Watch an AI assistant answer a question in your category and you have probably seen it happen. It quotes a Reddit thread, a review platform and a comparison article, names a couple of brands, and never cites your website, even though your site says all of it more clearly. It looks arbitrary. It is not.
Why AI answers pull third-party reviews and skip your official site
Every engine that answers a recommendation question has the same problem to solve first: who can it trust? A brand describing its own product is the least reliable witness available. Every company says it is the best, the most trusted, the leader in its field. None of that is checkable, so the engines discount it. An independent review, a Reddit thread, or a comparison written by someone with no stake in the sale carries a signal your homepage cannot, because it comes from a party with no reason to flatter you.
That is why the reviews look random when you read the answer. They are not random to the engine. It reached past your site on purpose, for a source it could treat as evidence rather than marketing. Our analysis of AI citations, covered in why brands don't show up in AI answers, found that the large majority of citations in AI answers come from third-party sources rather than brand sites, with Reddit and review platforms carrying much of the weight.
That verification incentive is the whole reason. An engine that quoted vendors about themselves would be trivial to game and quickly useless, so leaning on independent sources is the design every major model has settled on.
What "ignoring your official site" actually means
It is easy to read a third-party citation as the engine refusing to look at your website. That is not what is happening. Official sites are read constantly. They are simply read for a different job.
Your site is where an engine confirms the facts. Third parties are where it forms the judgement. When ChatGPT works through a question it will run narrow checks straight against your pages to verify a price, a specification or a credential before it commits to an answer. What it will not do is take your word for whether you are the best option in the category. That verdict comes from the sources it treats as impartial.
The practical consequence is a split brief. On your own pages, state the checkable facts plainly and mark them up with structured data, so the engine can confirm them. Off your pages, build the independent presence that earns the mention. A site that buries its prices in a PDF fails the first job. A brand absent from every review platform fails the second, and no amount of homepage polish fixes it.
Why Google AI Mode cites some brands and products but not others
Google AI Mode looks the least predictable of the engines, and the reason sits in how it searches. Rather than run your question once, it breaks the question into a set of smaller ones and runs them together, a method Google describes as query fan-out (Google). Ask it for the best option for a specific need and it fires off separate searches for the parts, the features, the price, the use case, the alternatives, then merges what comes back into a single answer.
That mechanism explains the apparent randomness. A brand gets cited when it shows up clearly across the third-party pages those sub-queries return, and gets skipped when it is thin or absent from them. Two competitors can look almost identical on their own websites and get very different treatment, because one is well covered on the sources the sub-queries surface and the other is not.
It also explains why a strong Google ranking does not carry over. AI Mode is pulling passages that answer its sub-questions, wherever they sit, so a page that ranks first for the headline query can lose its place to pages that answer the smaller questions better. Winning here is about being present across the sources the fan-out reaches, not about holding one ranking position.
How the citation pool differs from engine to engine
No two engines draw from the same pool, which is why a brand can appear in one and vanish in another. The models cite at different rates and against different bars of evidence. Gemini names brands readily. ChatGPT cites a moderate share of the time. Claude is the most conservative and wants the strongest independent backing before it will name you at all. Google AI Overviews rarely recommends a brand outright. The pattern holds across categories: the thinner your third-party footprint, the more engines drop you, starting with the strict ones.
So build for the strictest bar rather than chase each engine separately. Satisfy a conservative model like Claude and the more generous engines tend to follow, because the evidence that convinces the cautious one is what lifts you everywhere: independent sources, in volume, describing you clearly and recently.
How to build a prioritised target list from what AI already cites
Most outreach lists start from a generic media database and a wish list of logos. The better starting point is the set of sources the engines already cite in your category, because those are the pages the models keep re-reading when they answer.
The method is simple enough to run by hand:
- Ask the real buyer questions. Put the questions your customers ask, best providers, who should we use, top options for a given need, to ChatGPT, Claude, Gemini, Perplexity and Google AI Mode. Repeat each a few times, because the answers vary between runs.
- Record every source. Note every domain each engine cites, not just the ones that mention you. That set is your category citation pool.
- Count and sort. Rank the domains by how often they appear across engines and phrasings. The ones cited most often are the ones worth the most effort.
- Mark the gaps. Against each domain, flag whether it already names you, names competitors but not you, or names nobody in your category yet. The domains citing competitors and not you are your first targets, because the engines already trust them for the topic and a mention there is worth more than a mention on a site nobody cites.
From there the work is earning a presence on those specific pages, through reviews, genuine community contribution, and inclusion in the comparison and best-of articles the engines lean on. We set out the full method, including the four source types that matter most and realistic timelines, in our AI citation strategy guide.
The quickest way to start is to see who gets cited in your category today. A free AI visibility check runs the questions across the major engines and shows which brands, including your competitors, get named, and the sources the answers lean on. That is the first draft of your target list.
Frequently asked questions
Why do AI answers pull random third-party reviews and ignore our official site?
They are not random. AI engines weight independent sources above self-published content because a vendor describing its own product is not evidence, while a review site or forum describing it is. When an engine answers a recommendation question it reaches for the sources it can treat as impartial: reviews, directories, forums and publishers. Your official site is still read, but for a judgement about who is best the engine leans on third parties that have no reason to flatter you.
Why does Google AI Mode cite some brands and products but not others?
Google AI Mode breaks one question into many smaller ones and runs them at once, a method Google calls query fan-out. It then assembles an answer from whatever each sub-query returns. A brand gets cited when it is present and clearly described across the third-party pages those sub-queries surface, and skipped when it is thin or absent from them. Ranking first in classic Google does not carry over, because AI Mode is selecting passages that answer sub-questions, not the ten blue links.
How can I build a prioritised list of publications and review sites to target for outreach based on which ones AI models cite?
Start from the sources the engines already cite in your category, not a generic media list. Ask ChatGPT, Claude, Gemini, Perplexity and Google AI Mode the questions your buyers ask, record every domain each one cites, and count how often each domain appears. Sort by frequency, then mark which already name you, which name competitors but not you, and which name nobody in your category yet. The domains that cite competitors but not you are the highest priority, because the engines already trust them for the topic.
Does an AI engine ignoring your official site mean it never reads your pages?
No. Official sites are still read, but mostly for facts the engine wants to confirm, such as prices, specifications and credentials. The judgement about whether you belong in the answer tends to come from independent sources. So your own pages need the checkable facts stated plainly, while your presence in reviews, forums and publishers is what earns the mention.