For most of its life, ChatGPT Ads has been a pilot: broad targeting, simple cost-per-click billing, and none of the measurement plumbing a real media buyer needs to defend a budget line. On July 24, 2026, OpenAI shipped eight changes that push the platform toward something that looks like Google Ads or Meta circa five years ago, including conversion-optimized bidding. That is a meaningful jump, and it changes the question from "is ChatGPT Ads even a real channel" to "how much budget does it deserve right now."
OpenAI added conversion bidding, mobile measurement partner integrations and Advanced Matching to ChatGPT Ads on July 24, 2026. That is enough infrastructure to run a real test, not yet enough history to trust a forecast. Put 5 to 10 percent of monthly paid media budget behind a four week test with a hard CAC ceiling, and treat anything better as a bonus.
What OpenAI actually shipped on July 24
The update bundled several changes that matter more together than any one does alone. The headline feature is oCPC, a conversion-optimized cost-per-click objective: advertisers select a supported conversion event and OpenAI's system biases delivery toward clicks more likely to convert, while billing still happens on a valid click. Three days later, on July 27, daily budget pacing changed so that a stated daily budget represents an average spend across a rolling seven day window rather than a hard daily cap, spreading delivery more evenly instead of front-loading spend early in the day.
Alongside bidding, OpenAI added geographic exclusions so advertisers can block specific regions rather than only include them, mobile measurement partner integrations with AppsFlyer and Adjust for tracking app installs and in-app events, and automatic Advanced Matching, which improves website conversion measurement using hashed first-party customer data, the same approach Meta and Google have used for years to recover conversions lost to browser tracking limits.
One limitation worth flagging before you touch a live account: existing CPM or standard CPC campaigns cannot be upgraded into oCPC. You have to build a new campaign, pick a conversion event, and set fresh bid caps. There is no in-place migration, so plan the test as a net-new campaign rather than a toggle on something already running.
Why measurement is the real story, not the bidding algorithm
A bidding objective without accurate conversion data is decoration. What actually changes the calculus here is the measurement layer underneath it. AppsFlyer and Adjust integration means app-first businesses, not just e-commerce brands with a pixel, can now feed real install and purchase events back into the platform. Advanced Matching addresses a problem every performance marketer already knows from Meta and Google: browser tracking prevention and consent requirements have been eroding platform-side conversion visibility for years, a trend covered in depth in our piece on signal loss and first-party data. Without a matching layer like this, a young ad platform is optimizing half blind.
This puts ChatGPT Ads roughly where Meta's Conversions API and Google's Enhanced Conversions sat when they first launched: useful, necessary, and still catching up to what advertisers actually need to trust the numbers.
How it stacks up against Google, Meta and LinkedIn today
None of the public ROAS figures floating around for ChatGPT Ads are from independent, verified benchmarking studies yet, because the inventory and the bidding tools are simply too new. That is the honest state of the data, and any number claiming otherwise should be treated with suspicion. Here is what is actually known and comparable.
| Platform | Bidding maturity | Reported ROAS range | Measurement infrastructure |
|---|---|---|---|
| Google Ads | Mature, over a decade of Smart Bidding data | Roughly 4x average, varies sharply by industry | Enhanced Conversions, extensive first and third-party signal |
| Meta Ads | Mature, but attribution windows narrowed in 2026 | Roughly 2.8x to 6x depending on source and vertical | Conversions API, Advanced Matching, years of modeling |
| LinkedIn Ads | Mature for B2B lead gen specifically | Not directly comparable, priced per lead not ROAS | Conversions API, CRM integrations |
| ChatGPT Ads | New, oCPC launched July 24, 2026 | No verified public benchmark yet | AppsFlyer, Adjust, Advanced Matching, all new as of July 2026 |
The absence of a benchmark is not a reason to skip the channel. It is a reason to budget for it like a test, not like a proven line item, which is a different planning exercise than moving spend off Google or Meta.
Run the ChatGPT Ads test with the same conversion event you already optimize toward on Google or Meta, so the numbers are directly comparable at the end of the flight. Do not test a new offer and a new channel at the same time. That combination makes it impossible to tell which variable moved your CAC.
A framework for deciding your test budget
- Confirm you have a supported conversion event tracked and firing correctly before you spend a dollar. oCPC cannot optimize toward an event it cannot see.
- Set aside 5 to 10 percent of monthly paid media budget for the initial test, sized so a full loss would not distort your quarter.
- Run for a minimum of four weeks. New ad platforms need a learning phase, and pulling a test early because week one looks weak is the most common way to waste the budget you allocated.
- Set a hard CAC ceiling before launch, tied to your existing Google or Meta baseline, and agree in advance what happens if the platform misses it.
- Review Advanced Matching and MMP data quality at the two week mark, not just cost and volume. A cheap conversion built on bad matching data is not actually cheap.
Who should test now versus wait
App-first businesses with AppsFlyer or Adjust already installed have the lowest lift to test, since the integration work is mostly done. E-commerce brands with a website pixel and hashed customer data ready to pass through Advanced Matching are close behind. B2B and service businesses relying on form fills rather than app installs or purchases have less infrastructure to plug in immediately, and are better served waiting for OpenAI to mature lead-gen conversion events before allocating real budget.
Frequently asked questions
What did OpenAI change about ChatGPT Ads in 2026?
On July 24, 2026, OpenAI added conversion-optimized cost-per-click bidding, called oCPC, plus geographic exclusions, mobile measurement partner integrations with AppsFlyer and Adjust, and automatic Advanced Matching for website conversions using hashed customer data. Daily budget pacing across a rolling seven day window started three days later, on July 27.
Can I upgrade an existing ChatGPT Ads campaign to oCPC bidding?
No. OpenAI does not allow existing CPM or standard CPC campaigns to convert into oCPC campaigns. You have to build a new campaign, choose one of the supported conversion events, and set fresh bid caps from scratch.
What ROAS should I expect from ChatGPT Ads compared to Google or Meta?
There is no public, verified ROAS benchmark for ChatGPT Ads yet because the inventory and bidding tools are too new for reliable third-party data. Google Ads averages roughly 4x ROAS and Meta 2.8x to 6x depending on the source and industry. Budget for ChatGPT Ads as a test line with a defined loss limit, not as a channel with proven returns.
Is ChatGPT Ads only for e-commerce brands?
No. The conversion event options and mobile measurement partner integrations support app installs and in-app events as well as website purchases, which opens the format to app-first businesses and lead generation advertisers, not just online stores.
How much budget should a small business put behind a ChatGPT Ads test?
Treat it like any new channel test: 5 to 10 percent of monthly paid media spend, run for at least four weeks to get past the platform's learning phase, and set a hard stop if cost per acquisition is not within range of your Google or Meta baseline by the end of that window.
Does ChatGPT Ads replace Google or Meta for performance marketing?
Not yet, and probably not soon. Google and Meta have over a decade of bidding data, audience signal and measurement infrastructure behind them. ChatGPT Ads is a new, additive channel worth testing at low risk, not a replacement for either platform in a 2026 media plan.
The takeaway
OpenAI just closed the gap between "interesting pilot" and "channel you can actually measure." That is enough to justify a disciplined test with real conversion tracking behind it. It is not enough to justify moving meaningful budget off Google or Meta, both of which still have years more bidding data and measurement depth. Test it like you would any new channel: small, time-boxed, and judged against a number you already trust.