ChatGPT Ads Manager in 2026: How Conversational Intent Changes Everything About Where You Spend

Why I'm Writing This Now

I've run over $50M in ad spend across Meta, Google, Bing, and TikTok. Every new channel launch follows the same pattern: early hype, vague documentation, a handful of brands throwing money at it without a framework, and then either it matures into a real performance channel or it quietly fades. ChatGPT Ads Manager is in that early phase right now — and the stakes are different because the intent model is fundamentally unlike anything we've worked with before.

This isn't another "AI will revolutionize advertising" piece. I don't care about revolution. I care about whether this channel can deliver measurable outcomes for the budgets I manage, and what it takes to find that out without wasting six weeks of spend. What follows is a tactical breakdown of what's available, how conversational intent actually works, where the gaps are, and how I'd structure a test if I were allocating budget today.

What ChatGPT Ads Manager Actually Is

OpenAI launched a self-serve ChatGPT Ads Manager at ads.openai.com in May 2026. It's US-only for self-serve buyers right now, with no minimum spend — you can start with $50 if you want. The UK, Canada, Australia, and New Zealand are available only through managed or agency arrangements, which tells you OpenAI is being cautious about scaling internationally before they've nailed the US model.

The core ad format is sponsored chat cards that appear beneath ChatGPT's replies. These show up for logged-in adult users on the Free and Go tiers — not Plus or Pro subscribers, which is a deliberate choice. OpenAI is keeping the paid tiers ad-free for now, presumably to avoid alienating the subscribers who fund their core business.

Here's what the inventory looks like at launch:

• Sponsored chat cards: Text-plus-image cards inserted after ChatGPT's response, clearly labeled as sponsored. The user sees the ad as part of the conversation flow, not as a sidebar interruption.
• No search results placement yet: Unlike Google, there's no separate "ads above organic results" slot. The ad lives inside the conversation thread. This is a meaningful structural difference I'll dig into later.
• No display or video formats: It's card-based, static, and compact. If you're used to Meta's creative flexibility, this feels constrained — and that constraint is intentional.

Reports suggest the platform has already crossed $100M+ in annualized revenue [unverified — this figure comes from third-party reporting, not an official OpenAI statement]. Whether that number is exact or slightly inflated, the trajectory is real: advertisers are buying, and OpenAI is scaling the infrastructure.

How Targeting Works: Context Hints, Not Keywords

This is the part that throws most marketers. There are no keyword lists. No audience segments. No interest targeting. Instead, ChatGPT Ads Manager uses context hints — you describe the conversational intent you want to reach, and OpenAI's system matches your ad to relevant chat sessions.

A context hint isn't a keyword. It's a description of the conversation topic and user intent. For example:

• Keyword approach (Google): Target "CRM software"
• Interest approach (Meta): Target people interested in "business software"
• Context hint approach (ChatGPT): "User is asking ChatGPT for recommendations on CRM tools for a small team with budget constraints"

The difference is depth. A keyword matches a word. An interest matches a demographic profile. A context hint matches a conversation — the specific problem the user is describing to ChatGPT in that moment. That's why CPC starting bids are in the $3–5 range and CPM sits around $60 [approximate early beta figures — these will likely shift as supply and demand stabilize]. You're paying for intent specificity, not broad reach.

For anyone who's built custom audience segments in Meta or run long-tail Google campaigns, context hints will feel both familiar and strange. Familiar because you're describing who you want to reach. Strange because you're doing it in natural language instead of selecting from a dropdown menu. The craft shifts from "which boxes do I check" to "how well do I understand the conversations my customers are having with AI."

The Broader AI Platform Ad Landscape

ChatGPT isn't the only AI platform selling ads. Here's the current map:

Perplexity

Perplexity has been running sponsored follow-up questions and cited brand mentions in its AI search results. The model is closer to Google — ads appear alongside search-style results rather than inside a conversation thread. Perplexity's advantage is that its users are explicitly searching for information, so intent is more direct. The disadvantage is smaller reach compared to ChatGPT's user base. If you want to test AI search ads with a more familiar structure, Perplexity is the lower-risk starting point.

Reddit

Reddit is rolling out AI-powered shopping ads, announced at Cannes in 2026 [specific Cannes announcement details need separate verification]. The angle here is different: Reddit's AI ads leverage conversation context from subreddit discussions, but the format is still traditional Reddit ad units (promoted posts, comment threads). It's AI-enhanced targeting on a conventional ad platform, not a new ad format native to AI interaction. For a deeper look at how Reddit's ad ecosystem is evolving, see Reddit Ads in 2026.

Where this is heading

Expect Google to integrate AI Mode responses with ad inventory within months — they've already been testing sponsored citations in AI Overviews. Microsoft will likely follow with Copilot ad placements. The pattern is clear: every major AI interface will have paid placement, and each one will handle intent matching differently based on how their users interact with the AI.

Conversational Intent vs Search Intent vs Interest Intent

This is the conceptual shift that matters most for how you structure campaigns. Let me map the three models:

Google Search: Keyword Intent

The user types a query. The query is short, explicit, and transactional. "Buy running shoes" → the intent is clear. You match the keyword, you bid on the keyword, you write ad copy that responds to the keyword. The entire system is built on query-to-offer matching. It's efficient because the user has already declared what they want.

Meta: Interest and Behavior Intent

The user never declared intent. Meta infers it from browsing behavior, engagement patterns, and demographic signals. You target "people likely to be interested in running shoes" based on past actions. The system is built on prediction-to-offer matching. It works at scale because Meta has enough behavioral data to make good predictions, but the intent is always inferred, never explicit.

ChatGPT: Conversational Intent

The user is describing a problem, a scenario, a decision they're trying to make. They're not typing a keyword — they're having a conversation. "I'm training for a half marathon and my current shoes keep causing shin splints, what should I look for in a replacement?" That's not a keyword. It's a rich, contextual intent signal that contains the problem, the context, and the desired outcome all at once.

The system is built on conversation-to-offer matching. Your ad appears because the conversation context matches your context hint, not because the user typed your target phrase or fits your audience profile.

Why this changes campaign structure:

• You don't need 50 keyword variants to capture intent. One well-written context hint can cover the conversational range.
• Your ad copy needs to respond to the conversation, not just the keyword. A user asking about shin splints needs different messaging than one asking about shoe brands.
• Broad awareness campaigns don't work here. Conversational intent is specific — your ad needs to be specific too.

I've scaled Meta accounts from $10k/month to $500k/month by combining manual bidding with custom automation. The automation handled budget pacing, audience refresh, and creative rotation that Meta's native tools couldn't do well enough. That experience taught me something relevant here: the channels where you can most precisely match intent are the ones where manual strategy beats pure automation at scale. ChatGPT's context hint system gives you that precision — but only if you understand conversational intent well enough to write the hints.

The Measurement and Attribution Gap

Here's the problem nobody has solved yet. When a user chats with ChatGPT, sees your sponsored card, clicks through to your site, and converts — that's the happy path. But what actually happens most of the time?

• The user sees your card, reads it, and continues the conversation instead of clicking. They ask ChatGPT more questions about your product. They might convert days later from a different session, different device, different channel.
• The user clicks your card, bookmarks your site, and converts next week. The attribution window may or may not catch that.
• The user asks ChatGPT about your brand in a later session (without seeing your ad), and ChatGPT recommends you based on the earlier conversation context. That's an organic AI citation influenced by your paid placement — and it's completely unattributed.

ChatGPT Ads Manager offers a conversion pixel and Conversions API, similar to Meta's CAPI. That handles direct click-to-conversion measurement. But the indirect influence — the conversation that continues, the later-session recall, the organic AI recommendation — is invisible in current attribution models.

This is worse than the attribution gap on Meta or Google because AI chat sessions are multi-turn, ongoing, and cross-session. A Google search is a single event. A Meta ad impression is a single event. A ChatGPT conversation can span days and evolve. The standard attribution frameworks weren't built for this.

What I'd do about it right now:

• Install the conversion pixel and Conversions API — get the direct measurement working first.
• Run a holdout test: geo-based or audience-based, comparing periods with ChatGPT ads active vs paused. This is the only reliable way to measure total incremental impact including indirect effects.
• Track brand search volume on Google during your ChatGPT ad test. If ChatGPT ads are influencing later conversions, you'll see lift in branded search even if direct attribution looks weak.
• Don't expect the platform's attribution to tell the full story. It won't.

Creative Format Constraints: What Works in Sponsored Chat Cards

The sponsored chat card format is narrow. You're working with:

• A short headline (roughly 25–30 characters based on what I've seen in documentation)
• A brief description (roughly 75–100 characters)
• A small image or logo
• A clear "Sponsored" label that users will see before they engage

This is not Meta's creative playground. You can't do carousel storytelling, video hooks, or long-form copy. The constraint is real, and it forces a specific creative discipline: respond directly to the conversation context with a concise, useful offer.

What works based on the format's constraints:

• Direct utility over brand storytelling. The user is in a problem-solving conversation. Your card should offer the next step in solving that problem — not tell a brand story. "Free CRM trial for teams under 10" beats "The #1 CRM platform trusted by thousands."
• Specificity over generality. Context hints target specific conversations. Your creative should match that specificity. If you're targeting conversations about budget-friendly CRM options, your card should mention pricing or team size — not generic CRM benefits.
• Clear next action. The user is mid-conversation. Give them a reason to break out and take action. Free trial, comparison tool, specific pricing page — something that moves them from chatting to evaluating.

If you need to generate dozens of specific creative variants fast — because context hints require more tailored messaging than broad keyword campaigns — check out AI Ad Creative at Scale for a workflow that handles that volume.

Budget Allocation: How to Test Without Burning Money

New channels always tempt two extremes: throwing money at them to "see what happens" or waiting too long and missing the learning window. Here's the framework I'd use, based on how I've tested every new channel from TikTok to Bing over the past decade.

Phase 1: Structured Test (2–3 weeks, $500–$1,500)

Goal: answer the question "does conversational intent in this channel produce clicks and conversions for my offer?"

• Run 3–5 context hints targeting different conversation scenarios for your core offer.
• Use CPC bidding at the floor ($3–5 range). Don't bid up for volume — you're testing signal, not scale.
• Write 2–3 creative variants per context hint, each responding directly to that conversation type.
• Track direct conversions via pixel + CAPI. Track branded search lift on Google as a proxy for indirect impact.
• Set a clear kill criteria: if you see less than 0.5% click-through rate across all hints after 10,000 impressions, the intent matching isn't working for your offer. Pause and reassess.

Phase 2: Validation (2–4 weeks, $2,000–$5,000)

Goal: confirm the signal is real and understand which conversation contexts perform.

• Double down on the context hints that showed click-through rates above 1% in Phase 1.
• Pause the ones that didn't work. Don't keep funding weak signals hoping they'll improve — they won't.
• Run a holdout test: pause ChatGPT ads for one week while keeping everything else constant. Compare conversion volume and branded search. This tells you whether the channel is driving incremental results or just cannibalizing existing demand.
• If holdout shows clear incremental lift, you have a real channel. If it doesn't, you have an attribution mirage — stop scaling.

Phase 3: Scale (ongoing, budget based on validated ROAS)

Goal: allocate budget proportionally to validated incremental ROAS.

• If validated incremental ROAS is within 30% of your Meta or Google benchmarks, allocate 5–10% of total digital budget to ChatGPT ads. Not 50%. This is a complementary channel, not a replacement.
• If validated incremental ROAS is 50%+ below benchmarks, keep it at test budget levels and revisit quarterly. The channel may mature — CPCs may drop, targeting may improve, formats may expand. Don't kill it, but don't scale it.
• Build custom reporting that pulls ChatGPT ad data alongside your other platforms. I migrated my own reporting from manual exports to a real-time dashboard pulling from 5 ad platforms via API, normalized in Python, visualized in Looker Studio. Cut reporting time from 4 hours/week to near-zero. You'll need that efficiency when you're managing 6+ channels including AI platforms. Add ChatGPT's data to that pipeline early.

Risks: Brand Safety, Hallucination Placement, and Targeting Limits

The risks on AI platform advertising aren't the same as on social or search. They're specific to the conversational medium, and most marketers aren't thinking about them yet.

Brand safety in AI conversations

Your sponsored card appears after ChatGPT's response. That response could be accurate, incomplete, or flat-out wrong. If ChatGPT hallucinates a recommendation that contradicts your ad, your brand is sitting next to misinformation — and the user may not distinguish between the AI's answer and your paid placement.

Example: a user asks about investment strategies. ChatGPT gives a plausible but incorrect answer. Your financial services ad card appears right below it. Even though your ad is accurate, the context is contaminated. This is why financial services, health, dating, and politics are restricted categories on ChatGPT Ads Manager [restricted sectors list from third-party reporting — verify against OpenAI's official advertising policies page before relying on this for campaign planning]. OpenAI is being smart about this. If you're in an unrestricted category, you still need to think about what conversations your ads might appear alongside.

The hallucination-adjacent placement problem

This risk doesn't exist on Google (where ads appear next to organic results that are generally accurate) or Meta (where ads appear in a feed context unrelated to information accuracy). On ChatGPT, your ad is contextually bound to the AI's response. If that response is wrong, your ad inherits that context whether you like it or not.

There's no current solution for this beyond monitoring. I'd recommend:

• Regularly search ChatGPT for the conversation topics you're targeting with context hints. Read the responses. If you see consistent hallucination patterns in those topics, pause those hints.
• Avoid targeting conversations where the AI is likely to give controversial or uncertain answers. Stick to topics where ChatGPT is reliably accurate.
• Watch for user screenshots on social media showing your ad next to a bad AI response. That's a PR risk that standard brand safety tools won't catch.

Limited targeting granularity

Context hints are powerful for intent matching, but they don't give you the demographic, geographic, or behavioral filters you're used to on Meta and Google. You can't exclude specific audiences, can't layer intent with demographics, and can't create lookalike segments. For now, you're trading targeting precision for intent precision. That trade-off works for some offers and fails for others.

If your product requires demographic qualification (e.g., B2B software targeting specific company sizes, or financial products targeting specific income levels), context hints alone may not be sufficient. You'll need to qualify users on your landing page or through your conversion flow, which means your ChatGPT ad click-through rates may look decent but your conversion rates may be lower than what you see on channels with tighter targeting.

GEO vs Paid: When to Optimize for Organic AI Citations vs Buy Ads

This is the strategic decision that most content about ChatGPT advertising ignores. You have two paths to visibility on AI platforms:

• Paid: Buy sponsored chat cards through ChatGPT Ads Manager. You control the placement, the message, and the timing. You pay for it.
• Organic (GEO — Generative Engine Optimization): Make your content citable so that ChatGPT, Perplexity, and other AI systems reference your brand in their natural responses. You don't control the exact placement or wording, but you don't pay for it either.

The decision framework isn't "which is better." It's "which solves the problem I have right now."

Start with paid when:

• You need immediate, guaranteed visibility on AI platforms. GEO takes months to build authority and citation patterns. Paid gives you placement today.
• Your offer is transactional and time-sensitive — free trials, product launches, seasonal promotions. You need users to act now, not discover your brand organically over weeks.
• You're testing whether AI platform intent works for your category. Paid is a faster diagnostic than waiting for organic citations to accumulate.

Start with GEO when:

• You're building long-term brand authority in a category where AI systems will repeatedly reference trusted sources. If you want ChatGPT to recommend your brand consistently, you need the content foundation that makes you citable.
• Your budget is limited and you can't justify test spend on a new channel yet. GEO costs content production time, not ad spend.
• You're in a restricted ad category (financial services, health) where paid placement isn't available but organic citations are.

The best approach is both, sequenced: Test with paid first to validate the channel and learn which conversation contexts matter. Then invest in GEO for those same contexts so you build organic citation equity over time. Eventually, your organic visibility reduces your paid spend requirements — the same pattern we've seen with Google SEO and SEM for 20 years.

For the organic side of this strategy, SEO in the age of AI search covers how to make your content citable by AI systems. That's the companion piece to what we're covering here.

Connecting This to Your Existing AI×Ads Workflow

ChatGPT advertising doesn't exist in isolation. It's part of a broader shift where AI is both the interface where ads appear and the tool that manages campaigns on existing platforms. Here's how the pieces connect:

Managing existing campaigns through AI

While you're testing ads on AI platforms, you can also use AI to manage your Meta and Google campaigns more efficiently. I've been experimenting with Meta Ads MCP — a protocol that lets you control Meta ad campaigns through AI interfaces. The potential is real: automated budget adjustments, creative rotation, audience management through conversational commands. But I'm not trusting it with live spend yet, and I explained why in that post. The point is that AI is entering the advertising stack from both sides — as ad inventory and as campaign management infrastructure.

AI marketing agents for autonomous campaigns

The next step beyond MCP-style interfaces is fully autonomous AI agents that plan, execute, and optimize campaigns. That's still early — most "AI marketing agents" right now are glorified chatbots that generate copy suggestions. But the trajectory is toward agents that can manage multi-channel budgets, including ChatGPT ad spend, as part of an integrated system. When that matures, the conversational intent data from ChatGPT ads becomes input for the agent's optimization logic across all your channels.

The full picture

Think about it as three layers:

• Layer 1 — Ad inventory: ChatGPT, Perplexity, Reddit AI ads. New places to reach users.
• Layer 2 — Campaign management: MCP interfaces, AI agents. New ways to manage existing campaigns.
• Layer 3 — Organic visibility: GEO, citable content. New ways to earn unpaid AI visibility.

All three layers are evolving simultaneously. The marketers who will benefit most are the ones who understand how they interconnect rather than treating each as a separate experiment.

What I'd Do This Week

If you're running significant Meta or Google spend and you haven't touched ChatGPT Ads Manager yet, here's the sequence:

1. Set up a $500 test. Three context hints, two creative variants per hint, CPC floor bidding. No optimization, no scaling — just data collection.
2. Install the pixel and Conversions API. Get direct measurement running before you spend a dollar. Without it, you're flying blind.
3. Track branded search on Google during the test. This is your best proxy for indirect impact from ChatGPT ad exposure.
4. Read the conversations. Search ChatGPT for the topics you're targeting. Understand what the AI is actually telling users before your ad appears next to those responses.
5. Run a one-week holdout after Phase 1. Pause ChatGPT ads, keep everything else constant, compare results. This is the only reliable test for incremental impact.

Don't overthink it. Don't underthink it either. The channel is real, the intent model is different, and the early data will tell you whether it belongs in your mix. I've seen enough channel launches to know that the marketers who test rigorously in the first 90 days either find a genuine performance edge or avoid wasting six months of budget on a channel that doesn't work for their category. Both outcomes are valuable.

If you want to go deeper on the adjacent pieces — either using AI to manage your existing Meta campaigns or building organic AI visibility alongside your paid strategy — those posts break down the next steps from each angle.

Frequently Asked Questions

• How do ChatGPT ads work in 2026?

  ChatGPT Ads Manager (ads.openai.com) runs sponsored chat cards that appear beneath ChatGPT's responses for logged-in adult users on Free and Go tiers. You create context hints describing the conversational intent you want to reach — there are no keyword lists or audience segments. The system matches your ad to relevant chat sessions based on conversation context. CPC starts around $3–5, CPM around $60 (approximate early beta figures). Self-serve is US-only; international markets are available through managed arrangements.

• What are context hints in ChatGPT ads?

  Context hints are natural-language descriptions of the conversation intent you want your ad to appear alongside. Instead of targeting keywords like 'CRM software' (Google) or interests like 'business software' (Meta), you write something like 'User is asking ChatGPT for recommendations on CRM tools for a small team with budget constraints.' The system matches your ad to chat sessions where that conversational intent is present. This gives you deeper intent specificity than keywords or interests, but requires understanding how your customers actually describe their problems to AI.

• How much do ChatGPT ads cost?

  CPC bidding starts in the $3–5 range and CPM sits around $60, based on early beta data — these figures will likely shift as the platform matures and supply/demand stabilize. There's no minimum spend for self-serve buyers. You can start testing with $50, though I'd recommend a structured test budget of $500–$1,500 across 3–5 context hints to get meaningful signal before deciding whether to scale.

• How is AI platform intent different from Google Search intent?

  Google Search intent is keyword-based — the user types a short, explicit query like 'buy running shoes' and you match that keyword. AI platform intent is conversational — the user describes a problem in natural language like 'I'm training for a half marathon and my shoes keep causing shin splints, what should I look for?' That conversational signal contains the problem, context, and desired outcome all at once. It's richer than a keyword but requires different campaign structure: fewer targeting inputs, more specific creative, and ad copy that responds to the conversation rather than just the search term.

• Should I allocate budget to ChatGPT advertising?

  If you're running significant Meta or Google spend, yes — but as a structured test, not a budget reallocation. Start with $500–$1,500 across 3–5 context hints targeting different conversation scenarios for your core offer. Run for 2–3 weeks at CPC floor bids. Track direct conversions via pixel/CAPI and branded search lift on Google as a proxy for indirect impact. Then run a one-week holdout (pause ChatGPT ads, keep everything else constant) to measure true incremental impact. If validated incremental ROAS is within 30% of your existing channel benchmarks, allocate 5–10% of digital budget. If it's 50%+ below, keep at test levels and revisit quarterly.

• How do you attribute conversions from AI chat ads?

  Direct click-to-conversion attribution works through ChatGPT's conversion pixel and Conversions API, similar to Meta's CAPI. But the bigger attribution challenge is indirect influence: users who see your ad, continue the conversation instead of clicking, and may convert days later from a different session or channel. Standard attribution frameworks weren't built for multi-turn, cross-session AI interactions. The most reliable approach is a holdout test — pause ChatGPT ads for a period while keeping other channels constant, then compare total conversion volume. Also track branded search lift on Google during active ChatGPT ad periods as a proxy for indirect influence.

• When should I use GEO vs paid ads on AI platforms?

  Use paid when you need immediate guaranteed visibility, your offer is transactional and time-sensitive, or you're testing whether AI platform intent works for your category. Use GEO (Generative Engine Optimization — making your content citable by AI systems) when you're building long-term brand authority, your budget is limited, or you're in a restricted ad category where paid placement isn't available. The best approach is both, sequenced: test with paid first to validate the channel and learn which conversation contexts matter, then invest in GEO for those same contexts to build organic citation equity over time.

Related reading: Build vs Buy vs Automate: A Decision Framework for 2026