The Real Cost of ChatGPT Ads: A DTC Math Breakdown

OpenAI Ads · The math

Open OpenAI Ads. Pull up the CPC range. Multiply by your category's typical landing-page conversion rate. Divide that into your AOV. Whatever number comes out is the answer to whether this channel works for your brand today. For most DTC operators reading this, the answer is going to be one of two things. Either your AOV is north of $150 and the math closes inside a quarter, or your AOV is under $50 and the math does not close until something material changes about either CPCs, your LTV, or both.

This post is the math. Every input is either cited or explicitly labeled as an assumption you should swap with your own number. The point is not to tell you whether to spend on OpenAI Ads. The point is to give you a one-page worksheet so you can decide in five minutes instead of in three quarterly review cycles.

The verified numbers

Three inputs anchor everything else. Get these right and the rest is arithmetic.

CPC. Digiday's reporting on the April 2026 CPC bidding rollout puts the typical click price in the three-to-five-dollar range. Search Engine Land covered the same launch. We will use $4 as the midpoint for the math below. If your category is more or less competitive, swap it.

CPM. Launch CPMs were near sixty dollars, dropping to roughly twenty-five within nine weeks as inventory expanded. CPM matters less for our math because CPC is the bid model brands run today, but the trend line matters for the cost-decay argument in §4.

CTR. Independent estimates put ChatGPT's commercial-link CTR at roughly 1.3 percent, against Google Search's roughly 29.2 percent. That gap is the single most important number in this post. It means roughly one in seventy-seven impressions becomes a click on ChatGPT, versus one in three on Google. Your ad has to earn its way into the model's recommendation a lot more often to do the same volume.

One number you have to supply yourself: landing-page conversion rate. No public benchmark substitutes for your own data here. For the math below we will assume 2.5 percent, which is a common DTC baseline. If your real number is 1 percent, every CAC figure below doubles. If your real number is 5 percent, every CAC figure below halves. Use your number.

The math at $20 AOV

Imagine a candle brand. AOV is $20. Gross margin is 60 percent, so contribution margin per order is $12 before any acquisition cost. The brand decides to test OpenAI Ads.

Every click costs $4. At a 2.5 percent landing-page conversion rate, the brand needs forty clicks to produce one order. Forty clicks at $4 each is $160 in click cost per order. That is the brand's customer acquisition cost on this channel.

CAC of $160 against a contribution margin of $12 per order means the brand loses $148 on every first purchase. That gap closes only through repeat purchase. To recover the $160 CAC, the customer has to buy roughly fourteen more times at the same margin. Most candle brands do not see fourteen-purchase LTV.

The candle brand should not run OpenAI Ads today. The CPC math at this AOV does not close, period. The brand has three real options. Wait for CPCs to fall. Increase AOV through bundles or subscriptions. Or skip the paid channel and invest in becoming recommendable to the model organically, which is a different problem with its own technical baseline (robots.txt config, JSON-LD product schema, server-side rendering for crawler access).

The pattern at AOV under $50 generalizes. With $4 CPC and 2.5 percent LP conversion, CAC sits at $160. Any brand whose contribution margin per order is below $80 is losing money on every acquired customer and needs at least one repeat purchase to break even. If your repeat-rate within ninety days is below 50 percent, OpenAI Ads at current CPCs is structurally bad math for your brand.

The math at $200 AOV

Now imagine a premium skincare brand. AOV is $200. Gross margin is 65 percent, so contribution margin per order is $130. Same channel, same CPCs.

Forty clicks at $4 each is still $160 per order. CAC of $160 against contribution of $130 per order means the brand loses $30 on the first purchase. That gap closes inside two repeats. If the second order produces another $130 contribution, the brand is up $100 by purchase two. Most premium DTC brands see north of 30 percent ninety-day repeat rate, so the math closes in the normal LTV window.

Better yet, OpenAI Ads buyers behave differently than Meta buyers in one important way that helps this calculation. The compressed conversation-to-merchant-site path means buyers arriving from a model recommendation are pre-qualified in a way that scroll-driven Meta traffic is not. We covered the buyer-side dynamics in the audience piece. The implication for the math: your landing-page conversion rate on ChatGPT traffic will likely be at the high end of your channel benchmarks, not the low end. If your 2.5 percent assumption turns out to be 4 percent on this traffic, your CAC drops to $100, and the math closes on the first order.

The premium skincare brand should test OpenAI Ads now. The math closes inside a single repeat. The brand should size the test budget for daily variance (see the pillar explainer for what 'daily variance' means in this context), and run it for at least four weeks before drawing conclusions.

The pattern at AOV over $150 generalizes. With $4 CPC and 2.5 percent LP conversion, CAC stays at $160. Any brand with contribution margin per order above $100 is paying a manageable acquisition cost, especially if conversion on this traffic runs above your channel baseline. The decision becomes whether the volume on the channel is large enough to matter, which is a different question from whether the unit economics work.

Why CPCs will fall (and when)

Three forces push the cost curve down. They will not arrive on a predictable schedule, but the direction is not in serious doubt.

Inventory expansion. OpenAI is rolling ads out to more queries, more surfaces, more user tiers as the system matures. More inventory at a similar bid level means lower clearing prices. The CPM trajectory from $60 to roughly $25 in nine weeks is the signal: inventory growth is currently outpacing brand adoption, which compresses prices.

Auction maturity. Early auctions on every channel in history have been overpriced because the bidders did not yet know what a click was worth. As brands run the math we just walked through and re-price their bids to reflect reality, the average bid drops. Search ads in 2003 went through this. Meta in 2009 went through this. The same pattern will play out here on a twelve-to-twenty-four-month timeline.

Competitor adoption. This one cuts the other way. As more brands enter the auction, bid pressure increases. Whether this offsets the first two forces depends on which side scales faster. Our read: inventory and maturity dominate in the first eighteen months, then competition takes over.

The honest forecast is that current $3 to $5 CPCs are at the high end of where they will settle. A reasonable expectation is sub-$3 by late 2026 and sub-$2 in mature categories by mid-2027. None of these numbers are guarantees. The math sections above use $4 because that is the price today. Redo the math when your number changes.

When the math works today

This is the section most operators will skim back to. Five concrete decision rules. If your brand meets at least three, OpenAI Ads is worth testing this quarter. If it meets fewer than three, wait two quarters and rerun the math.

If your brand fails all five rules and CPCs do not fall by year-end, the answer is to skip the paid channel and invest the same dollars in becoming organically recommendable. The attribution problem (covered in the dark funnel post) makes the organic side hard to measure, and the Amazon-defaulting dynamic from rule 4 is covered in detail in the Amazon problem post. The unit economics work better at sub-$50 AOV than the paid side does.

Three numbers decide this. CPC, your LP conversion rate, and your contribution margin per order. Multiply CPC by clicks-needed (which is one divided by LP conversion). Divide that into contribution margin. If the ratio is above one, the math works today. If it is below one but your repeat rate covers the gap, the math works inside two quarters. If neither, wait two quarters and rerun.