Key takeaways
- Google AI Mode ads run through campaigns you probably already have (Performance Max, Shopping, AI Max, broad-match Search). ChatGPT ads are bought separately, in OpenAI's Ads Manager or through partners.
- Google places ads above, below or inside the AI answer. ChatGPT shows a labeled sponsored unit below the response, separate from the answer.
- Google matches on query and answer content. ChatGPT has no keywords: it weighs conversation context, advertiser hints, bids and, if personalization is on, signals from wider ChatGPT use.
- Reporting is the weak spot on both. Google blends AI placements into search totals. OpenAI gives aggregated views and clicks only.
- Promptwatch's data shows ads in about 20% of ChatGPT web-search responses over a 90-day window, so this is no longer a rounding error.
Why compare these two at all
For twenty years a search ad sat beside the answer. In 2026 it moved inside it, or at least right underneath. Google runs sponsored placements in AI Mode. OpenAI opened ads in ChatGPT, started testing in the US on February 9, 2026, and added a self-serve Ads Manager in May. Perplexity tried ads and quit in February 2026, so it is not part of this comparison.
The two products look similar from the user's chair: an AI answer with something labeled sponsored nearby. From the advertiser's chair they are quite different. One is an extension of a mature auction. The other is a young platform with its own rules, its own buying flow and a smaller toolset. I'll walk through the differences that change how you plan, and flag where the numbers come from vendors selling the surface.
The quick comparison
| Dimension | Google AI Mode ads | ChatGPT ads |
|---|---|---|
| How you buy | Existing Google Ads campaigns, no separate AI Mode manager | OpenAI Ads Manager (ads.openai.com) or agency and ad tech partners |
| Eligible setups | Performance Max, Shopping, AI Max for Search, Search with broad match | Ads Manager campaigns, including a multi-image format in image generation |
| Placement | Above, below or within the AI answer | One ad unit below the response, visually separate |
| Matching | Relevance to the query and the AI response | Conversation intent, landing page, advertiser hints, bids, optional personalization |
| Keywords | Yes, via your existing campaign logic | No keywords |
| New formats | Conversational Discovery, Highlighted Answers | Visual ad format, auto-generated variants |
| Who sees ads | Search users in supported markets | Free and Go plans (not Plus, Pro, Business, Enterprise, Edu) |
| Reporting | Blended into search totals | Aggregated views and clicks only |
| Pricing model | CPC in the existing auction | CPM and CPC bidding (beta self-serve) |
Pricing is where sources disagree. Early launch coverage cited roughly $60 CPMs and a $200,000 minimum commitment. OpenAI's May 2026 buying update added self-serve Ads Manager and CPC bidding alongside CPM, so treat the old pilot numbers as history, not current terms. Confirm with OpenAI or a partner before you forecast.
How Google AI Mode ads work
There is no AI Mode ads manager and no opt-in checkbox. Google confirmed in 2025 that advertisers running Performance Max, Shopping and Search campaigns with broad match, including AI Max for Search, are eligible to appear in AI Overviews and AI Mode. If you run those campaign types, you may already be serving there without knowing it.
At Google Marketing Live on May 20, 2026, Google announced two new formats built with Gemini:
- Conversational Discovery ads, where Gemini writes creative tailored to the specific question the user asked.
- Highlighted Answers, where a qualifying ad can appear inside a list of recommendations the AI generates.
Both carry an independent AI explainer next to the advertiser's creative, and both are labeled Sponsored. Google's own pitch is that 75% of people report making faster, more confident decisions using AI Mode. That is Google describing its own product, so read it as marketing.

The matching catch on Google
The ad has to be relevant to both the query and the content of the AI response. Ads can appear in one position or the other, never both at once. In practice a tightly managed exact-match account can be nearly invisible inside AI answers, while a broad-match or AI Max account shows up constantly. If you want AI Mode inventory, your keyword strategy has to loosen up first.
How ChatGPT ads work
OpenAI's help documentation describes a single ad unit below the response, which can hold one image or several. Ads are labeled and shown separately from the answer. A multi-image visual format is also being tested during image generation.
Selection is not keyword-based. OpenAI says the system weighs:
- the conversation's context and intent
- the ad's landing page, title and copy
- advertiser-provided context hints and targeting selections
- when ads personalization is on, select signals from the user's broader ChatGPT use
Ranking combines relevance and advertiser bids. Personalization is not initially available in the EEA or Switzerland.
Who actually sees the ads
This matters more than most media plans admit. Free and Go plans see ads. Plus, Pro, Business, Enterprise and Edu accounts do not. Under-18 accounts and Temporary Chats do not show ads either, and there are no ads near personal health, mental health or politics. Political advertising is not allowed at all. If your buyers are the kind of people who pay for ChatGPT, this placement may miss them.
How often ads appear
Promptwatch Data tracked this in its ChatGPT ads over time report (May 20 to August 17, 2026). Ads appeared in an average of 20.1% of ChatGPT responses where ChatGPT searched the web, returned at least one citation and completed ad extraction. None appeared until May 27, when the rate hit 1.69%. It passed 30% on July 1 and has reached daily highs of 43.9%. The past 7-day average at the time of the report was 32.4%.
A caveat from the same report: the mix of ad-bearing responses (73.3% organic prompts, 18.0% brand-specific, 8.6% competitor comparison) reflects what was in the monitor, not how likely each prompt type is to trigger an ad. Also note the timing difference. OpenAI says testing began February 9, while Promptwatch only measured ads in web-search responses from late May.
Placement and the user experience
Google's new formats sit inside the response. A Highlighted Answers ad lives in an AI-written list, so the line between recommendation and ad is thinner. ChatGPT keeps the sponsored card below the answer. Both label ads, but the trust dynamics differ. The comment sections under coverage of this topic are full of users worried about paid answers mixing with organic ones, and one source cites 63% of consumers saying ads reduce trust. I would not lean on that figure too hard, because it comes from a vendor blog, but the instinct is real.
OpenAI states that ads run on separate systems from the chat model, that advertisers cannot shape or rank answers, and that ads do not influence answers. Users can turn off personalization, dismiss ads, see why an ad was shown and clear ad data in Settings. With personalization off, ads still use the current chat's context.
Creative control and AI-written copy
In both cases the model writes or frames part of the ad. On Google, Conversational Discovery generates creative tailored to the question, and Google has floated an AI Brief as a way to set brand guardrails. Agencies at the event were open that nobody yet knows how well those guardrails hold.
OpenAI's October 2026 updates go further. The system can write new headlines and descriptions, resize images, translate copy and send the click to a different page on your site. OpenAI says plainly that these versions can run automatically without you reviewing each one. If you work in a regulated category, that sentence should make you pause. Get compliance sign-off on the specific formats, and document what the AI-generated explainer can say next to your name.
Reporting and measurement
Google does not break out AI Overviews or AI Mode as a placement in standard reporting. Impressions blend into your existing search totals. OpenAI gives advertisers aggregate, non-identifying data such as views and clicks, which it frames as a privacy feature. It also limits what you learn about why an ad served.
One agency that ran a live ChatGPT campaign reported a gap between OpenAI's numbers and Google Analytics large enough to pause the test. That is one agency, but it matches what I'd expect from a channel with no automatic tagging. OpenAI tells advertisers to add UTM parameters to each creative by hand.
Some practical measurement habits:
- Tag every ChatGPT creative with UTMs, and use a dedicated landing page or promo code as cheap insurance.
- Track referrals from chatgpt.com, copilot.microsoft.com and Google AI Mode as their own channel, so a lift from paid is not mistaken for a lift in free citations.
- Compare platforms on the same business outcome (orders or qualified pipeline), not on clicks. A hypothetical $10 CPM with a 1% CTR implies $1 per click, and a 2% conversion rate on top of that implies $50 per conversion. Those are arithmetic examples, not benchmarks.
- Since October 2026, third-party attribution and incrementality partners can connect to ChatGPT ads, which helps. Google still has a long head start on modelled conversions.
The tools below cover the attribution side if native reporting is not enough.

Intent: demand capture versus demand shaping
Someone typing "waterproof jacket buy online" has decided. Google is built to be in front of that person. Someone asking an assistant what to wear hiking in Scotland in October has decided nothing, and will ask three follow-ups before reaching a shortlist. That is demand shaping, and it takes longer to show up as revenue.
Do not over-literalize this. A short Google search can be exploratory and a long conversation can be close to purchase. Write the same customer problem both ways, a short query and a conversational request, use the differences to shape your message, then test whether they change qualified demand. For most ecommerce brands the sensible order is Google first and fully optimized, then ChatGPT as a capped experiment.
Landing pages and crawlability
Both platforms reward a landing page that finishes the thought. OpenAI tells advertisers to use product or content pages rather than a homepage. It also requires that landing pages not block its OAI-AdsBot and OAI-SearchBot user agents. Google's Gemini-built formats pull context from your site and feed to write the explainer, so a site that shuts crawlers out risks ineligible ads or a thin explainer.
Check your robots.txt and any bot blocker before you spend a cent. Promptwatch can show which AI crawlers are actually hitting your site and where they fail, through its AI crawler logs.

For ecommerce, product data does the heavy lifting on both channels. Titles that describe the product, more than one image, and stock levels that are true right now matter more than any bidding tweak.
Paid placement is only half of AI visibility
Paid slots are a small slice of how brands show up in AI answers. Organic citations still drive most of the exposure. Promptwatch Data tracks which sources ChatGPT and Google's AI surfaces cite, and the picture keeps shifting: in its social media citations report, ChatGPT drew 0.46% of citations from social platforms between Sep 28 and Oct 4, 2026, while Google AI Overviews drew 12.2%. A brand that buys ads but never gets cited is paying to rent what competitors earn.
That is why teams pair ad tests with visibility tracking. You want to know whether a paid test moved your mention rate or just your spend. If you are comparing options, the GEO software directory at bestgeosoftware.com lists the main platforms. Promptwatch is the one that goes beyond monitoring: it traces citations, flags content gaps, and its agents write and publish the pages AI cites. Tools like Profound and Peec AI are mostly dashboards by comparison.
A simple decision framework
| If you are... | Start with | Why |
|---|---|---|
| Already running PMax, Shopping or AI Max | Audit what is serving in AI Mode | You are probably eligible already |
| An ecommerce brand with a clean feed | Google first, small ChatGPT test | Feed quality drives both, Google has better attribution |
| A brand whose buyers pay for ChatGPT | Weight ChatGPT lower | Paid tiers do not see ads |
| In health, finance or another regulated field | Slow down, get legal review | Auto-generated variants and AI explainers need sign-off |
| Lacking attribution infrastructure | Fix tracking before buying | Native ChatGPT reporting is thin |
What to do this quarter
- Audit which Google campaigns can serve in AI Mode and loosen match types only where margins allow.
- Open an OpenAI Ads Manager account or talk to a partner, and cap the test budget at an amount you can afford to learn from.
- Hand-tag every creative, build a dedicated landing page per test, and unblock OAI-AdsBot and OAI-SearchBot.
- Run at least a few weeks before judging, since daily ad rates in ChatGPT swing widely.
- Track organic AI visibility alongside paid so you can separate the two.
One more thing, and it is a bit uncomfortable: the cheapest improvement is still rewriting your pages for a reader who just finished a paragraph of AI answer, not one scanning a results page. That helps whether the click arrives through a paid card or a citation.
If you want help planning across paid and organic AI search, 1001 SEO Media works on both the GEO side and the classic SEO side.



