How AI Chooses Which Brands to Recommend:
Reviews, Mentions and What Studies Show

AI assistants recommend brands that other people already talk about. ChatGPT says it uses product data plus third-party content and summarizes reviews from public review sites. Google says AI Overviews and AI Mode pull from its index and run several related searches per question. Citation studies point the same way: reviews, Reddit, YouTube and editorial lists carry real weight.

Nobody outside OpenAI or Google knows the exact ranking rules. This post separates three things: what the platforms document, what independent studies observed, and what a mid-size Shopify brand can sensibly do about it. For the full picture across feeds, schema and crawlers, start with our AI search optimization guide for Shopify.

What do ChatGPT and Google say they use to pick brands?

Both platforms publish some of their inputs. Neither publishes weights. Here is what the official pages say as of October 2026.

OpenAI also says product results "are not ads" and are not influenced by partnerships. You cannot pay for placement in the organic carousel. The mechanics of the ChatGPT merchant side are in our post on getting products into ChatGPT shopping.

What sources do AI answers actually cite for shopping questions?

The platforms differ, and the studies show it.

BrightEdge analyzed tens of thousands of ecommerce prompts during the 2025 holiday season. Google AI Overviews cited retailers about 4% of the time. ChatGPT cited retailers about 36% of the time. Google leaned on YouTube, Reddit, Quora and editorial sites. ChatGPT leaned on large retailers. Search Engine Land summed it up: "Google leans on what people say, while ChatGPT focuses more on where you can buy it."

Peec counted 30 million cited sources across ChatGPT, AI Mode, Gemini, Perplexity and AI Overviews in the US in March 2026. The top five domains were Reddit, YouTube, LinkedIn, Wikipedia and Forbes. Review sites like G2 and Yelp made the top 10.

Ahrefs studied about 75,000 brands and correlated their traits with AI mentions. YouTube mentions correlated at about 0.74. Branded web mentions correlated at 0.66 in ChatGPT and 0.71 in AI Mode. Backlinks were "very weak." Ahrefs adds the usual caveat: correlation is not causation, and its sample skews to larger sites.

The pattern across all three: AI answers echo what the web says about you, more than what you say about yourself.

Why do AI citation patterns keep changing?

Because the retrieval layer changes without notice. Two examples, both about Reddit in ChatGPT.

Semrush tracked 230,000+ prompts from July to October 2025. Reddit appeared in close to 60% of ChatGPT responses in early August 2025. By mid-September it was around 10%.

A year later, Promptwatch data reported by Semrush showed Reddit holding 3.8% of ChatGPT citations from July 18 to August 7, 2026. From August 14 to 17 it averaged 0.5%, an 86% drop. Over the same window Reddit fell 11% in AI Overviews and 31% in AI Mode. Promptwatch called the figure provisional.

Two lessons for a founder:

• Do not bet on one source. A brand that only seeded Reddit threads lost most of its ChatGPT exposure in a week.
• Measure monthly, not daily. Week-to-week swings are platform noise. Track the trend over a quarter.

Spread mentions across reviews, video, editorial and forums.

Do on-site reviews help you get recommended by AI?

Yes, in three documented ways.

1. ChatGPT summarizes reviews from public review sites. Reviews that sit in crawlable HTML on your product pages, or on a public review platform, can be read. Reviews that load only inside a script widget after the page renders can be harder for crawlers to read.
2. Google's product ratings need 50 reviews. Below that, no stars on Shopping surfaces. Most Shopify review apps can send a feed to Google, or you can upload one in Merchant Center under Products, then Product reviews.
3. Review text answers fan-out questions. "Does it run small?" and "Is it good for sensitive skin?" are side questions AI Mode searches for. Your reviews often answer them in plain words.

What to check in Shopify admin this week:

• Apps: does your review app output Product and AggregateRating structured data, and does it sync to Google Merchant Center? Our Shopify product schema guide shows how to test this with Google's Rich Results Test.
• Request flow: is a review request going out 7 to 14 days after delivery, with a reminder? Our Shopify review strategy guide covers timing and photo reviews.
• Coverage: how many products have zero reviews? A $1.5M skincare brand with 40 SKUs and reviews on only 6 has 34 product pages with nothing third-party for an assistant to quote.

Which off-site reviews and listings matter?

Reviews on your own site are one input. Assistants also look elsewhere.

• Retailer listings. BrightEdge found ChatGPT cites retailers in about 36% of ecommerce answers. If you sell on Amazon, Target or Walmart, those listings and their reviews are part of your AI footprint. Keep titles, specs and images consistent with your Shopify store.
• Public review platforms. Third-party platforms that host verified buyer reviews are crawlable and independent. Google's store ratings pull from review aggregators.
• Google Customer Reviews. It feeds store-level ratings on Google surfaces, per Google's store ratings page.

If your PDP says 3 sizes and your Amazon listing says 4, an assistant gets two conflicting facts. Fix the source of truth, then sync. Our post on product descriptions for AI search covers how to write specs that survive being quoted out of context.

How do Reddit and forum mentions affect AI answers?

Reddit was the most cited domain in Peec's March 2026 study. It is also the most volatile, as the numbers above show. Treat it as a place to listen first.

What works for a mid-size brand:

• Search before you post. Find 10 to 20 threads where people ask about your category. Their questions and complaints are your PDP and FAQ backlog.
• Answer as yourself. A founder or team member replying under a clearly labeled account, with specs and honest tradeoffs, is legitimate. Most subreddits have their own self-promotion rules. Read them first.
• Fix the product issues people raise. A thread saying "the lid leaks" will keep getting cited until the lid stops leaking and newer threads say so.

What gets brands banned and creates legal risk: fake customer accounts, paid posters without disclosure, and employees posing as buyers. The FTC rule covered below treats undisclosed insider reviews as a violation.

How do you get into editorial "best of" lists?

Editorial lists show up in Google's AI answers often. BrightEdge named sites like Wirecutter, CNET and The Spruce Eats among Google's frequent shopping sources.

A realistic approach for a brand doing $50k to $300k a month:

• Pick 15 to 25 target lists. Search your category plus "best" and record which publications show up in AI Overviews and ChatGPT answers.
• Make the reviewer's job easy. Send a one-page spec sheet: materials, dimensions, certifications, price, warranty, return window, and who it is not for.
• Offer samples, not payment for placement. Many publishers run affiliate programs. Joining one is normal. Paying for a positive verdict without disclosure is not.
• Follow up on updates. Lists get refreshed. A new formula or size is a reason to reach out again.

Expect 3 to 6 months from first pitch to inclusion. Some lists never respond.

Does YouTube matter for AI search?

The data says yes. YouTube mentions had the strongest correlation with AI visibility in the Ahrefs study, at about 0.74. YouTube was the second most cited domain in Peec's data and a top Google source in BrightEdge's holiday analysis.

For a Shopify brand, that points to two moves:

1. Creator reviews. Send product to 10 to 30 creators who review your category. Unboxings, comparisons and "6 months later" videos produce the kind of detail assistants quote. The FTC says free product is a material connection, and for video "the disclosure should be in the video and not just in the description" (FTC Disclosures 101).
2. Your own channel. Short how-to, sizing and comparison videos with accurate titles and descriptions. A clear title like "Size 8 vs size 9: fit test on 3 foot shapes" beats a brand film.

Our D2C UGC strategy guide covers creator briefs and rights.

What is a realistic 90 day plan for a mid-size brand?

Here is a plan for a fictional $3M a year supplement brand with 25 SKUs and a 2 person marketing team.

Re-run the prompt set monthly and look for a 3 month trend. Our guide to tracking ChatGPT traffic in GA4 shows how to measure the traffic side.

What should you never do to get recommended by AI?

Fake or bought reviews. The FTC's Consumer Reviews and Testimonials Rule took effect October 21, 2024. It bans fake reviews (including AI-generated ones), incentives conditioned on positive sentiment, undisclosed insider reviews, company-controlled "independent" review sites, suppressing reviews through threats, and buying fake followers or views. In December 2025 the FTC sent warning letters to 10 companies and noted penalties of up to $53,088 per violation.

Review gating. The rule has no specific ban on asking only happy customers. FTC staff warn it can still violate the FTC Act. Ask every buyer.

Hidden text and prompt stuffing. Text hidden from shoppers but written for bots falls under Google's "hidden text and link abuse" spam policy. Instructions aimed at AI models buried in your page are the same idea with a new name.

Paid "AI SEO" guarantees. Nobody outside OpenAI or Google controls these rankings. A vendor promising a top spot in ChatGPT is selling something they cannot deliver.

Where should you start?

Start with reviews. Every platform reads them in some form. Get hero SKUs past 50 reviews, keep the text crawlable and sync it to Merchant Center. Then add creators and editorial outreach. If you want a second opinion on where your store stands, our free Shopify store grader is a quick start, or you can book a call with our team.