ChatGPT Ads vs Google Ads: what actually transfers
A practical comparison for advertisers who know search: targeting, pricing, creative, measurement — what carries over from Google Ads and what will actively mislead you.
If you’ve run Google Ads, you have instincts. Some of them will serve you well inside ChatGPT, and some will quietly cost you money. Here’s the sorting, column by column.
Targeting: the big rewrite
Google: keywords plus match types. You bid on compressed queries and the machine expands them.
ChatGPT: context hints — natural-language descriptions of conversations. No match types, no expansion settings. The compression instinct is the one to unlearn: “plumber near me” is a fine keyword and a weak hint. The situation is the targeting: “water coming through the ceiling” beats “plumber” every time.
What transfers: intent thinking. The person mid-emergency, the person comparing quotes and the person who knows your name are still three different campaigns’ worth of psychology. What changes is that you now write it out in sentences instead of triangulating it with match types and negatives.
Price: higher floor, less competition
Google’s auctions are two decades deep and priced accordingly, niche by niche. ChatGPT’s auction is young but not cheap — the platform’s suggested CPC sits around $3, and lowball bids don’t serve at all rather than serving cheaply. Fewer advertisers are competing, but the floor is real.
The honest comparison isn’t CPC against CPC anyway — it’s moment against moment. A Google click catches someone searching. A ChatGPT card catches someone describing their problem to an assistant they trust, often before they’ve searched for anything. Whether that earlier, warmer moment is worth $3 for your business is exactly what a first campaign tells you.
Creative: one unit, thankfully
Google wants responsive search ads, sitelinks, callouts, image extensions and a stack of headlines to rotate. ChatGPT has one unit: image, title, short body, link. Constraints worth knowing — the rendered card truncates titles around 35 characters (well before the API’s 50 limit), images need 256 × 256 minimum, and text baked into images is a review rejection waiting to happen.
A/B testing exists only if you build it: run two or three cards per ad group and read the difference yourself. There’s no rotation optimizer yet.
Measurement: bring your own discipline
Google spoils you with conversion columns. On ChatGPT, conversion tracking is still gated and early for most accounts — so click-through and post-click behavior carry the weight for now. Two practical consequences:
- Tag everything. Untagged ChatGPT traffic lands in analytics as “direct,” which makes the channel look like it does nothing. UTM parameters on every landing URL, from day one, no exceptions.
- Judge on a week, not a day. Volume is thinner than search. Day-two panic reads are how new advertisers convince themselves a working channel is broken.
Where the leverage is
On Google, most spend competes against category veterans with polished funnels and a decade of Quality Score. ChatGPT advertising in 2026 looks more like search did early: the mechanics are learnable, most competitors haven’t shown up, and craft — good hints, honest cards, patient reading of small numbers — still moves the needle disproportionately.
That’s the actual argument for testing now, and it expires. We built AthillyAds to compress the craft part: it writes intent-grouped hints from your website, builds the cards, checks what review rejects, and coaches your first week. The first campaign is free, so the cost of the test is a lunch break.