Do affiliate review sites decide which brands AI recommends?

For many "best X for Y" questions, engines that search the web pull in roundup and review pages and cite them. If those pages name you, you have a route into the answer. If they don't, your own site is often not what gets retrieved.

That is an association, not a lever with a known size. Nobody outside the engines can see which pages were retrieved for a real user's question. Engines also rewrite the question before searching: OpenAI says ChatGPT search typically rewrites your query into one or more targeted queries that it sends to its search providers.

So affiliate sites matter where an engine retrieves them, and only there. The practical question is not "do review sites decide" but "which pages does each engine cite for my buyers' questions, and what do they say about me." For the trade-off between the two channels, see your own site vs third-party mentions.

An AI answer passes through four stages. Third-party pages only matter in some of them.

Retrieval is discretionary per query. As of October 2026, ChatGPT decides for itself when to search, and web search in Claude.ai is a user toggle. An answer given without search never touches an affiliate page at all.

What this means: a strong roundup placement helps only on answers where search fires and that roundup is retrieved. It does nothing for answers drawn from the model's training data.

Several findings point the same way, and none of them proves cause.

• Citations concentrate. A study of AI search systems (arXiv 2507.05301) found the top 20 most frequent news sources accounted for 67.3% of all citations for OpenAI models. That data covers news queries from March to May 2025, not product recommendations, but it shows engines lean on a small set of familiar sources.
• Ranking is a weak guide. Ahrefs found in March 2026 that 37.9% of URLs cited in AI Overviews also appeared in the first 10 results. Most cited pages sat lower or beyond.
• Brand size confounds everything. Large brands appear on more lists and in more answers. No published study separates the list placement from the brand's fame.
• Engines differ. In Otterly's experiment, Microsoft Copilot kept citing two sites after Google deindexed them, while Google's AI features almost never did.

One widely shared study on review profiles never defines its citation metric, so treat its numbers as unusable. The same caution applies to whether online reviews affect AI recommendations.

You can't see retrieval, but you can see citations. Here is the method.

1. Write the questions buyers actually ask, in their words. Say you sell payroll software: "best payroll tool for a 20-person agency" is one.
2. Ask each engine with web search on, logged out, with memory off. Logged-in answers can draw on saved context and won't repeat.
3. Record every cited URL, and whether you were named.
4. Repeat over several days. Answers vary between runs.
5. Group the URLs by type: affiliate roundups, review platforms, forums, publishers.
6. For each page, note whether it names you and whether its facts about you are right.

The playbook for finding the pages AI engines cite covers this in more detail. For a quick snapshot, the free checker on this site shows whether engines name you for up to three buyer questions. Once you have the list, work through getting listed on pages AI engines already cite, starting with the pages that appear across several engines.

Affiliate commissions and sponsored listings are normal business. The risk is what the placement says.

A page that states a specific fact gives an engine a sentence it can quote. That cuts both ways. A placement with a wrong price, an invented feature or a made-up rating hands the engine a wrong sentence about you. AI platforms already get specific facts wrong about specific brands, and third-party pages are one source of those errors.

Avoid three things:

• Placements that misstate your product. Check the copy before you pay, and after it goes live.
• Mass-produced roundup networks. Google's spam policies treat many pages generated to manipulate rankings without value as scaled content abuse. In Otterly's test, Google removed both mass-produced AI sites with no manual action warning.
• Offers to hide instructions for AI in a page. Injected content is illegitimate. Research on LLM search agents (arXiv 2606.16821) found attack success varied by model, and results shift with every safety update. It is a reputational risk, not a tactic.

If a cited page already gets you wrong, ask the site to fix the wrong facts. If you suspect a page mentioning you carries hidden instructions, scan third-party pages for injected instructions.

Should I pay to be included on an affiliate best-of list? Only if the page is already cited for your buyers' questions and will describe you accurately. Paying for a list no engine retrieves buys nothing for AI answers. Paying for a list that misstates your price or features creates a wrong fact engines can repeat. Check citations first, read the draft copy, and monitor the page after it publishes. More on this in [whether best-of lists help AI recommend you](/questions/best-of-lists-ai-recommendations).

Does my own website still matter if engines cite review sites? Yes. Review sites often describe you by reading your site, and engines that do fetch your pages need to reach them. Check that your robots.txt allows the retrieval bots, such as OAI-SearchBot for ChatGPT search, and that your firewall doesn't block them. Put key facts like pricing in the HTML before JavaScript runs, because most AI crawlers don't execute JavaScript.

Why does AI recommend a competitor that appears on fewer lists? List count isn't the input. The engine retrieves a few pages per rewritten query, and a competitor may sit on the one page that gets retrieved. Brand prominence also plays a part, and answers without search come from training data. Compare the cited URLs for the exact question rather than counting lists.

Can I see which pages an engine read but didn't cite? Not in the consumer apps. Some developer APIs, including OpenAI's web search tool, return the full set of URLs consulted separately from the citations, and the consulted set is usually larger. That shows retrieval for API queries you run yourself. It is a useful view, but it isn't identical to what the consumer app does for a real user.