ChatGPT Ads vs search ads: what changes when there is no keyword

The one difference that drives all the others

In search advertising you choose keywords, and an auction runs each time someone types a query that matches one. The keyword is the unit of buying, targeting and measurement.

ChatGPT Ads remove that unit. There is no keyword to bid on. OpenAI's system reads the intent and context of the current conversation and decides whether an ad fits, then which one. Everything below follows from that single change.

How the ad is matched

A search ad is matched to a string of words. A ChatGPT ad is matched to a described need.

OpenAI considers the context and intent of the current conversation, your landing page, your title and copy, your context hints and your targeting selections. When the user has personalization switched on, select signals from their broader ChatGPT use can join that mix. Those signals stay inside ChatGPT and never reach the advertiser.

The practical effect: copy that plainly states what you offer, who it is for and when it helps has more surface area to match than a short slogan. You are describing a fit, not naming a term.

How the auction works

Both models are auctions, but they weigh different things.

When several ads are eligible, OpenAI weighs relevance together with advertiser bids in what it describes as a relevance-weighted, second-price auction. Relevance and bid both count, and neither wins alone. You cannot buy your way past a poor match, and a great match still needs a bid to compete.

Buying is on impressions or valid clicks. Conversion-optimized campaigns still bill per click or per impression, never per conversion.

How the creative is built

Search ads assemble headlines and descriptions, often mixed and tested by the platform. ChatGPT ads are built from a fixed set of parts:

• Advertiser name and favicon
• Title and copy
• Landing page URL
• Image asset

There is no AI-generated creative. Ads Manager can prefill a draft from your website metadata, but OpenAI states it does not generate new copy or imagery. The text and image are yours to write and approve, which makes the wording and the landing page the two things you actually control.

Where the ad shows and who sees it

A ChatGPT ad can appear below the end of a response, labeled as sponsored and visually separated from the answer. It never changes the answer itself.

OpenAI shows ads only to users on the Free and Go plans. Paid tiers above Go do not see ads, and neither do accounts identified as belonging to people under 18. That is a different audience shape than search, where anyone running a query can see an ad.

How you measure it

Search gives you keyword-level detail. ChatGPT Ads reporting is aggregated: impressions, clicks, spend and conversions. Advertisers never receive chats, chat history, memories, names, emails, precise location or IP addresses.

So the measurement work moves to your own site. OpenAI's reporting shows the click; whether that click became a lead depends on your pixel, the Conversions API and preserving the click reference when the visitor lands. Treat that setup as part of launch, not an afterthought.

What to keep, what to unlearn

Three habits carry over, and three do not.

Carry over: bid discipline, a fast and honest landing page, and rigorous conversion tracking. These matter in both worlds.

Unlearn: keyword lists, exact-match thinking and search-term reports. In ChatGPT Ads the equivalent work is writing specific copy and precise context hints, then reading aggregated results rather than query strings.

The goal is the same as any lead channel: the right message reaching the right need at a price that pays back. The levers are what changed.