ChatGPT Ads adds view-through conversions on a fixed one-day window

ChatGPT Ads adds view-through conversions on a fixed one-day window

Google AdsAugust 20, 2026
By Antonio Fernandez

TL;DR

  • OpenAI switched on view-through conversion reporting in ChatGPT Ads on a window fixed at one day from an eligible ad impression, surfaced only through documentation updated on 18 and 19 August 2026.
  • The metric appears at campaign, ad group and ad level as VTA (1d), reachable through the Conversions column or Edit columns, and it exports to CSV.
  • View-through conversions are excluded from the main Conversions total, cost per acquisition, post-click conversion rate, and from bidding, billing and conversion optimisation.
  • Conversion-optimised campaigns, live since 5 June 2026, keep training on click-attributed events alone, and no pixel or Conversions API changes are needed.
  • OpenAI published no eligibility criteria, no markets, and no definition of the click window even though the interface shows 30 days.

OpenAI switched on view-through conversion reporting in ChatGPT Ads, credited on a window fixed at one day from an eligible ad impression. The change surfaced only through documentation, and the trade publication PPC Land pieced the timeline together: the ChatGPT Ads Measurement Pixel and Conversions API developer pages now describe view-through attribution, the Ads Manager Beta measurement help article was updated on 18 August 2026, and the conversion measurement article was updated on 19 August 2026. There was no blog post, no press release and no advertiser email.

The metric is fenced off from everything that decides delivery. ChatGPT Ads view-through conversions are excluded from the main Conversions total, from cost per acquisition, from post-click conversion rate, and from bidding, billing and conversion optimisation. It is a number you can read, not a number the platform acts on.

What was actually added

The metric appears at campaign, ad group and ad level as VTA (1d). It is reachable by hovering the Conversions column, by adding a column through Edit columns, and it exports to CSV alongside the other reporting fields. The one-day window runs from an eligible ad impression and, according to the documentation, operates independently of the click window. Where a single conversion qualifies for both a click and an impression, the documentation states the click takes precedence, so the same conversion is not counted twice across the two attribution paths.

No implementation work is required. Advertisers already sending order_created or lead_created events through the ChatGPT Ads Measurement Pixel or the Conversions API get the reporting automatically, provided their account is eligible. OpenAI has not published what makes an account eligible. PPC Land reports the documentation limits the feature to eligible accounts without defining account tiers or geographic availability.

The gap between a reported metric and an actionable one

This is the part worth slowing down on, because it is the difference between this launch and how view-through reporting works on mature ad platforms. Elsewhere, a view-through window is wired into the machine: it feeds automated bidding, it moves the cost per acquisition figure that budgets are defended with, and changing the window changes what the system optimises toward. Here, OpenAI published the number and then explicitly kept it out of the loop.

Conversion-optimised campaigns in ChatGPT Ads, live since 5 June 2026, continue training on click-attributed events alone. Nothing about the model's inputs changed on 18 or 19 August 2026. The same is true of billing. What changed is that an advertiser can now see a count of conversions that followed an impression without a click, and can see it broken out at three levels of the account.

Read as our own analysis rather than as anything OpenAI stated: a reported-only metric is still useful, because it gives you a rough sense of whether impressions are doing work that clicks are not capturing. It is only dangerous when someone forgets the fence and starts treating the number as performance.

The table below sets out where VTA (1d) does and does not appear, using what the documentation states.

The gap between a reported metric and an actionable one
WhereIs VTA (1d) included?
Campaign, ad group and ad level reporting columns and CSV exportsYes, as VTA (1d), via the Conversions column hover or Edit columns
Main Conversions totalNo, that total stays click-attributed
Cost per acquisition and post-click conversion rateNo
Bidding, billing and conversion optimisationNo, conversion-optimised campaigns still train on click-attributed events

What to do in your reporting this week

  1. Do not add VTA (1d) into blended ROAS or CPA sent to a client. Mixing a metric that cannot influence delivery into a cost figure inflates apparent performance while changing nothing about what the platform buys. If you report it, report it in its own row, labelled as view-through and dated, and say plainly that it does not feed bidding.
  2. Check what click window your own account shows. The interface shows a 30-day click window in the screenshots PPC Land documented, and OpenAI's documentation never defines that default or the permitted range. Since you cannot look the answer up, look at the account.
  3. Expect a reporting-automation break. Any script, sheet import or dashboard that keys off column labels now faces a column that the documentation calls VTA (1d) and the interface calls "View-through conversions (1 day)". Match on both strings or you will silently drop the column when someone changes the export path.
  4. Give attribution 24 to 48 hours before you read it. OpenAI advises that window for attributed conversions to appear, so a same-day pull will understate the count and a day-over-day comparison built on same-day data will be wrong in a direction that looks like decline.
  5. Decide now whether the number enters your client narrative at all. Adding a metric to a report is easy and removing one is a conversation. A one-day view-through count on a platform that is still building out measurement is a candidate for internal use before it becomes a client-facing line.

A one-day window, read honestly

One day is a short window. Longer view-through windows are common practice on other ad platforms, and a shorter window credits fewer conversions to impressions than a longer one would over the same period. We are not going to name a competing platform's current default here, because those settings change and an out-of-date comparison is worse than no comparison.

The honest reading of a fixed one-day window is that it is conservative. It will not credit the person who saw an ad in a ChatGPT answer, thought about it for three days and then bought directly. Whether that person exists in meaningful numbers is exactly what the metric cannot tell you, because the window is fixed and OpenAI has not said whether it will become configurable.

There is a second reading worth holding at the same time. On a conversational surface, an ad impression sits inside an answer the user asked for, which is a different context from a display impression served beside unrelated content. Nobody has published data on how impression-only exposure behaves in that setting. Anyone claiming to know how much credit a one-day window is missing on ChatGPT Ads is guessing.

What is new here, and what was already covered

Relevant Audience has followed ChatGPT Ads since the format opened, including oCPC bidding and advanced matching on 17 August 2026, the switch from fixed daily budget caps to seven-day averages, app attribution through AppsFlyer and Adjust, and ads reaching free and Go users in the EEA. This is the same measurement build-out continuing rather than a new direction.

What is genuinely new on 18 and 19 August 2026 is impression-based attribution appearing in the reporting at all, the VTA (1d) column at three levels of the account, the CSV export, and the stated click precedence rule. What is not new is the bidding model, the events you send, or the pixel and Conversions API implementation. If you already read the oCPC coverage and set your account up then, there is no new setup work here.

What OpenAI did not say

OpenAI published no eligibility criteria, so there is no way to tell from the documentation which accounts get the column and which do not. OpenAI named no markets and no account tiers. OpenAI did not say whether the one-day window will ever become configurable, and did not say whether view-through will eventually feed bidding or the main Conversions total. The documentation does not define the click window default or its permitted range even though the interface displays 30 days. There was no blog post or advertiser email announcing any of it, which is itself worth noting: an advertiser who does not read the developer documentation would have no way to know the column exists.

What this means for Thai marketers

The measurement question comes before the media question here, and that framing is our own rather than something the source reported. ChatGPT Ads is a young inventory line, and the first thing a Thai agency or in-house team should establish is whether their account even has the column, since OpenAI has published no eligibility criteria and named no markets. Thailand is not mentioned in the documentation either way.

The reporting-hygiene point travels well. Monthly client reports in Thailand often present a single blended cost per lead across search, social and now AI surfaces, and a metric that is excluded from bidding and billing does not belong inside that number. The discipline of keeping platform-reported view-through separate from click-attributed performance is the same discipline that keeps a Google Ads account honest when a platform adds a generous attribution model.

There is a strategic read for anyone working on visibility inside AI answer engines. Impression-level reporting inside ChatGPT, however fenced off, is the first measurement OpenAI has offered of exposure without a click on a conversational surface. That is the same phenomenon organic teams have been arguing about since AI answers started resolving questions without sending traffic. A one-day, bidding-excluded count is a thin instrument, but it is more instrument than existed a week ago.

Frequently asked questions

Do I have to set anything up to get view-through conversions?

No, advertisers already sending order_created or lead_created events get the reporting automatically if their account is eligible, and no changes to the Measurement Pixel or Conversions API are needed. OpenAI has not published what makes an account eligible.

Is this available in Thailand?

The documentation does not say, because OpenAI named no markets and published no eligibility criteria. The only reliable check is to open your own ChatGPT Ads reporting and look for the VTA (1d) column through Edit columns.

Will view-through conversions change how my campaigns bid?

No, view-through conversions are excluded from bidding, billing and conversion optimisation, and conversion-optimised campaigns running since 5 June 2026 continue training on click-attributed events alone. Nothing about delivery changed when the column appeared.

Can I change the one-day window?

The window is fixed at one day from an eligible ad impression, and OpenAI has not said whether it will become configurable. The click window is separate, and the documentation does not define its default or permitted range even though the interface shows 30 days.

Should I include VTA (1d) in the numbers I send clients?

Not inside a blended CPA or ROAS figure, because the metric does not influence delivery and folding it in inflates apparent performance. Reporting it as its own clearly labelled line, with a note that it is excluded from bidding and billing, is the defensible way to show it.

Getting the measurement right before the spend grows

New ad surfaces reward the teams that fix reporting early, because a metric that quietly enters a dashboard is very hard to remove once a client has seen the bigger number. If you want your ChatGPT Ads, paid search and AI-visibility reporting to agree with each other before the next quarter's budget conversation, Relevant Audience builds that measurement layer as part of ongoing account work.

Antonio Fernandez

Antonio Fernandez

Founder and CEO of Relevant Audience. With over 15 years of experience in digital marketing strategy, he leads teams across southeast Asia in delivering exceptional results for clients through performance-focused digital solutions.

Share to:
Copy link: