ChatGPT ads vs Google Ads
Not a takedown of either. A straight comparison, so you can decide which budget belongs where.
Google Ads buys keywords; ChatGPT ads buy conversations described in plain language. Google has deeper reporting, mature attribution and far more inventory, while ChatGPT has cheaper clicks in most categories, roughly double the conversion rate on assistant referrals, and almost no competition — which is the entire reason to be there now.
The side-by-side
| ChatGPT ads | Google Ads | |
|---|---|---|
| Targeting | Context hints in plain language | Keywords, audiences, demographics, remarketing |
| Negatives | None. You tighten the hint instead | Negative keywords and lists |
| Auction | Relevance-weighted, second price | Ad Rank, Quality Score, second price |
| Bidding | CPC, CPM, oCPC, Maximize results | The full automated suite |
| Typical CPC | $3–5, $8–18 in software and finance | Often higher in the same B2B categories |
| Formats | Chat card, product feed, multi-product carousel | Search, Shopping, Display, Video, Demand Gen |
| Reporting | Thin but improving. One-day view-through | Deep, mature, sometimes too much |
| Geo | Country level, plus platform | Radius, postcode, everything |
| Competition | Thin, and closing | Fully priced |
What transfers from your Google playbook?
More than people expect, once you stop looking for the keyword box.
- Account structure discipline. Tight ad groups mattered on Google and matter more here, because reporting is thinner
- Landing page relevance. Google reads it for Quality Score; OpenAI reads it for matching
- Measurement rigour. Server-side conversions, clean events, honest attribution windows
- Feed hygiene. If your Shopping feed is good, product feed campaigns are a short hop
- Incrementality thinking. Same discipline, newer channel
What does not transfer?
- Keyword research. There are no keywords. Volume tools tell you nothing about which conversations exist
- Negative keyword hygiene. No such control. Precision comes from writing a narrower hint
- Audience layering. No demographics, no lookalikes, no remarketing lists, no cookies
- Bid automation instincts. Maximize results has no CPA or ROAS guarantee, so handing it a cold account behaves differently from tCPA on a mature one
- Attribution habits. View-through is one day, and pixel-only counting under-reports
So which should a budget go into?
Both, and not by moving money out of what works.
Google is where demand that already knows the words goes. ChatGPT is where demand that is still forming the question goes — comparing, narrowing, deciding. Those are different moments in the same purchase, and the assistant one arrives earlier.
The practical answer for most companies: keep search running and take a $3,000–10,000 monthly test into ChatGPT while the clearing price is what it is today.
Book a 20-minute call. We will show you what a click costs in your category and what your current search CPA would have to be for this to beat it.
Get my ads runningWhat about Copilot and Google AI Mode?
Both are already buyable, and both go through accounts you probably have.
- Microsoft Copilot — served through existing Microsoft Advertising campaigns. Low friction, low extra work
- Google AI Mode and AI Overviews — ads appear in AI results and are bought through Google Ads
- Claude and Perplexity — deliberately ad-free. The only route there is being recommended organically
That fragmentation is the actual state of the category. Nobody’s native dashboard covers it, which is why we built one view across all of them.
Last verified 23 Aug 2026 against OpenAI’s published documentation and our own live account. This channel changes monthly, so we re-check every figure on this page and restamp the date.
Sources: OpenAI Ads Manager documentation; third-party CPC surveys, 2026; Similarweb and Criteo referral conversion data; and our own live account.