What do ChatGPT ads cost, and what drives the price per click?

ChatGPT ads are bought through an auction. You can bid on CPM or CPC, and you can let it optimize toward a conversion goal. OpenAI publishes no fixed rates. How the channel itself works is in advertising in ChatGPT. As a guide, OpenAI recommends a maximum bid of $3 to $5 per click for manual CPC bidding; for CPM campaigns OpenAI names no default amount. The bidding models and those guide amounts are in Ads in ChatGPT: the basics in OpenAI’s help center.

Sources for the dates and amounts: OpenAI’s help center on ads in ChatGPT, the announcement of the new buying models, and reporting on the rollout and the self-service Ads Manager. Checked September 5, 2026, and the European self-serve launch in Digiday, August 31, 2026, checked October 29, 2026.

An auction, not a fixed price

You bid, and the auction decides the final price based on competition and relevance. So there is no rate card. If you are in a high-demand category, you usually pay more than in a niche category.

Important: a maximum bid is not a price. In an auction you pay what it takes to win, not the most you were willing to pay. OpenAI uses a second-price auction with relevance weighting.

CPM, CPC or conversion optimization: which model does OpenAI use?

All three, and you choose.

• CPM (cost per thousand impressions): you pay for reach. Choose this if your goal is brand awareness and a single click is worth little to you.
• CPC (cost per click): you only pay when someone clicks through. Choose this if you want traffic to a landing page and that page is in order.
• Conversion optimization (in practice steering on CPA): you let the auction steer on a measured conversion. Choose this once you report enough conversions for the model to learn.

The third model only works if you actually report conversions back, with the OpenAI pixel or the Conversions API. See conversion tracking for ChatGPT ads.

What you actually pay

Honest answer: it cannot be captured in one number, and anyone who gives you a precise figure is making it up. The channel is young and there are few independent benchmarks.

What can be checked:

• OpenAI recommends a starting bid of $3 to $5 per click for manual CPC bidding.
• For CPM campaigns you set a maximum bid yourself; OpenAI names no default amount.
• Early campaigns reported average click prices of $2 to $5.

Four factors decide what you pay:

• Category. The higher the value of a customer in your market, the harder people bid.
• Competition at the moment. In a young channel there are still gaps in the auction. They fill up as more advertisers join.
• Relevance. The better your ad fits the conversation, the more efficiently you buy.
• Quality of your landing page. Not as an official score factor, but in what your click brings in.

What a click is worth depends on your landing page

The click price is set in the auction; the value of that click is not. A visitor from ChatGPT comes out of a conversation: they have context, expect a direct answer and may already have seen a recommendation. If they land on a page that starts over with a brand story, you are paying for a click you throw away yourself.

There is a second reason to be sharp on this. If your landing page only builds its content after JavaScript runs, a crawler sees an empty page and your organic chance on exactly the same question is zero. You then buy a click for a question you could have appeared on for free. See landing pages for ChatGPT ads and the technical scan.

Setting a budget for a first test

Start small. Set a test budget that buys enough clicks to learn something, and measure which pages and conversation contexts convert. Then scale what works.

Two things to arrange up front, not afterwards: put the pixel on your site before you turn the campaign on, and determine your break-even CPA (average order value times gross margin). Without that second number, you do not know whether a click price is high or low.

And the other AI channels?

ChatGPT is not the only channel where you can buy ads, and the pricing models differ.

Perplexity: no ads for sale anymore

Perplexity sold ads on a CPM basis, but stopped accepting new advertisers in October 2025 and is winding down ads further according to reporting from February 2026. So Perplexity is not a buying channel. If you want to be visible there, it can only be organic: see getting found in Perplexity and the overview advertising in AI search engines.

Google AI Overviews: the existing Google Ads system

Google has built ads in AI Overviews into the existing Google Ads ecosystem. You do not manage a separate platform: AI Overview placements run along in your regular Search campaigns, with the same CPC model and the same auction.

That also means Quality Score applies in full. Advertisers with relevant landing pages, a good CTR history and strong ad quality pay less per click. What you pay differs per keyword and market; there is no fixed surcharge.

The channels side by side

A comparison on absolute amounts would suggest there are fixed prices. There are not. What you can compare is how you buy and what drives the price.

Channel	Bidding models	What drives the price	Intent level
ChatGPT ads	CPM, CPC, conversion optimization	Auction per conversation context; category, competition and relevance	High
Perplexity ads	No longer for sale since October 2025	n/a	n/a
Google AI Overviews	CPC, through the existing Google Ads auction	Keyword auction and Quality Score	Very high
Google Search (standard)	CPC	Keyword auction and Quality Score	High
Google Display	CPM, CPC	Reach and audience; little intent signal	Low
Meta ads	CPM, CPC	Audience and creative quality	Medium
LinkedIn ads	CPM, CPC	Job and company targeting; structurally more expensive per reach	Medium-high

So the relevant comparison is not CPM against CPM, but cost per quality interaction. A more expensive impression in a conversation where someone is actively making a choice can work out cheaper than a cheap impression next to content nobody wants.

Attribution: where it really goes wrong

The fundamental problem

The biggest challenge with AI ads is not the price but attribution. How do you value an ad someone sees in a ChatGPT conversation, after which that person visits your site three days later through a branded search and converts?

Last-click attribution gives that conversion to the branded search. The ad disappears from your reporting, and the channel is structurally undervalued. Exactly the same happens with your organic AI visibility, for exactly the same reason: see tracking ChatGPT traffic in Google Analytics.

Multi-touch attribution models

• Linear attribution: every touchpoint gets equal weight. With four touchpoints, the AI ad gets a quarter of the value.
• Time-decay attribution: touchpoints closer to the conversion weigh more, but earlier touchpoints keep value.
• Data-driven attribution: a model decides from your own historical data what weight each touchpoint deserves. Only works with enough conversion volume.

Start with linear attribution if you have little data, and switch to data-driven once you have enough conversions per month for the model to learn.

View-through conversions

AI ads produce relatively many view-through conversions: people see the ad, do not click, and come back later anyway. If you only measure click-through, you miss that part completely. How large that part is for ChatGPT is not publicly established, so do not assume a fixed number; measure it in your own data.

Set your view-through conversion window to at least seven days, and longer for a longer sales cycle.

Branded search volume as a proxy

A pragmatic way to see the indirect impact: track your brand-related search volume in Search Console from the start of your campaign. If it rises while you changed nothing else, you are probably seeing the effect your attribution model misses. It is not proof, but it is a signal, and it is free.

Five ways to make your budget more efficient

1. Choose your bidding model by goal, not habit. CPM for reach, CPC for traffic, conversion optimization once you report enough conversions.
2. Set up measurement before you turn the campaign on. A week without a pixel is a week without learning data.
3. Steer on the landing page, not only the bid. That is the biggest lever, and it costs no media budget.
4. Exclude irrelevant context. Not every conversation that mentions your category is a buying moment.
5. Combine with organic visibility. You do not need to buy clicks from people who already see you named. See ChatGPT ads or GEO.

ChatGPT ads versus organic visibility

Every click from an ad costs money, month after month. A mention you earn organically in AI answers does not. And an ad only reaches users on Free and Go: whoever pays for ChatGPT never sees your ad, but does see the answer.

That is why GEO is the cheapest way to become visible in ChatGPT. Ads accelerate, organic visibility stacks. How to weigh the two in dollars is in ChatGPT ads or GEO.

Prepare without an ad budget

The smartest thing you can do costs no ad money: make sure you are already visible organically. Measure with the prompt tracker for which questions you are mentioned now and for which you are not.