ChatGPT Has Started Showing Ads: Separating Ads from Recommendations Based on 100 E-Commerce Related Queries

Key Takeaways

• Starting in 2026, ads began appearing at the bottom of ChatGPT answers. The rollout began in the U.S., and in May, a self-service ad platform opened, allowing anyone to run ads directly.

• When we measured 100 U.S. e-commerce queries using our own GEO engine, ads appeared on 49 of them. However, among those 49 ad placements, the advertised brand also appeared in ChatGPT’s recommendation list in only two cases.

• The pattern remained similar even when users asked about a specific brand. When we asked for “La Mer alternatives,” an ad for a men’s shaving brand appeared. When we asked where to buy the Dyson V15, an ad for an affiliate review site appeared.

• Ads are paid exposure. Recommendations are earned through content and reputation. The fact that these two barely overlap is exactly what brands need to pay attention to now.

1. ChatGPT Has Started Showing Ads in Answers

For a long time, one of ChatGPT’s strengths was that it provided “ad-free answers.” That premise changed in 2026. OpenAI officially announced the introduction of ads in January, and starting February 9, ads began appearing for users in the U.S. on the Free and Go plans. On May 5, OpenAI added a self-service ad platform and cost-per-click pricing, allowing anyone to buy ads without a minimum spend. (OpenAI official announcement)

Ads appear as separately labeled cards at the bottom of answers, and according to OpenAI, they do not influence the answer itself. That may be the stated policy, but to understand which brands appear as ads for which questions, we needed to look directly. So we measured it.

2. We Directly Measured 100 U.S. E-Commerce Queries

Using our own GEO engine, we submitted 100 e-commerce purchase queries to ChatGPT in the U.S. and collected both the ad cards shown alongside the answers and the brands ChatGPT recommended in the body of the response.

The queries were divided into five categories, with 20 queries in each: electronics and audio, appliances, beauty and skincare, home and living, and fashion and apparel. The questions were designed to have clear purchase intent, such as “best ~ under $X.” We also included some brand-specific questions, such as “where to buy Dyson V15,” to examine whether competitor conquesting was happening.

The data was collected on May 27, 2026.

One thing should be noted upfront. This is a U.S. snapshot from a specific point in time. ChatGPT ads may be affected by personalization and real-time bidding, so the results can vary depending on the user and the timing, even for the same query. Therefore, the numbers below should not be read as permanent ratios, but as a snapshot of what was happening in the U.S. at the end of May 2026.

3. Advertised Brands and ChatGPT-Recommended Brands Barely Overlapped

49% — Queries with ads: Ads appeared on 49 out of 100 e-commerce queries.

96% — Ads ≠ recommendations: In 47 out of 49 ad placements, the advertised brand did not appear in the recommendation list.

4/5 — Brand conquesting: Among five brand-specific queries where ads appeared, four showed ads for a different brand.

Out of the 100 queries, 49 had ads attached. But when we compared those 49 ads against the brands ChatGPT recommended in the main answer, the advertised brand appeared in the recommendation list in only two cases.

Ad exposure by category was as follows.

Below are representative examples comparing the ad card with ChatGPT’s recommendations.

Users who see an ad may easily assume that “ChatGPT is recommending this.” But in most cases, the brands ChatGPT actually recommended in the answer were different.

4. How ChatGPT Ads Are Changing Advertising

Across the 100 queries, several patterns emerged. ChatGPT ads resemble traditional search ads in some ways, but they also behave differently in several important respects.

Ads Are Not the “Answer”

Among the 49 queries that had ads, the advertised brand also appeared in ChatGPT’s recommendation list in only two cases. In the remaining 47 cases, or 96%, the brands shown in ads and the brands recommended in the answer were different.

ChatGPT recommends products in the main answer based on budget and use case, while ads appear separately in their own space. The recommendation that users trust and the placement advertisers pay for are not the same thing.

Ads Can Be Misaligned with the Query

For the query “La Mer alternatives,” which asks about luxury skincare, an ad for MANSCAPED, a men’s shaving brand, appeared. MANSCAPED was the most frequently appearing advertiser in the entire measurement, and it even appeared on questions far from its own category, such as skincare.

This suggests that when advertisers bid aggressively, ads can appear in placements that are not closely aligned with the user’s context.

Competitor Conquesting Is Common

Among five brand-specific queries where ads appeared, four showed ads for a different brand. For “where to buy Dyson V15,” an affiliate review site appeared. For “Casper mattress sale,” a department store ad appeared.

The kind of competitor keyword conquesting that has long been common in search advertising is now also appearing in ChatGPT.

Retailers and Affiliate Sites Are Entering as Advertisers, Not Just Manufacturers

Many of the frequent advertisers were not product manufacturers, but retailers that sell multiple brands, such as DSW, Home Depot, Macy’s, and Wayfair, as well as affiliate review sites such as Consumer Picks.

Even for robot vacuum queries, the advertiser was not a vacuum manufacturer but an affiliate review site promoting an “expert-ranked Top 10” list.

Ad Frequency Varies by Category

Ads appeared on 49 out of 100 queries, or 49% overall, but there were differences by category. Fashion and apparel had the highest exposure rate at 60%, while appliances had the lowest at 40%.

This appears to be an early-stage pattern in which ads are concentrated in categories where advertisers have already entered.

There were also signs in the links indicating that these were indeed ChatGPT ads. The click URLs contained campaign tags pointing to OpenAI’s ad pilot. One beauty brand’s link contained a marker suggesting a quarterly pilot, while a large retailer’s link included a tag indicating traffic routed through ChatGPT.

5. Where Did ChatGPT’s Recommended Brands Come From?

If not from ads, then where did the brands recommended in ChatGPT’s answers come from? The sources cited in the answers provide a clue.

Running shoe and walking shoe recommendations cited Reddit posts, RunRepeat, and media outlets such as Forbes.

Robot vacuum recommendations referenced RTINGS, Tom’s Guide, and Forbes.

Vitamin C serum recommendations cited beauty media such as Allure and NewBeauty, as well as brand-owned websites.

For Dyson, ChatGPT used Dyson’s own site, Reddit, and Wired as sources.

The brands that appeared in recommendations entered ChatGPT’s answers through community reputation, expert media reviews, and well-structured brand-owned content. If ads are paid exposure, recommendations are earned through content and reputation.

This difference matters because of durability. Ads disappear when the budget stops. Recommendation positions, however, are assets built through accumulated content and reputation. The work of organizing content and data so that AI cites and recommends your brand is the core target of Generative Engine Optimization, or GEO.

6. What Brands Should Do Now

Ads and recommendations are not an either-or choice. It is perfectly reasonable to consider the newly opened ChatGPT ad channel. But before doing so, brands need to be clear about one thing: buying ads does not mean ChatGPT will place your brand in its recommended answer.

As our measurement of 100 queries shows, recommendations are a separate space, and that space is earned through content and reputation.

ChatGPT ads have not yet appeared in Korea. Put differently, this is still a window of opportunity to secure organic recommendation positions before ad competition begins. The first thing brands should do is check which brands ChatGPT currently recommends for questions in their category, then organize their content and data so that their own brand can be included in those recommendations.

Villion is currently collecting data and running experiments to respond to ChatGPT Ads.

Frequently Asked Questions

Can ChatGPT ads be seen in Korea?

As of May 2026, they began in the U.S. first and have not yet appeared in Korea. In other words, this is still a period when brands can secure organic recommendation positions before ad competition begins.

If I buy ChatGPT ads, will my brand be recommended?

No. Ad cards and ChatGPT’s recommendations in the main answer are separate. When we measured 100 U.S. e-commerce queries, ads appeared on 49 of them, but the advertised brand also appeared in the recommendation list in only two cases. Ads buy exposure; they do not buy recommendation positions.

Then how can my brand get into ChatGPT recommendations?

The sources ChatGPT used as evidence included community reputation, such as Reddit, expert media reviews, such as RTINGS, Forbes, and Allure, and well-structured brand-owned content. The starting point is to organize your content and data so that your brand is clearly and consistently represented across these sources.