Does AI Recommend Amazon Over DTC Brands?

In 391,073 citations collected from AI shopping answers between April and July 2026, ChatGPT linked to amazon.com six times. That finding, from a study by LLM Pulse, cuts against the common belief that AI assistants hand every shopping question to Amazon. The same study found most of Amazon's citations came from Google's AI Mode and AI Overviews.

The belief is not wrong everywhere, and it is not right everywhere either. Which retailer an assistant favors depends on the engine, the category and who lets the engine's crawler in. This brief sorts out what the 2026 data shows and what a DTC brand can do about the part that hurts it.

Why do AI assistants recommend Amazon and big retailers instead of DTC brands?

AI assistants favor Amazon and big retailers mainly in Google's AI Overviews and AI Mode, and in Microsoft Copilot for at least beauty queries, while ChatGPT and Perplexity in 2026 cite Amazon rarely and lean on review sites, YouTube, Reddit and retailers such as Walmart and Best Buy.

The picture changed over a few months. In holiday 2025, BrightEdge reported that ChatGPT cited retailers about 36% of the time, with Amazon as its most cited shopping source.

By early 2026, Tinuiti's data put Amazon at about 0.3% of ChatGPT citations for commercial prompts, down from roughly 0.5%, while Amazon held around 3% in AI Mode and AI Overviews.

The likely reason is access. Modern Retail reported in November 2025 that Amazon had blocked more of OpenAI's crawlers in its robots.txt file, including the ones ChatGPT uses to fetch pages for answers. Neither company commented, and the studies use different prompt sets, so treat the link as probable rather than proven.

Copilot is the clearest exception. In April 2026 data from Tinuiti, reported by MediaPost, Amazon held 21% of citation share in Copilot for beauty queries.

AI names DTC brands but cites someone else

The bigger problem for a DTC brand is not Amazon. It is that an assistant can recommend the brand by name while linking to other sites as proof.

Gradial's study of 28 retail and consumer brands found an average mention rate of 44% against a citation rate of 8% across OpenAI, Anthropic and Perplexity. Review sites, editorial publications, data aggregators and Amazon took the citation slots, while product and category pages were rarely cited.

BrightEdge saw the same split in 12 weeks of ecommerce answers, writing that the engines converge on which brands belong in an answer and diverge on what they cite in support. The brand gets named, and the click goes elsewhere. For a shopper, the cited review site or retailer becomes the next step, and a marketplace with stock and fast shipping is well placed to win the sale there.

What makes big retailers easier to cite

The research points to signals that work as correlations rather than proven causes.

Review volume matters. In Profound's analysis of 812,190 ChatGPT product cards from June 2026, the top-ranked cards had a median of 787 reviews against 352 for cards ranked fourth or lower. A large retailer collects reviews for many brands on one page, which a single DTC product page rarely matches.

Product data matters as well. The same Profound analysis found that 87.3% of product cards came from crawling the web rather than from merchant feeds, so what a store publishes on its pages shaped what ChatGPT showed. Profound later reported that from 10 July 2026 most ChatGPT Shopping picks came from merchant feeds, so a complete catalog feed now matters as much as the pages. Shopify said on its Q2 2026 call that AI searches using its Catalog data converted at twice the rate of those using scraped data, a company claim with no base disclosed.

Access also counts. Amazon's block shows that being cited depends on letting the engine read the page. A DTC brand controls its own robots.txt and feeds, which a marketplace seller does not.

The weighting also moves. BrightEdge found YouTube's share of AI Overviews citations for ecommerce prompts went from 31% to 65% between May and July 2026, with nothing changed on the sites themselves. A single snapshot of who gets cited is out of date within weeks.

What a DTC brand can do about it

A DTC brand can start with the cheapest fixes.

1. Check crawler access. Confirm the store's robots.txt does not block OAI-SearchBot, ChatGPT-User or PerplexityBot by accident, and that product pages load without scripts hiding prices or stock.
2. Clean the product data. Complete titles, attributes, variants, price and stock in Shopify and Merchant Center, since the assistants read what the store publishes. Getting a DTC brand ready for AI shopping agents covers the full list.
3. Build reviews where they count. Grow review volume on hero products, on the brand's own pages and on the third-party sites that appear in your category's answers.
4. Earn the citations that name you. Creator videos, editorial roundups and comparison pages are the sources assistants link to when they recommend a brand.
5. Track it by engine. Run the same category questions monthly in ChatGPT, Perplexity, AI Mode and Copilot and log which domains are cited, because the answer differs by engine. How AI shopping assistants are changing product discovery explains why the product page still decides the sale.

Frequently asked questions

Should a DTC brand sell on Amazon to show up in AI answers?

Selling on Amazon helps most in Google's AI Overviews and AI Mode, and in Copilot for beauty queries, where Amazon is cited more often, and helps little in ChatGPT in 2026. Weigh it as a channel decision on margin and control first, since the AI benefit depends on which engines your customers use.

Can a brand pay to be cited in AI shopping answers?

No platform publishes how it chooses cited sources in these answers, and the studies in this brief measured organic answers. Check each platform's current merchant and ads terms before planning around paid placement.

How often should we check which sites AI cites in our category?

Monthly is a practical minimum. Citation patterns shifted by more than 30 percentage points in ten weeks in one study, so a quarterly check can miss a change that matters.