Key takeaways
- ChatGPT Ads went from zero to a $1 billion annualized run rate in under 200 days, and Promptwatch's data shows ads now appear in roughly 30% of ChatGPT web-search responses, up from essentially none in April 2026.
- Google has quietly made AI Mode and AI Overviews a real ad surface: ads sit alongside 25.5% of AI Overview responses, Shopping carousels now render inside AI Mode, and standard exact/phrase-match Search campaigns became eligible to serve in AI Mode in September 2026 without a Performance Max rebuild.
- Perplexity tried ads, got spooked by trust data, and pulled them in February 2026. Microsoft Copilot never built a separate ads manager at all, it just makes existing Microsoft Advertising campaigns eligible.
- US AI search ad spend is still tiny in absolute terms ($2.08 billion in 2026, 1.3% of total search ad spend) but is forecast to hit $25.93 billion by 2029, a roughly 10x jump in three years.
- Consumer trust is the wildcard: 63% of US adults say ads inside AI answers make them trust those answers less, which means the brands that win early will be the ones that treat these placements carefully, not the ones that spam them.
Why this matters right now, not next year
I'll be honest, when ChatGPT Ads launched quietly as a pilot in February 2026, a lot of marketers I talked to shrugged it off. Free-tier placements, no scale, no proof it would matter. Six weeks later OpenAI announced it had already hit $100 million in annualized revenue. By the end of August, that number was $1 billion. Tens of thousands of advertisers were running campaigns across 40-plus countries. That's not a pilot anymore, that's a channel.
The reason this moment matters is that the window to learn an advertising surface before it gets crowded and expensive is always short, and it's always shortest right after launch. Anyone who ran paid search in 2002 or paid social in 2008 knows the pattern: early CPMs are cheap because competition hasn't arrived, early algorithms are forgiving because there isn't enough data to punish mistakes, and early case studies get outsized attention because there aren't many yet. That window on AI search ads is open now. It will not stay open.
What's actually live across each platform
It helps to separate hype from what's shipped. Here's where things stand as of October 2026.
ChatGPT: the fastest-moving surface
OpenAI's ad rollout has been aggressive. Ads started appearing on Free and Go tier accounts in the US in February, and by late summer the ad rate had climbed in visible steps rather than a smooth curve, a burst in late June, then a jump past 30% on July 1 that stuck. Promptwatch's ChatGPT ads tracking puts the most recent 7-day average at 32.4%, up over 12% week over week, with a daily peak of nearly 44%.
What's notable is where those ads show up. About 73% appear on organic, non-branded prompts, roughly 18% on brand-specific prompts, and under 9% on competitor-comparison prompts. That means most of the ad inventory is being won on broad, top-of-funnel questions, not on your brand name or your competitor's name. If you're only watching branded queries, you're missing most of the action.
On October 5, OpenAI pushed further with a new visual ad format inside ChatGPT's image generation flow, plus a brand-suitability pilot with DoubleVerify and Integral Ad Science, which tells you OpenAI is trying to make this feel like a mature ad platform, not a side experiment. Pricing details floating around (roughly $60 CPM, a six-figure minimum spend for early self-serve access) are third-party estimates OpenAI hasn't officially confirmed, so treat those with some skepticism, but the direction is clear: this is becoming a real media buy with real measurement partners like Triple Whale, Northbeam, and LiveRamp plugged in.
Shopping is the other thread worth watching inside ChatGPT. Product cards and price comparisons still only show up in a low single-digit percentage of all responses, but that rate has doubled overnight before, then dropped back weeks later, which tells you OpenAI is actively tuning the feature rather than letting it run organically. On commercial and transactional prompts specifically, the trigger rate is far higher than the blended average suggests. If you sell physical products, getting your Product schema, pricing, and merchant feed clean now is the same move that worked for brands who got ahead of Google Shopping in its early days.
Google: AI Mode just became a real ad surface
Google moved slower in public messaging but just as decisively in product. Ads alongside AI Overviews went from 5.17% of responses in early 2025 to 25.5% by this analysis, nearly a 5x jump in about a year. In February, Google launched shopping ads with "Direct Offers" inside AI Mode, and by August 31 it confirmed Shopping ads now render in a carousel format directly inside AI Mode responses, a first for the platform.
The bigger structural change came in September. On September 1, Google began automatically migrating eligible Search campaigns to AI Max for Search, reporting a 7% average conversion lift, climbing to 27% for campaigns that had been stuck relying on exact or phrase match keywords. Then on September 9, Google's Ads Liaison confirmed that standard exact and phrase match keyword campaigns are now eligible to serve text ads directly in AI Mode for direct-intent queries, without requiring an AI Max or Performance Max rebuild. Conversational and exploratory queries still need AI Max or PMax.
There's a catch worth flagging loudly: a late-September analysis found only 1.28% overlap between the products shown in the standard Google Shopping carousel and the products surfaced in AI Mode for the same query on the same day. AI Mode shows roughly 3.9 products per query versus 27.8 in standard Shopping, and AI Mode prices run about 21.6% higher on average. That's a near-total disconnect. If you only manage one Merchant Center feed and one bidding strategy, you may be functionally invisible in AI Mode even while looking healthy in regular Shopping results.
Perplexity and Microsoft: the quieter paths
Perplexity tested sponsored follow-up questions (in a Related Questions section that reportedly drives 40% of all platform queries) through 2024 and 2025, then pulled all ads in February 2026 citing user trust. It's now leaning into being the ad-free alternative, chasing $500 million in annualized subscription revenue off 780 million-plus monthly queries. Worth remembering if you're betting heavily on one platform's monetization model staying put.
Microsoft never built a separate Copilot ads product. Existing Microsoft Advertising campaigns simply become eligible to appear inside Copilot responses using your existing text, image, and product assets, no way to guarantee placement in a given conversation, and no published universal CPC or CPM for the slot. It's less a new channel than an extension of one you may already be running.
Quick comparison
| Platform | Ad status in Oct 2026 | Scale signal | How you get in |
|---|---|---|---|
| ChatGPT | Live, expanding fast | ~30% of responses show ads; $1B annualized run rate | Self-serve Ads Manager, product feeds, custom audiences |
| Google AI Mode / AI Overviews | Live, structurally integrated | Ads beside 25.5% of AI Overview responses | Exact/phrase match Search (direct intent), AI Max/PMax (conversational) |
| Microsoft Copilot | Live, passive | No separate buy, rides existing campaigns | Standard Microsoft Advertising eligibility |
| Perplexity | Pulled, Feb 2026 | Positioning as ad-free | Not available |
The trust problem nobody's solved yet
Here's the part that should temper any rush to dump budget into these placements. Ipsos surveyed 1,085 US adults in February 2026 and found 63% say ads inside AI search results make them trust those results less, 27% strongly agree. That's a meaningfully different reaction than ads in a traditional SERP, where people have decades of learned tolerance for a labeled ad block. AI answers are read as advice. Advice that's secretly for sale reads as a betrayal, even when it's disclosed.
This isn't a reason to avoid the channel. It's a reason to be careful about creative and placement. OpenAI's own case data from October shows WeightWatchers got a 15.3% lower attributed CPA on ChatGPT Ads than its blended paid-search benchmark, and wellness brand Dose saw 67% of incremental purchases come from net-new customers. Those are real results. But they came from brands that treated the placement as a genuine answer to a genuine question, not an interruption. The brands that get burned here will be the ones that try to game conversational trust the way banner ads gamed display inventory fifteen years ago.
Why ad spend and AI visibility are now the same conversation
A year ago, paid search and GEO (generative engine optimization) were separate budget lines run by separate teams. That line is blurring fast, and for a practical reason: whether your ad even gets a chance to show depends increasingly on the same signals that determine whether you get cited organically. Google's AI Max rewards structured data, strong assets, and clean conversion signals, the exact inputs that also drive organic AI citations. If your product pages are a mess, you're losing on both fronts simultaneously.
This is where it's worth watching citation data closely rather than guessing. Promptwatch's August 2026 analysis of ChatGPT's citation mix found domains in the DR 46-75 range account for nearly half of all citations, while the very top tier, DR 91-100, dropped from about 7% to roughly 3% of citations in a single week. That's a real shift in who gets surfaced, and it happened on the same day Reddit's share of ChatGPT citations collapsed from about 4% to 0.5%, which Promptwatch documents in detail in its "Reddit citations are dropping in ChatGPT" report. Platform-level algorithm changes are moving fast enough that a snapshot from three months ago can already be stale.
If you're serious about understanding where your brand actually shows up, in paid placements and in organic AI answers, tools built specifically for AI visibility monitoring are worth adding to your stack rather than trying to eyeball it manually. Promptwatch tracks ChatGPT, Gemini, Claude, Perplexity, Grok, Copilot, and Google AI Overviews and AI Mode in one place, including a dedicated Ads Radar that shows which competitors are winning ad slots on your tracked prompts and a Shopping tracker for product recommendation placements.
Other tools worth knowing in this space
The AI visibility tooling market has gotten crowded, and most of them are monitoring-only, they'll tell you where you stand but won't help you fix it. A few worth knowing by category:
| Tool | Focus | Notable detail |
|---|---|---|
| [tool:profound] | Multi-engine monitoring | Starts at $99/mo, ChatGPT-only on entry tier |
| [tool:peec-ai] | Multi-language tracking | $95-495/mo, cut annual pricing 16% in 2026 |
| [tool:athenahq] | Citation engine benchmarking | Free credits tier, reported 45% net gain in answer share in a 30-day test |
| [tool:otterly-ai] | Budget multi-engine tracking | Includes Microsoft Copilot from $29/mo entry tier |
| [tool:ahrefs-brand-radar] | Add-on to existing SEO suite | $398-699/mo add-on, needs base Ahrefs plan |
For a broader view of where these platforms stack up against each other on crawler logs, content generation, and agentic optimization rather than just tracking, the GEO software directory at bestgeosoftware.com is a useful starting point.
A practical playbook for the next two quarters
1. Stop treating AI ad spend as a rounding error, but don't overcommit either
At $2.08 billion, US AI search ad spend is still under 1.5% of total search ad spend. It's not going to replace your Google Search budget this year. But the forecast trajectory to $25.93 billion by 2029 means the brands building creative, measurement, and feed infrastructure now will have a multi-year head start over anyone waiting for the channel to "prove itself" first.
2. Fix your feed before you fix your bids
Whether it's ChatGPT Shopping or Google AI Mode's carousel, the gating factor is data quality: accurate pricing, availability, Product schema, and a merchant feed that doesn't lie. Google's own data shows only 1.28% overlap between standard Shopping and AI Mode shopping results for the same query, which means a feed optimized only for classic Shopping is quietly failing in the newer surface.
3. Build reporting that separates AI-influenced traffic from everything else
Push offline conversions, CRM deals, and store visits into your ad platforms so smart bidding optimizes for actual profit instead of vanity clicks. Compare performance on queries where an AI Overview or AI Mode response appears against queries where it doesn't, that gap tells you where AI is cannibalizing your paid clicks and where it's opening new ones.
4. Watch the trust signal, not just the CPC
Ads inside AI answers carry reputational risk that traditional SERP ads don't. If 63% of consumers trust AI answers less once ads appear, the brands that win long-term will be disciplined about relevance over frequency. Bidding aggressively on every broad prompt just to appear more often is a short-term tactic with a long-term cost.
5. Treat GEO and paid AI search as one strategy, not two budgets
The signals that get you cited organically (structured data, clean entity information, authoritative content) overlap heavily with the signals that get your ads eligible to show in AI Max, AI Mode, and ChatGPT's commercial surfaces. Agencies that can run both sides together, rather than siloed PPC and SEO teams working from different playbooks, are going to outpace competitors who keep the two disciplines separate. If you need outside help pulling that together, 1001 SEO Media runs technical SEO, content, and GEO work as one coordinated strategy rather than bolting AI tracking onto a traditional SEO retainer.
Where this is headed
The honest read is that nobody, including OpenAI, has fully figured out the economics here yet. EMarketer's Nate Elliott called OpenAI's $1 billion run rate "terribly disappointing" against its internal $2.5 billion 2026 target, which tells you even the companies building these ad products are still calibrating expectations against reality. Perplexity walked away entirely. Google is iterating in near-weekly product announcements. That volatility is exactly why marketers who build the measurement and creative muscle now, while the playbook is still being written, end up with more leverage than the ones who wait for a settled set of best practices that may not arrive for another year or two.