Key takeaways
- ChatGPT now shows ads in roughly 20-30% of web-search responses (with daily spikes above 40%), while Google serves ads both above/below AI Overviews (all 200+ markets) and, in a smaller set of countries, inside the AI Overview itself.
- The two ad systems behave nothing alike: ChatGPT typically shows one sponsored card per answer, Google's AI surfaces behave more like a carousel, averaging over three ad items per ad-bearing response.
- Neither OpenAI's Ads Manager nor Google Ads' own reporting gives you competitive visibility. You only see your own spend and impressions, not who else is bidding on the same prompts.
- There's no sanctioned ads-reporting API for either surface. Every monitoring approach, DIY or vendor-built, relies on direct UI capture of what real users would see.
- A workable dashboard needs three layers: a capture layer (prompt panels or scraping), a normalization layer (so ChatGPT and Google data share a schema), and a visualization layer (Looker Studio, a BI tool, or a dedicated AI-visibility platform).
Why this suddenly matters
A year ago, "ads in AI answers" was a hypothetical line in an OpenAI blog post. It isn't anymore. ChatGPT turned on ad testing for US Free and Go users on February 9, 2026, and by August had expanded to the UK, Mexico, Brazil, Japan, South Korea, most of Southeast Asia, Taiwan, and India. Google has run ads above and below AI Overviews since the feature launched, and has been quietly expanding ads inside the overview text itself across a dozen markets including the US, Canada, Australia, and India.
This is a different ad buy than a Search Ads campaign. The matching logic is different, the placement density is different, and the parts of your market that get blocked from advertising entirely are different. If you're running paid search and you're not watching both surfaces, you're missing a channel that's already live and already shifting share away from the ads you understand.
I'll be honest: the first time I looked into this, I assumed it would be a minor variant of normal SERP tracking. It isn't. ChatGPT ads and Google AI Overview ads are run by two companies with opposite instincts about density, transparency, and which verticals get to participate at all. A dashboard that treats them as the same thing will mislead you.
How the two ad systems actually differ
Google's mechanics are documented, if buried. Ads above and below AI Overviews run through the standard auction and are eligible in every market where AI Overviews exist, over 200 of them. Ads inside the AI Overview are far more restricted: English only, and limited to Australia, Canada, India, Indonesia, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, Philippines, Singapore, and the US. Only Text and Shopping ads from Search, Shopping, Performance Max, or Dynamic Search campaigns qualify, exact-match-only campaigns don't. Advertisers can't target the AIO placement directly, and they can't opt out of it if they're running an eligible campaign type.
ChatGPT works on a completely different model. There's no keyword list. OpenAI matches ads using "context hints" advertisers write themselves, checked against the conversation topic, prior chats if personalization is on, and the landing page content, then ranks eligible ads by relevance and bid (CPM or CPC). Ads appear as a clearly labeled "Sponsored" card below the end of the response, never inside it, and never in the Atlas browser during this test phase. No ads are shown to under-18 accounts or anywhere near health, mental health, or political topics, and political ads are banned outright.
The practical gap shows up in density. According to Adthena's analysis of roughly 850,000 AI search queries between March and May 2026, ChatGPT and Google's AI surfaces served ads at almost the same overall rate (4.47% vs 4.14% of queries), but the experience looks totally different once an ad shows up. ChatGPT averages about 1.06 ad items per ad-bearing response, essentially one ad, no carousel. Google's AI Overviews and AI Mode average 3.53 items per ad-bearing answer, room for several brands to appear at once.
Verticals get treated unevenly too. ChatGPT returned zero ads in Legal, Pharma, Banking, and Nonprofit queries in that same dataset, and near-zero in Healthcare (0.46%). Google's AI surfaces still served ads in some of those same categories in the US, including Legal at 5.62%. That's a deliberate OpenAI policy choice, not noise in the sample.
| Dimension | ChatGPT Ads | Google AI Overviews / AI Mode Ads |
|---|---|---|
| Ad density per response | ~1.06 items (single card) | ~3.53 items (carousel-like) |
| Overall ad rate (Mar-May 2026, Adthena) | 4.47% of queries | 4.14% of queries |
| Geographic reach | US, Canada, Australia, NZ, UK, Mexico, Brazil, Japan, South Korea, SE Asia, Taiwan, India (rolling out) | Ads above/below in 200+ markets; ads inside AIO limited to ~12 countries, English only |
| Matching mechanism | Context hints + conversation topic + bid, no keyword list | Standard auction (above/below); keywordless matching (broad match, PMax, Shopping, DSA) for inside-AIO |
| Blocked verticals | Legal, Pharma, Banking, Nonprofit (near-zero), no political ads at all | Fewer categorical blocks for above/below; sensitive categories excluded from inside-AIO only |
| Native reporting visibility | Own impressions/clicks/CPC only, via Ads Manager | Own Top Ads reporting; no segmented reporting for inside-AIO ads yet |
Promptwatch's own tracking of ChatGPT ad frequency over the trailing 90 days (May 20-Aug 17, 2026) backs up how volatile this still is: ChatGPT served zero ads until May 27, then climbed in bursts to a daily high of 43.9%, with the 30-day average sitting around 29.6% and rising roughly 22.6% month over month. That kind of swing means a single-day snapshot of "ad presence" is close to useless. You need a rolling window.
The blind spot neither vendor will fix for you
Here's the part that actually justifies building a dashboard instead of just checking your own account dashboards. Adthena's research flags it bluntly: native ad managers, both OpenAI's and Google's, only show your own spend, impressions, clicks, CPC, and CTR. Neither gives you whole-market competitive visibility: who else is bidding on your prompts, which creative is winning the slot, how their messaging is positioned.
And there's no API shortcut. As of this writing, neither OpenAI nor Google exposes "ads in AI Overviews" or "ChatGPT ads" as a labeled, queryable metric to third parties. Every monitoring tool in this space, including the enterprise ones, is built on direct UI capture, either synthetic query panels or real-user panels, not a sanctioned reporting feed. That's worth internalizing before you build anything: you're not integrating with an API, you're replicating what a human would see by running prompts and reading the response.
What to actually put in the dashboard
Think of the build in three layers.
1. Capture layer
This is where you run (or buy access to) a panel of prompts across both platforms, on a schedule, and log the raw response including whether an ad appeared, which advertiser it named, the landing URL, and any creative text. For ChatGPT, you need to restrict this to regions where ads are actually live, running a prompt panel against US traffic when ads haven't launched there yet will just produce zeros and waste your sampling budget. For Google, remember the above/below placement is live everywhere AIO exists, but the inside-AIO placement only fires in about a dozen English-language markets, so segment your capture by placement type, not just by platform.
2. Normalization layer
ChatGPT and Google don't return comparable units. ChatGPT gives you a single sponsored card with its own ad id, context hint, and bid type (CPM/CPC); Google gives you ad slot position (above AIO, below AIO, or inside AIO) plus the standard Ads auction fields. You'll want a shared schema along the lines of: platform, placement type, query, timestamp, advertiser domain, ad copy, position, vertical. This is the layer most DIY builds skip, and it's the reason side-by-side comparisons usually end up as two separate spreadsheets instead of one dashboard.
3. Visualization layer
Looker Studio remains the default free option for blending data sources into one report, and it's the approach most practitioners reach for when there's no native combined dashboard from either vendor, pulling platform data through connectors (Windsor.ai, Coupler.io, and similar tools are common for this) and layering it against GA or CRM data to see correlation with actual conversions. If you'd rather not assemble this yourself, a few vendors have already built the capture and normalization layers for you.
Tools that already do some of this
| Tool | Covers ChatGPT ads | Covers Google AI Overview/Mode ads | Competitive visibility | Notes |
|---|---|---|---|---|
| Native Ads Manager (OpenAI) / Google Ads | Own data only | Own data only | No | Free, but zero competitor view |
| Similarweb AI Ads | Yes | Yes | Yes | Built on real-user panel data; enterprise/contact-sales pricing |
| Adthena ChatGPT Ads Intelligence | Yes (300k+ daily prompts) | Separate "AI Overview Impact" product | Yes | Strong on creative analysis (tone, CTA usage, headline length) |
| Finseo | Yes | Yes | Yes | Logs advertiser, landing URL, and win frequency per tracked prompt |
If your goal is narrower, specifically watching how your own organic AI visibility holds up as paid placements eat into the same answers, that's a slightly different job than ad-tracking, and it's where a platform like Promptwatch fits. Promptwatch isn't built as a pure ads tracker, but its ChatGPT Ads and ChatGPT Shopping monitoring sits next to its citation and crawler data, so you can see in one place whether a sponsored card showing up on a prompt is correlated with your organic citation share dropping on that same prompt, something a standalone ads tool can't show you because it has no visibility into your organic performance at all.
That combination matters more than it sounds. An ad appearing above or below an AI Overview doesn't touch the organic answer, but an ad appearing inside an AI Overview, or a sponsored card pushing your organic ChatGPT citation further down the response, is a real visibility cost. Promptwatch's own data on ChatGPT ad frequency shows this isn't a rare event anymore: ads are now appearing in roughly a third of web-search responses on a rolling 30-day basis, up sharply from where it started in late May. Pairing that trend line with your own citation data tells you whether paid placements are actually displacing your brand or just running alongside it.
Metrics worth putting on the dashboard
Don't just track "did an ad appear." Track these, side by side, per platform:
- Ad appearance rate, as a 7-day and 30-day rolling average, not a daily snapshot (daily rates have swung from under 2% to nearly 44% for ChatGPT in a single quarter)
- Ad density per response (single card vs. carousel count)
- Advertiser identity and landing URL, so you can see who's actually bidding against you
- Prompt-type mix (organic, brand-specific, competitor-comparison), since ad behavior differs meaningfully across these
- Vertical-level blocking, especially if you're in Legal, Pharma, Banking, Healthcare, or Nonprofit, where ChatGPT currently shows near-zero ads but Google's AI surfaces still do
- Geographic coverage, tracked against the live rollout list for each platform, so you're not drawing conclusions from a market where ads haven't launched yet
A few pitfalls to avoid
First, don't trust a single day of data. The volatility in ChatGPT's rollout alone, zero ads one week, 44% the next, means any dashboard built on daily snapshots will produce whiplash numbers that don't mean anything. Use rolling windows.
Second, don't assume parity across markets. A dashboard built on US data will overstate how far along Google's inside-AIO ads are (still English-only, ~12 countries) and may understate how far ChatGPT ads have spread (now live across a much longer country list than most people realize, including India and most of Southeast Asia).
Third, treat third-party growth stats with some skepticism unless they're sourced directly. One widely cited claim has ads alongside AI Overview SERPs growing from roughly 3% in January 2025 to 40% in November 2025, then settling back to about 25.5% by March 2026. That's a Semrush-sourced, third-party figure, directionally useful but not independently verified, and worth labeling as such on any dashboard you share internally.
If you want a broader view of which GEO and AI-visibility platforms handle this kind of monitoring well, the directory at bestgeosoftware.com is a reasonable place to compare options beyond the ones named here. Building this dashboard yourself is absolutely doable with Looker Studio and a disciplined capture schedule. The harder part isn't the chart, it's making sure you're sampling the right markets, the right verticals, and enough days to see past the noise.