~/neuralshit
← all posts

how does claude decide which brands to recommend

claudegeoai-recommendationsai-seo
A dim supermarket shelf where a single product is lit by a hard spotlight.

Claude has no ad product — you can’t pay to be recommended. It surfaces brands the way a well-read friend would: by the consensus of independent, credible mentions across the web it was trained on and retrieves from. The brand discussed consistently and positively by many third-party sources wins — not the one with the slickest landing page.

can you pay claude to recommend your brand?

No. Anthropic has publicly pledged that Claude stays ad-free — no sponsored slots, no advertiser influence, no product placement you didn’t ask for. The stated logic is a conflict of interest: the moment an advertiser can pay for a mention, the recommendation stops being about you and starts being about the highest bidder. Anthropic’s money comes from API and enterprise seats, not ad inventory, and enterprise buyers wouldn’t touch a model that sells its answers. So there’s no “buy your way in” lever here. Every recommendation is organic — which means the only way in is to earn it.

what actually makes claude recommend a brand?

Consensus. Claude’s picture of your brand is assembled from everything credible that’s been written about it — across its training data, and when web search is on, live retrieval from authoritative sources. It is not reading your homepage and taking your word for it. It is reading what the rest of the internet says about you when you’re not in the room. A brand mentioned consistently and positively across many independent, trustworthy places becomes the default answer. One polished page you fully control counts for almost nothing against that weight of outside evidence.

how is claude different from chatgpt or perplexity here?

Two differences that change the game. First, Claude leans balanced and multi-brand — ask it to compare and it tends to hand you a shortlist with tradeoffs, not one hero pick, so “become the single answer” is a weaker play here than on other engines. Second, unlike Perplexity or ChatGPT, Claude doesn’t surface citations in a way you can cleanly reverse-engineer and optimize into. There’s no obvious citation slot to farm. Both facts push the same direction: shallow tricks matter less, and broad, real reputation matters more.

The same way you’d earn a recommendation from a sharp human: be verifiably good, and get other people to say so.

  • Earn independent mentions. Digital PR, guest articles, being covered by real sources in your niche. Third-party validation is the currency — it’s what the model reads as “this is legit.”
  • Reviews and consistent presence across the channels your category lives on. Claude reads the aggregate, not your ad copy.
  • Depth and clarity in your own content, so when it does get read, it’s extractable and unambiguous rather than marketing fog.
  • Make sure the AI can read you at all. If retrieval bots can’t fetch your pages, you’re invisible before the reputation game even starts — here’s how to make your site visible to AI crawlers.

the honest takeaway

You can’t hack Claude — you can only deserve it. There’s no ad auction, no citation slot, no llms.txt incantation that makes Claude vouch for a brand the web doesn’t already vouch for. The optimization is the reputation. Build the thing that people independently recommend, and Claude will too — because that is, quite literally, all it’s doing: repeating the consensus back to whoever asked.