ChatGPT Ads Hit $1B in 2026: What Advertisers Get Now

OpenAI’s advertising business went from quiet beta to a $1 billion annualized run rate in under a year. As of this month, European advertisers can buy ChatGPT ads through a self-serve dashboard — no sales rep, no minimum spend, no waitlist. That combination matters more than the headline number. A billion-dollar run rate proves the inventory converts well enough that OpenAI is scaling it aggressively; self-serve access means a business owner with a $500 test budget can be live this afternoon. The auction is thin, CPMs are soft, and most of your competitors have not logged in yet — exactly the window where cheap early inventory turns into disproportionate returns.

What’s new about ChatGPT ads

Two things changed at once. First, OpenAI disclosed that its advertising business crossed a $1 billion annualized revenue run rate — roughly $83 million in a single recent month, extrapolated forward. That is a faster ramp than Instagram’s early ad business and faster than TikTok’s. Ads inside ChatGPT are not a side experiment tolerated to appease investors; they are becoming a core revenue line alongside subscriptions and API usage. A product generating that kind of money gets engineering headcount, targeting improvements, and reporting tooling.

Second, and more actionable: the OpenAI advertising platform opened self-serve buying to advertisers across the EU, UK, and EEA. Buying ChatGPT ad placements previously meant a managed-service relationship — you talked to a salesperson, committed to a spend floor, and got a media plan. ChatGPT self-serve ads in Europe collapse that into a credit card and a campaign builder. You pick objectives, set budgets, define audiences by topic and intent signal rather than demographic guesswork, and upload creative. The ads appear as sponsored suggestions and clearly labeled recommendation cards inside conversational responses, primarily on the free tier. Paid subscribers see far less inventory, and Pro tiers currently see none.

The format is genuinely different from search or social. A ChatGPT ad is not a blue link fighting nine other blue links, and it is not an interruption in a scroll feed. It appears when a user has articulated an intent in natural language — “what’s the best CRM for a two-person agency,” “help me plan a kitchen renovation budget” — and the model surfaces a relevant sponsored option alongside its organic answer. Intent is richer than a keyword. That is why early ChatGPT ad revenue run rate growth has been so steep: click-through rates on well-matched placements run meaningfully above display benchmarks, because the ad answers a question the user just asked in their own words.

Why it matters

• Auction density is the whole opportunity. Google Ads and Meta are mature auctions where you pay a price reflecting full competitive pressure. ChatGPT’s auction has a fraction of the advertisers bidding against you. Early Meta advertisers in 2012 and early TikTok advertisers in 2019 both captured a two-to-three-year arbitrage window. This is the same shape of opportunity.
• Intent quality beats audience targeting. You are not buying a lookalike audience and hoping. You are buying placement against a stated problem. For considered purchases — software, services, home improvement, B2B tools — that beats a demographic segment.
• Self-serve removes the gatekeeper. Small and mid-sized businesses were structurally locked out while buying required a managed relationship. The Europe rollout, and the US self-serve tier before it, mean a solo operator and a Fortune 500 brand now use the same interface.
• It creates a second front in AI visibility. Businesses have spent 2025 and 2026 worrying about whether AI assistants mention them organically. Paid placement is the deterministic version of that problem. If you cannot earn the mention, you can buy adjacency to it.
• Attribution is still immature — which cuts both ways. Reporting is thinner than Google’s. That deters sophisticated buyers with strict MER targets, which keeps competition low. Tolerate coarser measurement, test with holdouts, and cheap inventory is your compensation.
• The window is closing, not opening. A $1B run rate growing this fast means the advertiser count is climbing. Every quarter you wait, the auction gets more expensive. Testing now is cheaper than testing in six months.

How to advertise on ChatGPT today

Here is the practical path from zero to a running campaign. Execute this section rather than skim it.

1. Create an advertiser account. Go to the ads dashboard at ads.openai.com and sign in with the account that owns your business relationship. Complete business verification — legal entity name, tax ID, and for EU advertisers, VAT registration. Verification typically clears in one to three business days. Do this first; everything else waits on it.

2. Install the conversion pixel before you spend a euro. Attribution is the weak point of this channel, so instrument it properly on day one. Drop the base tag in your site’s <head>:

   <script async src="https://ads.openai.com/pixel.js?id=YOUR_PIXEL_ID"></script>
   <script>
     window.oaiq = window.oaiq || [];
     function oai(){ oaiq.push(arguments); }
     oai('init', 'YOUR_PIXEL_ID');
     oai('track', 'PageView');
   </script>

   Then fire a conversion event on your thank-you or order-confirmation page:

   oai('track', 'Purchase', {
     value: 149.00,
     currency: 'EUR',
     order_id: 'ORD-10482'
   });

3. Build your first campaign around intent clusters, not keywords. This mental shift separates advertisers who succeed here from those who port Google campaigns over and wonder why performance is flat. Instead of bidding on “project management software,” describe the situations where your product is the answer. Write out ten to fifteen real questions a prospect would type into ChatGPT, in full sentences, and use those as your targeting seeds.

4. Create the campaign via API if you want repeatability. The dashboard handles a first test, but scripting it lets you version-control your setup and clone it across markets:

   curl -X POST https://api.openai.com/v1/ads/campaigns \
     -H "Authorization: Bearer $OPENAI_ADS_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "name": "EU-Test-Q3",
       "objective": "conversions",
       "daily_budget": { "amount": 4000, "currency": "EUR" },
       "bid_strategy": "target_cpa",
       "target_cpa": { "amount": 3500, "currency": "EUR" },
       "geo": ["DE", "NL", "IE"],
       "conversion_event": "Purchase",
       "status": "paused"
     }'

   Budgets are in minor units — 4000 is €40.00 per day. Create it paused, review it in the UI, then activate.

5. Write creative that sounds like an answer, not an ad. Tone is the biggest performance lever in this channel. Your copy sits next to a helpful conversational response, and copy that screams “LIMITED TIME OFFER” reads as noise and gets ignored. Write the way a knowledgeable friend recommends something. A useful prompt for drafting variants:

   You are writing a sponsored recommendation that will appear
   inside a ChatGPT answer.
   
   Product: [your product]
   The user just asked: "[the exact question you're targeting]"
   
   Write 5 variants, each under 90 characters, that:
   - Name the specific problem the user described
   - State one concrete differentiator (not "best in class")
   - Read like a recommendation, not a banner
   - Contain no exclamation points, no urgency language
   
   Then rank them by how likely a skeptical reader is to click.

6. Set a holdout and run for at least 14 days. Attribution is coarse, so judge the channel on incrementality rather than the dashboard’s self-reported ROAS. Hold out one comparable geo, run the campaign in the others, and compare total conversions. Budget €1,500–€3,000 for a first read. Anything less and you are reading noise.

7. Pull performance data on a schedule so you are not living in the UI:

   curl -G https://api.openai.com/v1/ads/reports \
     -H "Authorization: Bearer $OPENAI_ADS_KEY" \
     --data-urlencode "campaign_id=camp_abc123" \
     --data-urlencode "start_date=2026-08-01" \
     --data-urlencode "end_date=2026-08-31" \
     --data-urlencode "breakdown=intent_cluster" \
     --data-urlencode "metrics=impressions,clicks,spend,conversions"

   The intent_cluster breakdown is the most valuable field in the reporting suite. It tells you which phrasings of the underlying question actually convert, which you feed back into targeting and into your organic content strategy.

How ChatGPT ads compare to Google, Meta, and Perplexity

Factor	ChatGPT Ads	Google Search Ads	Meta Ads	Perplexity Ads
Targeting basis	Conversational intent clusters	Keyword match types	Interest and behavior signals	Query topic categories
Auction competition	Low — early advertiser base	Very high, mature	Very high, mature	Low, but small reach
Typical CPM range	Soft; well below search parity	High in commercial verticals	Moderate to high	Low
Self-serve access	Yes — US and now Europe	Yes	Yes	Limited, largely managed
Reporting depth	Basic; improving quarterly	Deep and granular	Deep and granular	Basic
Audience scale	Very large weekly active base	Largest	Largest	Niche, research-heavy
Creative format	Inline sponsored recommendation	Text link and shopping unit	Image, video, carousel	Sponsored follow-up question

The honest read: Google still wins on measurement and on capturing bottom-funnel commercial queries. Meta still wins on creative flexibility and audience scale for impulse categories. ChatGPT offers what neither does — placement inside a genuine consultative moment, at a price that has not yet been bid up. Treat it as a third leg, not a replacement.

What’s next for the OpenAI ads manager

Expect the format set to widen fast. The current inline recommendation card is deliberately conservative. OpenAI has been careful not to damage trust in the assistant, because the entire value of the placement depends on users believing the model’s answers are not for sale. Watch for shopping-style units with pricing and inventory pulled from merchant feeds, which would make the channel directly competitive with Google Shopping for e-commerce. Agentic commerce is the bigger prize: if ChatGPT can complete a purchase inside the conversation, the gap between ad impression and transaction collapses to nothing, and CPMs will reprice accordingly.

On the buy side, the OpenAI ads manager still lacks what sophisticated advertisers consider table stakes — proper multi-touch attribution, offline conversion imports, audience exclusion lists, and integration with the major bid-management platforms. Each of those shipping is simultaneously good news and bad news: better tooling, but also the trigger that brings agency money into the auction. Your cheap inventory has a shelf life measured in quarters.

Track the regulatory dimension too, particularly in Europe. The EU AI Act’s transparency obligations and the Digital Services Act’s rules on ad disclosure and profiling both apply to AI chatbot advertising in 2026, and the boundary between “sponsored recommendation” and “model output” is exactly the kind of thing regulators will want drawn clearly. Expect labeling requirements to tighten. Advertisers will survive that, but it may constrain how native these placements are allowed to look, which would in turn affect click-through rates. Build your business case on incremental profit at current performance, not on the assumption that today’s CTRs hold forever.

Frequently Asked Questions

How much does it cost to start advertising on ChatGPT?

There is no minimum spend on the self-serve tier. Practically, budget €1,500–€3,000 over two weeks for a first test that produces a readable signal. Below that you are gathering anecdotes rather than data, and you will not have enough conversions to judge whether the channel works for your offer.

Can advertisers outside Europe and the US buy ChatGPT ads?

Self-serve is currently live in the US, EU, UK, and EEA. Advertisers in other regions can generally still access managed buying through OpenAI’s sales team, and additional self-serve markets are expected to open through the rest of 2026. If you are outside the current footprint, start business verification now so you are not queued behind everyone else when your market opens.

Do ChatGPT ads influence the model’s actual answers?

No. OpenAI’s stated policy is that paid placement does not alter organic response content. Ads appear as separately labeled units adjacent to the answer. This separation is load-bearing for the entire product, so expect strict enforcement. It also means you cannot buy your way into being recommended organically; that remains an earned-visibility problem.

What kinds of businesses perform best on this channel?

Considered purchases with a research phase — B2B software, professional services, education and training, financial products, home improvement, healthcare services. Anything a person would naturally ask an assistant to help evaluate. Low-consideration impulse products perform worse here than on Meta, because the user is in problem-solving mode rather than browsing mode.

How do I measure results when the reporting is limited?

Use geo holdouts and compare total business conversions between test and control regions rather than trusting platform-reported ROAS. Install the conversion pixel for directional data, but make the spend decision on incrementality. Add a post-purchase “how did you hear about us” field — self-reported attribution is imperfect but catches assisted conversions the pixel misses entirely.

Should I move budget out of Google or Meta to fund this?

No. Fund the test from a separate experimental budget, or from the bottom-performing 5–10 percent of your existing spend. An early test on a thin auction is about asymmetric upside on a small stake, not reallocating proven demand capture into an unproven channel. If the holdout shows real incrementality after 30 days, scale it with real money.