ChatGPT Ads Cost Per Lead: No Lead Price, Just a Bid and a Budget Cap

Short answer: there is no cost per lead on ChatGPT Ads, because OpenAI does not sell leads. Every billing model documented on the pages we read prices an impression or a click. What it does publish is a recommended starting maximum bid of $3 to $5 USD per click, a $25 minimum daily budget on an account billed in US dollars, and a spending cap that works weekly rather than daily. Setting that minimum exposes you to up to $175 of billed media spend in a budget week. That $175 is a ceiling on what you can be charged, not a sum you are required to spend. What you get for that money, you work out yourself, afterwards, with tracking you install on your own site.

Everything below was read on the vendors' own pages on 22 September 2026. Where a number is not published, this article says so rather than borrowing one from a round-up.

What OpenAI actually bills you for

OpenAI's pricing article is blunt: ChatGPT Ads "supports impression and valid-click billing", and "your campaign objective determines how your ads are priced". The three objectives are Views (CPM), where you "pay per 1,000 impressions served", Clicks (CPC), and Conversions.

That third one is where the meter looks like it should change. It does not. Conversion campaigns come in two shapes: oCPC campaigns "are billed based on valid clicks and optimize toward conversions that happen after someone clicks your ad", oCPM campaigns "are billed based on impressions served". The pricing article puts it in one line, that conversion-optimized campaigns with click billing "charge for valid clicks, not conversions".

Four documented billing models, and all four charge for traffic. A lead submission is something the system can aim at, not something you buy by the unit. And OpenAI says existing CPM, CPC, oCPC or oCPM campaigns "cannot be changed to a different objective or billing model at this time", so the meter you pick on day one is the meter you keep.

The budget is weekly, not daily

OpenAI publishes a minimum daily budget in 23 billing currencies. For an account billed in USD it is currently $25, and every value in the table is a whole multiple of five in its own currency: USD, AUD, CAD and NZD all sit at 25, EUR and GBP at 15.

The word daily is doing less work than it looks. OpenAI's rules say a daily budget is "the average amount you want your campaign to spend per day over a seven day period", that "your maximum daily spend is twice your daily budget", and that "your maximum seven-day spend is seven times your daily budget". Its worked example runs $100 a day to a $700 limit. The week is fixed as well, Sunday through Saturday in your ad account timezone, so a campaign started midweek gets a partial allowance before the counter resets.

Run that on the floor. A $25 daily budget can bill $50 on a single day, and the limit that actually binds is $175 across the week. That holds only if you leave the budget alone, because OpenAI prorates the weekly limit from the moment you change it. Both figures are limits on media spend.

There is a trap attached. OpenAI warns that a campaign that "uses up its weekly allowance early... may stop delivering until the next budget week even while marked active". The pages we read do not say whether OpenAI notifies you, so plan to check delivery yourself.

What a 25 dollar day buys

Take the floor budget against the top of OpenAI's own recommended range. A maximum bid caps what one click can cost, so $25 fully spent buys at least five clicks at a $5 ceiling and at least eight at $3. Read weekly, which is how the cap actually works, $175 buys at least 35 clicks at a $5 ceiling and at least 58 at $3.

Treat all four as floors that assume the budget is fully spent, not forecasts. OpenAI describes a "relevance-weighted, second-price auction", and since the figure you set is a maximum bid, a click may cost less than it; what you actually pay against that bid is not published on the pages we read. Delivery also "need not be even within a day or across the week", and nothing guarantees the budget is spent at all. A few dozen clicks a week is thin traffic on which to read a conversion rate.

The default has no cost ceiling

This is the part a first-time buyer is most likely to miss. The pricing article says "advertisers set a maximum bid at the ad-group level". A separate bid-strategy article sits alongside it: Maximize results, an automatic bidder, "is selected by default" on eligible new ad groups.

Maximize results carries an explicit disclaimer. In OpenAI's words it "prioritizes result volume and does not guarantee delivery against a specific CPA, CPC, ROAS, or other cost-efficiency target at this time". If you need a ceiling, OpenAI points to "the manual or advanced bid controls available in Ads Manager", the named one being Manual: Max bid.

So on eligible new ad groups the out-of-the-box setting is the one with no cost-per-result promise attached, and the $3 to $5 guidance belongs to the path you choose deliberately. The bid-strategy article was last updated about a month ago and the pricing article within the last week, worth remembering where they pull in different directions.

Your lead has to be one of their events

If you want a cost per lead, you build it yourself. Ads Manager Beta reporting "currently includes impressions, clicks, spend, click-through rate (CTR), average CPC, average CPM, and conversions". That is a list of what reporting includes rather than a statement of everything it holds, and no cost-per-conversion column appears in it. Spend divided by leads is arithmetic you do in your own spreadsheet.

Two conditions sit in front of that. Conversion tracking has to be in place through the Conversions API, a JavaScript pixel, or both. And the event has to be a standard one. OpenAI names supported examples as "a purchase, sign-up, lead submission, or other supported standard conversion event", so an ordinary lead form is covered, while "custom conversion events are not currently supported for conversion-optimized campaigns". A lead defined more narrowly, a booked survey or a quote request carrying a job type, has to be mapped onto a standard event first.

One more disclosure decides how you read your dashboard on day one. Impressions, clicks and CTR "can appear within minutes", but "spend reporting can take longer, so a displayed spend of zero does not necessarily mean no charges have accrued". On audience: OpenAI says it does not show ads to users on Plus, Pro or any Business plan, or to accounts identified as belonging to users under 18, "based on account-level age information and age prediction where available". The pages we read describe who is excluded, not who the remaining audience is.

Sponsored agents are not priced anywhere we looked

The format that led OpenAI's 16 September announcement is the one you cannot cost out today. Sponsored Agents let someone "chat with an AI representative of a business in ChatGPT" after clicking an ad. OpenAI's help article on it runs 829 characters of article text, and measured on that text it contains no instance of bill, charge, price, pricing, cost, fee, pay, spend, budget, rate, free, CPC, CPM or a dollar sign. It says the format is "currently available only to selected advertisers in a limited alpha test", and that "we're not accepting early access requests at this time".

The launch post is similar: 5,734 characters of article text, with no instance of price, pricing, bid, budget, auction, minimum, per click, CPC, CPM or a dollar sign. That is not concealment, and it is worth saying so, because the rates are published in the help centre on a page openly subtitled as covering pricing. A launch post without a rate card is normal, and so is a closed alpha without one. What the pages we read do not answer is what Sponsored Agents will cost if the test opens up. We covered what shipped that day in our write-up of Sponsored Agents, and we track availability separately in our note on self-serve by country.

The charge trigger is the real comparison

For a local service business the question is rarely which number is smaller, but what has to happen before you owe money. ChatGPT Ads bills on the click or the impression, whether or not the person ever becomes a lead, books, or answers the phone. Angi Services, which is Angi's managed booking product and not its lead-selling product, runs the other way. On Angi's own account it costs a pro nothing to use "other than certain avoidable fees", the pro "is paid the amount listed by Angi for the job", and Angi takes "a portion of the booking price customers pay". It even carries the counterparty risk: the pro "will be paid as long as you complete the job, even if the customer does not pay Angi". We did not examine Angi's lead-selling product, and nothing here describes it. For how the lead marketplaces meter their charges, see what a home services lead actually costs.

On the pages we read, neither publishes a price per outcome. OpenAI publishes its inputs, the bid and the budget, and no cost per lead. Angi says it takes a portion of the booking, and the page we read does not say how big that portion is.

Which leaves one honest comparison, and it is a break-even rather than a price. It is not a comparison with Angi Services, which does not sell leads by the unit. It is a comparison with whatever you pay a lead marketplace, a product we did not examine here. If a lead costs you $L somewhere else, a ChatGPT Ads click at a maximum bid of $B has to convert at better than $B divided by $L to win.

• Against a $25 lead: better than 12 per cent of clicks at a $3 ceiling, better than 20 per cent at $5.
• Against a $40 lead: better than 7.5 per cent at $3, better than 12.5 per cent at $5.
• Against a $150 lead: better than 2 per cent at $3, better than 3.34 per cent at $5.

Those lead prices are illustrations, not quotes, because no marketplace here publishes one. Since $3 to $5 is a bid ceiling rather than a click price, a campaign clearing the auction below its maximum needs a lower conversion rate than the list demands. The comparison also assumes a lead is a lead: nothing we read establishes whether a marketplace lead is exclusive to you, so treat the two sides as like-for-like only if you know they are.

How to decide

Budget for up to $175 across a budget week rather than $25 for a day, and treat a few dozen clicks as the ceiling the arithmetic allows rather than a number to plan against. Move off the default bid strategy if you need a cost ceiling, and pick the billing model carefully, since OpenAI says existing campaigns cannot be switched to a different objective or billing model at this time. Before you write any copy, check that your definition of a lead survives translation into one of OpenAI's standard events.

The wider point outlives this platform. Getting recommended inside an AI answer is its own acquisition channel now, and the paid version is priced in clicks while the unpaid version is not priced at all, which is the subject of our AI search optimization work. If the calls are the part you care about, the economics of answering them sit on our AI calling pages for home services.

Two disclosures, since this article spends its length asking vendors to publish their numbers. This page carries no affiliate link and nobody paid for its conclusions, though it does link two of our own service pages and you should read it knowing that. And imisofts fails the standard this article sets, harder than either vendor does: we publish no price per outcome, and unlike OpenAI we publish no input price either, no rate card and no minimum. Asked what a booked job will cost, our answer today is that it depends on your market and your close rate, which is the answer we just called insufficient. Until we publish something, hold us to the same test and make us show our working before you sign anything.