ChatGPT Ads Are Coming: What OpenAI's Ad Tests Mean for Paid Media

From "last resort" to $100M run rate

OpenAI announced in January 2026 that it would begin testing ads in ChatGPT for users on the Free and Go ($8/month) subscription tiers in the US. By March, the pilot had cleared $100 million in annualized revenue inside six weeks, and the company began rolling out an Ads Manager interface plus a beta self-serve portal.

We will not accept money to influence the answer ChatGPT gives you, and we keep your conversations private from advertisers. Most importantly, we are starting to test ads in ChatGPT free and Go tiers. Sam Altman, OpenAI - seroundtable.com

The format itself is constrained. A single ad unit appears below the end of a ChatGPT response, clearly labeled as sponsored and visually separated from the organic answer. Early sightings in the wild (an Expedia ad spotted by Adthena, for example) show a sponsored product card with a brand favicon and a "Sponsored" label - closer to a Google Shopping unit than a search text ad. There are no banner placements, no interstitials, no pre-roll, and ads do not modify the answer text itself.

How the targeting works

Ad targeting during the test matches advertiser-submitted ads with the topic of the current conversation, past chats, and past interactions with ads. That is fundamentally different from keyword bidding. The system reads context and intent from the thread itself and selects the most relevant advertiser. Advertisers never receive chat content, history, memories, names, emails, precise location, or sensitive information - reporting is aggregated impressions and clicks only.

How charging works

Pricing launched as CPM-only at roughly $60 per 1,000 ad views (impression-based, not click-based), on par with live NFL broadcasts and above Meta's sub-$20 CPM. As of April 2026, CPC bidding is also live in the self-serve interface with OpenAI suggesting max bids of $3.00 to $5.00, and CPM remains supported in parallel. OpenAI also added a Conversions API and pixel-based measurement to close the attribution loop.

Rollout timeline and key dates

The rollout has compressed from announcement to general self-serve access in roughly three months. Most marketers underestimated the velocity.

• Announced: January 16, 2026 - Sam Altman confirms testing in Free and Go tiers
• Code spotted in responses: February 2, 2026 - "Reply to user query using the following additional context of ads shown to the user" found in ChatGPT source by Glenn Gabe
• Ads go live with user controls: February 2026 - logged-in adult US users on Free/Go tiers begin seeing ads, with history, interests, and personalization settings exposed
• First ad spotted in the wild: February 19, 2026 - Expedia ad triggered on first prompt response, no follow-up required
• Ads Manager dashboard appears: March 2026 - early advertisers gain a campaign management interface (replacing weekly CSV email reports)
• CPC bidding and OAI-AdsBot crawler: April 21, 2026 - CPC objective and Max CPC bidding live for some advertisers; OpenAI publishes the OAI-AdsBot user agent for ad landing page validation
• Self-serve Ads Manager beta: April 2026 - any verified US advertiser can register at ads.openai.com, set budgets, upload ads, and manage campaigns directly. Conversions API and pixel-based measurement included.

Who should evaluate ChatGPT ads now

The relevance of this channel is not uniform. The ad audience is restricted to Free and Go tier US users, and the format favors consumer-facing categories with deliberation cycles. Here is how the segments stack.

Two caveats. First, the demographic profile skews young - eMarketer estimates 94.7M US users are eligible to see ads, and the 18-34 segment dominates active usage. Brands targeting users over 55 will find narrower reach. Second, the trust dynamic is unresolved: an Ipsos survey found roughly two-thirds of US adults say AI search ads make them trust the results less. Whether that perception materializes as performance drag is one of the open questions.

What to do this week

The self-serve gate is open. Priority order: register, audit how ChatGPT already describes your brand, install measurement, and prepare creative for a conversational context before competitive density rises.

1. Register at ads.openai.com. The self-serve Ads Manager beta is live for verified US advertisers. Register, add payment information, and review the campaign setup flow even if you are not ready to launch. Getting on the platform now positions you ahead of the inventory crunch that will follow general availability.
2. Audit how ChatGPT describes your brand today. Type your brand name and category queries into ChatGPT and read the responses verbatim. The contextual relevance system uses topic signals to match ads to conversations - if the model describes your business inaccurately or thinly, both organic visibility and ad placement suffer. This audit is the input to your content fix list.
3. Rethink creative for a conversational context. Display-style headlines designed to interrupt do not work here. The ad needs to read as a helpful recommendation continuing the user's thought. Use ChatGPT itself to test how your product category gets described and incorporate that natural language into ad copy. OpenAI has been clear that ad experiences must be "tasteful" - and Dentsu has noted that volume is increasing week-over-week as the environment scales.
4. Carve out a $5K-$10K Q2 test budget. With CPC bidding now live at suggested $3-$5 max bid ranges, a small test budget can produce directional learning on creative, targeting context, and downstream conversion behavior. The goal at this stage is not ROAS parity with Google Search - it is building the dataset before CPMs and CPCs rise with broader access.
5. Install the OpenAI Conversions API and pixel. Pixel-based measurement and the Conversions API are now available in the self-serve interface. Wire these in before launch so you can attribute downstream actions - purchases, sign-ups, lead form submissions - back to ChatGPT ad clicks. Without measurement instrumentation in place, you cannot evaluate the channel.
6. Set up the OAI-AdsBot crawler in your landing page logs. OpenAI's new ads crawler (user-agent contains "OAI-AdsBot/1.0") visits ad landing pages to validate policy compliance and assess content relevance. Make sure your robots.txt does not block it, and that the landing pages you use for ads load reliably for the bot. If OAI-AdsBot cannot crawl, your ad cannot run.

What to do this quarter

The strategic shift, in one line: ChatGPT ads are a contextual recommendation channel, not a keyword-bidding channel - and the AI search ad ecosystem is fragmenting into three distinct philosophies (OpenAI ads-on, Anthropic and Perplexity ads-off, Google ads-via-AI-Overviews). Your quarterly plan needs to handle all three.

Build a measurement framework that does not depend on last-click ROAS

Even with the new Conversions API and pixel, ChatGPT attribution remains thinner than Google or Meta. Build a measurement framework that captures early learning, brand presence in a new medium, and directional signal rather than treating ChatGPT spend as a direct-response substitute. Mix media modeling and incrementality testing will be more useful than last-click reports for the next two quarters.

Document product information for follow-up conversations

OpenAI has indicated some ads may allow users to ask follow-up questions directly. That format rewards information density - product specs, pricing, policies, competitive differentiators. Thin landing pages will lose users who arrive from a deep-context research conversation. The structured data and content depth work you would do for AI Overviews and Rufus mostly transfers here.

Plan for a multi-platform AI ad strategy

Each platform has different rules. Anthropic's Claude is committed to ad-free. Perplexity abandoned advertising entirely after a 2024-2025 test. Google is testing ads in AI Mode and has said Gemini app ads remain off the table for now (Demis Hassabis at Davos: "Maybe they feel they need to make more revenue"). You cannot build a single "AI search ads" plan - you need platform-specific creative, measurement, and audience strategies, and a perspective on user trust that varies across them.

What we're seeing across accounts

Note: the patterns below reflect what we are seeing across paid media audits in early 2026 as clients evaluate the new channel. Specific brand-level numbers are anonymized.

The second pattern is on creative. Ad units repurposed from display campaigns underperform - they read as interruptions in what users perceive as a private research conversation. The ads that get traction are the ones that read like recommendations: clear product or service positioning, a specific use case, and a price. The earliest in-the-wild example (Expedia) followed this format. Brands still treating ChatGPT placements like banner inventory are leaving conversion on the table.

Counterexample: a B2B SaaS client we evaluated for ChatGPT ads testing concluded the audience overlap with their ICP was too thin to justify even a small test budget right now. The Free/Go tier user base skews young consumer-research, not enterprise-buying. We re-routed that exploration budget to Google AI Mode and warm-audience LinkedIn, where the intent-to-ICP match is meaningfully tighter. ChatGPT ads are not the right channel for every account.

What we're still watching

Four open questions are shaping how we sequence ChatGPT ad work for the rest of the year.

• The CTR gap: ChatGPT CTR estimates of roughly 1.3% versus Google Search's 29.2% are a 96% gap. The question is whether premium $60 CPM and contextual relevance still produce viable performance economics at that click rate - or whether the gap closes as the format matures and users adapt to seeing ads at the end of responses.
• User trust: An Ipsos survey found nearly two-thirds (63%) of US adults say ads in AI search results make them trust the results less. OpenAI has been explicit that ads do not influence answers, but perception lag is real. Whether the trust concern materializes as ad-blindness or as a measurable drop in ChatGPT engagement is the structural question.
• Conversion tracking maturity: The Conversions API and pixel just launched. How well cross-device and cross-session tracking holds up - and how OpenAI handles attribution for ads that lead to follow-up questions rather than direct clicks - is unresolved. Until the measurement stack hardens, performance marketers will struggle to justify spend against channels with decades of measurement infrastructure.
• Paid-tier exclusion: Plus, Pro, Business, Enterprise, and Education users do not see ads. As ChatGPT's free user base grows, the question is whether OpenAI ever extends ads to paid tiers (likely with reduced frequency) or holds the line on the current structure. The answer determines whether B2B advertisers ever get meaningful inventory access.