ad slot inside a ChatGPT, Claude, or Perplexity answer the way there's a sponsored listing at the top of a Google search. These are answer engines, not auction-based results pages. Nobody bought their way into that citation. They earned it with a page the AI system could actually use.

Here's what's actually happening. When someone asks an AI system a comparison or recommendation question, the system doesn't crawl the web and rank every possible store the way a search engine ranks a results page. It searches for content that already answers the question in a form it can extract, verify, and attribute. Then it synthesizes a response built substantially around whichever page did that job best, and cites that page as the source. This is the same retrieval logic behind why entire stores go invisible to AI search, just applied to a single, sharper moment: the exact comparison question a buyer asked.

That means the real question is never "who has the stronger brand" or "who has the bigger site." It's narrower and more answerable: which page, out of everything indexed, most directly and specifically answers this one question? If your competitor published that page and you didn't, the AI system never had a reason to consider you. Not because you lost a comparison. Because you were never in the running.

What the winning page has that yours doesn't

Look at the page your competitor got cited from, and it's rarely their homepage or a product listing. It's almost always one of three formats: a direct comparison ("Brand A vs Brand B"), a buying guide ("Best X for Y use case"), or an FAQ-heavy page that answers the exact question in a labeled block. Each of those formats hands the AI system something a product page cannot: an answer already shaped like the question.

Break down the structural gap:

• Real comparison content. The winning page names the alternatives, states what's different between them, and reaches an actual conclusion. It doesn't just describe one product in isolation. It answers "which one and why," which is the exact shape of the question that was asked.
• Schema markup. Structured data tells the AI system, programmatically, what kind of content is on the page. Product schema, FAQPage schema, and comparison-relevant structured data remove the guesswork. A page with the right schema is a page the AI system can parse with confidence instead of trying to infer intent from unstructured prose.
• FAQ format. FAQ sections are pre-shaped as question and answer pairs, which is precisely the pattern AI retrieval is built to match against a user's question. A page with an FAQ section handed the AI exactly what it needed. A page without one made the AI do extra work it usually won't bother doing when an easier source exists.
• Specific, answerable claims. "Great for camping trips" is not a citable claim. "Draws 15 bar of pressure, runs on 4 AA batteries, and holds a charge for approximately 25 shots between charges" is. AI systems extract and cite specific, checkable statements. Vague marketing language gives them nothing to quote.

Now look at what the losing store typically has instead: a product page with pricing, a photo gallery, and marketing copy that describes the product in general, feel-good terms. No comparison to alternatives. No FAQ section. No schema declaring what the page is. It's a perfectly good page for a shopper who already decided to buy from you. It's an unusable candidate for an AI system trying to answer "which one should I get."

Four AI engines, one shared instinct

ChatGPT, Claude, Perplexity, and Gemini don't retrieve sources identically, and understanding each one's habits matters once you're building content to compete for citations. Perplexity leans hardest on live web retrieval and tends to reward freshness and directly quotable claims. Claude weighs source structure and clarity of reasoning, favoring pages that lay out a comparison logically. ChatGPT search blends its own trained knowledge with live retrieval, and rewards pages that answer a question in the first few sentences instead of building up to it. Gemini pulls heavily from Google's index and structured data, so schema markup carries extra weight there.

We've written a dedicated breakdown for each of these, and if you want the mechanics of a specific engine, read them directly rather than treating this page as the full explanation: how ChatGPT search picks sources, how Claude decides citations, how Perplexity decides citations, how Gemini decides citations, and how Grok decides citations.

What matters here is the shared instinct underneath all four: every one of them is trying to find the single best, most specific, most trustworthy answer to the question it was actually asked. None of them is trying to find the biggest brand, the store with the most traffic, or the one that spent the most on ads. Build the page that answers the specific question, and you become a candidate across all four systems at once. Don't build four separate versions. Build one page that does the job right.

Side by side: the page that got cited vs the page that didn't

Here's the same comparison as a scorecard, factor by factor. If your current store page fails more than one or two of these, that's the reason the citation went elsewhere, not your brand, not your ad budget, and not your Google ranking.

The AI system was never comparing your store to your competitor's store. It was comparing your competitor's page to nothing, because you never gave it a page that answered the question.

How to fix it

The fix isn't a redesign of your homepage or a bigger ad budget. It's building the specific pages that answer the specific comparison and buying questions shoppers in your category actually ask an AI system. You have two paths.

Path 1: Do it yourself

Start by finding the real questions. Ask ChatGPT, Claude, and Perplexity the versus and best-for questions a buyer in your category would ask, and note who gets cited and why. Look at the cited competitor pages and reverse-engineer their structure: what comparison do they make, what specific claims do they state, do they have an FAQ section, what schema is in the page source. Then write your own comparison and buying-guide content for each real question, add schema markup including FAQPage schema, and replace vague marketing language with specific, checkable claims about your product. This works if you have the research bandwidth and the writing discipline to do it question by question, category by category, and to keep it current as competitors update their own pages.

Path 2: Let Ollie do it in 48 hours

Ollie is the AI behind RunOctopus. Tell Ollie what you sell, and it researches the actual comparison and buying-guide questions AI systems get asked in your category, then builds the comparison content, the FAQ sections, and the schema markup grounded in your real catalog, not generic template copy. The structural gap between your page and your competitor's closes in days instead of months of manual research and writing. The strategy is identical either way. The difference is how fast you become a candidate for the next citation instead of watching another one go to someone else.