If you land in a new city ww with no booking, hop into a cab and ask the driver ask the driver where to stay, chances are he’ll know “the best hotel in town” without even blinking.

Now swap the cab driver for ChatGPT.

Shoppers ask AI for the best cordless vacuum the same way tourists ask cabbies about dinner, and they trust the answer just as much.  

But a recent study traced 106,758 sources behind “best X” answers on ChatGPT, Google’s AI Overview and Perplexity and found that one in four of those sources disclose a commercial interest on the page. Most of the pages feeding AI recommendations are affiliate review page, that is, pages where a publisher tests products and earns a commission when you buy through its link.

And many cases are less subtle than you’d think. I searched “best mattress” on Google, and its AI Overview recommended Sleep Foundation as the primary source. Upon opening it, the page displayed the label “When you buy through our links, we may earn a commission. Products or services may be offered by an affiliated entity.”

To be fair, the publishers aren’t hiding anything. The label is right there at the top of their page.    

‍

That would make anyone think that AI product recommendations are biased. But are they? If so, how can you use them to get featured?

Let’s look at which publishers are shaping AI product recommendations in your category, so you know exactly whose attention to earn, what value to offer, and how to get your products right into the AI’s sight.

In this guide, we’ll explore all that and more.  

But before that...

Let’s look at what the study found  

When a shopper asks ChatGPT for the ‘best treadmill’, the AI runs a few quick searches, reads a bunch of pages, and serves up a neat little list of recommendations. It’s pretty much like a cabbie who has never eaten at a single restaurant but knows which are the best ones in town based on the flyers he read.

And these tools read a lot of flyers. Counting only sources the study could match to a page, nearly nine answers in ten (86%) use at least one commercial source, and ChatGPT does it most often (91%).  

But not every commercial source looks the same, so here’s the difference. A commercial source is any page that discloses it has money riding on your purchase. That includes sponsored posts, paid-program pages and “in partnership with” content. The biggest group by far is affiliate reviews, where the publisher earns a commission when you buy through its link. Together, these make up one in four of the sources AI tools cite for “best X” questions, and most of them are affiliate pages.

If we look at a single citation instead, that order flips. Perplexity has the highest commercial share per citation at 31.1%, against 22.5% for ChatGPT. Perplexity leans on commercial pages more than any other tool. Almost a third of the sources it cites (31.1%) have a commercial interest, compared with 22.5% for ChatGPT. And in one Perplexity answer out of every six, commercial pages make up most of the sources behind the recommendation.

These so-called ‘flyers’ also tend to come from a few well-known names. The study found ten publishers, review brands such as TechRadar and PCMag and commerce sections such as Forbes Advisor, account for a quarter of all commercially disclosed citations.

These so-called ‘flyers’ also tend to come from a few well-known names. The study found ten publishers, review brands such as TechRadar and PCMag and commerce sections such as Forbes Advisor, account for a quarter of all commercially disclosed citations.  

I know what you are wondering now; “why do these publishers with affiliated links get cited so often”. Well, we’ve got our next section dedicated to just that.

Why does AI lean on affiliate sites?

Nobody at OpenAI or Google decided that affiliate reviewers should run the shelf or at least they didn’t see the downside. It mostly comes down to who writes the pages the AI wants to read. Take away the hobbyists and the newsrooms, and most content on the internet gets made because someone gets paid for it. Testing 20 pairs of headphones, measuring their noise cancellation and writing 4,000 words on the results takes weeks of work, and almost nobody does that for the love of it. An affiliate publisher does, because every reader who clicks “buy” pays a little toward the next review. Poets would never have been poor if their poems came with a “buy now” link.

• Affiliate sites answer the exact question - A shopper asks for the “best noise-cancelling headphones,” and affiliate publishers have pages with nearly that exact title. Brands rarely publish “best headphones” lists, because nobody trusts a brand that ranks itself first. Though brands indeed have tried and failed.

• They’re built to rank in search - AI tools run a web search behind the scenes before writing an answer. Affiliate publishers have spent years optimizing for those same searches, so they’re often what comes back.

• They’re easy to quote - Roundups come with scores, specs, prices and “best for” labels. One research found that pages with statistics and citations get noticeably more visibility in AI answers, and a review page is stuffed with exactly that.  

• They stay fresh - Affiliate roundups get updated constantly, often with the year in the title. When a shopper types “in 2026,” those pages look like the most current answer.

• They look like a second opinion - AI tools tend to trust third-party sources over brand pages when answering “best” questions, because even a robot knows the restaurant isn’t going to tell you the place next door has better shawarma. The commission doesn’t show up in the source chip, so the AI and the shopper both see an outside reviewer.

So, how exposed would that make your industry?

The share of commercial sources varies far more by what you ask than by which tool you ask. Some categories are practically run by affiliate reviewers, while others barely see them. Here’s a table to help you get a better idea:

Source: GetCited.

A thing to note here is that a low number doesn’t mean a clean shelf. In these industries, AI often cites company websites and business directories, where commercial interests may not be clearly disclosed. That makes them harder for the study to identify, so some commercial influence could be going undetected.

But how do you detect it?  

How to detect who’s driving your customers

This is what I call the Source Chip Audit technique. All you have to do is follow these simple steps to find out who has been bringing you or can bring you your customers.

1. Write 10 to 15 “best X” questions your buyers would type, such as “best standing desk under $500” or “which rowing machine should I buy?”

2. Run each one on ChatGPT, Google and Perplexity. Use a temporary chat or a signed-out, incognito window, so your own history doesn’t steer the answer. This is to be done for each query.

3. Log every cited URL in a spreadsheet, along with the products the AI named.

4. Open each page and look for a disclosure. Check the top of the page, the byline and the footer for lines like “we may earn a commission,” “in partnership with,” “sponsored” or “paid program.”

5. Tally what you find. Most categories fall into one of three shelf types. Affiliate-run shelves are dominated by review publishers earning commission. Review-platform-run shelves lean on Trustpilot, G2, Yelp and Reddit threads. Provider-run shelves cite brand sites, retailers and directories.

6. Repeat every quarter, since AI answers shift between runs and months.

What I found when I ran it

I started with one of the most common questions in consumer electronics: “Best noise-cancelling headphones in 2026.” I put it to all three tools on the same day, with a fresh chat or a signed-out, incognito window for each, then opened every cited page to look for a disclosure.

Three things stood out for me.

1. Every answer used at least one affiliate source - That’s three out of three, which lines up with the study’s finding that 86% of answers use at least one commercial source.  

2. Headphones ran hotter than the category average - Just over half of my sources carried affiliate links, against 41.3% for consumer electronics overall in the study. One query proves nothing on its own, but it fits the study’s point that individual products can run much higher than their sector.  

3. The same sites kept showing up - RTINGS appeared on all three tools, and NYT Wirecutter showed up on both Perplexity and Google. Only one brand made ChatGPT’s list (Bose), and only two retailer or brand pages made Google’s (Sony and Sennheiser). If you sell headphones, those few publishers are the ones you need to know.

A note on method: this is a single query run once, so treat it as a snapshot rather than a benchmark. AI answers change between runs, and your category may look different. That’s exactly why it’s worth running the audit yourself.

What to do based on what you find

That’s a very valid question to ask after all that work. Here’s what you can do:

• If your shelf is affiliate-run, the cabbies are the review publishers, and they’re worth getting to know. Send review units to the sites that show up most in your audit, and make sure your affiliate program pays them fairly and on time. If you run one, keep an eye on affiliate fraud so you’re paying the reviewers who actually send buyers.

• Then give reviewers what they won’t know by just testing your product. This includes your customer service, the warranty length, what’s covered, how a claim works and how fast repairs come back. It gives a reviewer a reason to write about you instead of the next brand on the list.

• If your shelf is review-platform-run, the AI is reading what your customers say about you. Claim your profiles, ask every buyer for a review and reply to the bad ones like a polite friend.  

• If your shelf is provider-run, then congratulations.  

What not to do

Also a valid thing to ask. Here’s the rundown:

• Don’t build your own “independent” review site – Whatever you do, just don’t do this. A comparison site that ranks your products first is illegal in the US. The FTC’s 2024 rule prohibits a business from misrepresenting that a website or entity it controls provides independent reviews or opinions about a category of products that includes its own.  

• Don’t hide text for the bots - Researchers showed that adding a carefully crafted “strategic text sequence” to a product page can significantly increase its chances of becoming the AI’s top recommendation. Though AI companies are already working to filter this kind of manipulation. A trick that gets you a top spot this quarter can get your pages penalized in the next.

• Think twice before renting a paid-program page on a big news site - It’s legal when labeled, and as the intro showed, AI tools do cite them. But Google has already punished the rented-authority model in regular search. Here’s the proof: after Google’s site reputation abuse update, Sistrix reported Time’s affiliate sections, lost 97% of their visibility and Forbes Advisor 43%.  

To sum it up,  

Your customers will keep asking ChatGPT, Google and Perplexity what to buy, and those tools will keep reading the same handful of reviewers before they answer. So stop guessing who’s recommending your products and go find out. Run the audit, learn which publishers drive your category, and give them something worth writing about.