ChatGPT Ads for Ecommerce: How They Work and What They Cost

ChatGPT started testing ads on February 9, 2026, for free-tier US users. By May, OpenAI had opened a self-serve Ads Manager with no minimum spend. Here's exactly how the ad format works, what it costs today, and whether it belongs in your 2026 budget.

---

TL;DR

• What: Sponsored placements inside ChatGPT — a clearly labeled, visually separated box below an AI answer. Ads never change what ChatGPT actually tells the user.
• Who sees them: Logged-in adult US users on the Free and ChatGPT Go tiers only. Plus, Pro, Business, Enterprise, and Education tiers are ad-free.
• How targeting works: Ads are matched to the topic of the conversation, the user's past chats, and past ad interactions — not keywords, not demographic profiles. Advertisers never see the underlying conversation.
• Cost today: Self-serve CPC bidding starting around $3–$5, CPM options as low as roughly $25 in some pilot categories, no minimum spend as of the May 2026 self-serve launch — down from a $60 CPM / $200,000 minimum at the February launch.
• Where it's live: US since February 9, 2026, expanded to Canada, Australia, and New Zealand in spring 2026, then the UK on June 6, 2026 (OpenAI's first European market), with Mexico, Brazil, Japan, and South Korea announced for rollout starting May 7, 2026.

---

The launch, in OpenAI's own words

OpenAI's official framing, published February 9, 2026: "Today, we're beginning to test ads in ChatGPT in the U.S. The test will be for logged-in adult users on the Free and Go subscription tiers... Ads do not influence the answers ChatGPT gives you, and we keep your conversations with ChatGPT private from advertisers."

That distinction — ads that don't touch the answer — is the single design decision the whole program is built around. It shows up twice more in OpenAI's own updates: a March 26, 2026 post reporting "no impact on consumer trust metrics, low dismissal rates of ads" from the early test, and a May 7, 2026 expansion post reiterating the same commitment as the pilot grew into new markets.

If you prefer not to see ads at all, OpenAI's own workaround is upgrading to Plus or Pro, or opting out within the Free tier in exchange for fewer daily free messages — a detail worth knowing if you're trying to understand how much of your own target audience is even eligible to see your ad.

---

Where the ad actually appears

Ads show up as a clearly labeled, visually separated box below the AI's answer — not inline, not blended into the response text. OpenAI's stated goal is that the answer itself is "optimized based on what's most helpful to you," with the ad living entirely outside that boundary.

That single placement pattern (one contextual box, one advertiser shown at a time when there are multiple matches) is a real constraint worth planning around: there's no equivalent yet to Google's multiple ad slots per results page, or Meta's dense, scrollable feed inventory. Whoever wins the single relevant slot for a given conversation gets the whole placement.

---

How targeting actually works

This is the part most performance marketers get wrong on their first pass, because it doesn't map cleanly onto keyword bidding or audience segments.

OpenAI's own description: ads are matched by "the topic of your conversation, your past chats, and your past interactions with ads." Their example — a user researching recipes may see an ad for meal kits or grocery delivery — is contextual, not keyword-based. There's no traditional keyword match type, and no third-party demographic targeting layer to configure.

Two structural privacy commitments shape what advertisers can and can't do:

• Advertisers never see the conversation. No chat content, chat history, memories, or personal details are shared — only aggregate performance data like impression and click counts.
• No targeting near sensitive topics. OpenAI does not show ads in accounts it predicts belong to users under 18, and ads are not eligible to appear near health, mental health, or political conversations.

Practically, this means your targeting lever is the content and framing of your ad submission — the merchant description and category mapping you provide — rather than an audience you build and refine over time the way you would in Meta Ads Manager.

---

What it actually costs right now

Pricing has moved fast and in one direction: down and more accessible.

Phase	CPM	Minimum spend	Bidding options
Launch (Feb 9, 2026)	~$60	~$200,000	Managed, invite-only
Self-serve (May 2026 onward)	as low as ~$25 in some pilot categories	None	CPC (~$3–5 starting bid), CPM, CPA (activated June 5, 2026)

The self-serve Ads Manager at ads.openai.com is the on-ramp most ecommerce brands will actually use — no agency relationship or minimum commitment required, in contrast to the invite-only managed launch. OpenAI has also brought in agency partners (Dentsu, Omnicom, Publicis, WPP) and ad-tech partners (including Adobe) for teams that want to buy through existing workflows instead of the self-serve tool directly.

Creative requirements are lean by design: a merchant description in the 40–80 word range, a square hero image, and category mapping — no video, no carousel formats yet, and reported production cost roughly 70% lower than a comparable Google Performance Max campaign.

---

Who's already running ads — our own crawl of the US market

Rather than rely on secondhand reports, we ran our own research pass: 2,560 high commercial-intent US prompts submitted directly to ChatGPT, capturing every sponsored placement returned. The result is a live, first-party read on who's actually spending.

Headline numbers from our crawl:

• 220 distinct US advertisers identified across 1,746 captured ad creatives
• 72 are direct consumer ecommerce brands — the category most relevant to most readers here
• Ad density is concentrated: B2B SaaS accounts for 37.9% of all captured ads, consumer commerce 30.7%, with the remainder split across consumer services, affiliate aggregators, marketing agencies, and B2B supply

The heaviest ecommerce spenders we observed (by creative volume): Harbor Freight, Home Depot, JCPenney, Peloton, DreamCloud, Macy's, Dell, SkinCeuticals, Wayfair, and Avocado Green Mattress. Categories skew toward home goods, apparel, fitness equipment, and beauty — all categories where shoppers plausibly use ChatGPT to compare options before buying, matching the "evaluation phase" framing covered earlier in this piece.

What they're actually advertising isn't generic brand awareness — it's specific product and category pages: Harbor Freight running cordless drill kits and pressure washers, Peloton pushing treadmill comparisons, DreamCloud running mattress-in-a-box explainer creative, JCPenney promoting specific footwear categories. The pattern across nearly every advertiser we captured is direct-response, not brand-building.

Consumer research backs up why retail leads: a 2026 Pacvue survey of 1,008 US consumers found 53% use AI tools to research products, and 28% turn to AI for shopping research daily.

---

Should you actually spend here in 2026

Three honest signals to weigh:

In favor: No minimum spend anymore, CPC bidding as low as $3–5, and a genuinely high-intent moment — ChatGPT ads reach people mid-evaluation, actively comparing options, not passively scrolling.

Against: Single ad slot per conversation means real competition for the relevant moment, targeting is contextual rather than granular, and the ad format itself is still limited to one static creative type. This is an early-stage channel, not a mature one.

The bigger picture: paid placement inside ChatGPT is one lever. The other is being the brand the answer itself recommends — unpaid, based on how well your product data and content are structured for AI citation. Most brands only think about the first lever. We work on both: AI-Sponsored Placements if you want to run paid campaigns, and Citation Rank to improve how often you show up organically in the answer itself, not just the ad slot beneath it.

If you're weighing ChatGPT Ads against your existing Google and Meta spend, book a demo — we'll walk through where your category is likely to see the strongest early return.

---