ChatGPT Ads may have started counting conversions from ads nobody clicked. A practitioner screenshot reported by Search Engine Roundtable on July 23 shows a new tooltip on the Conversions column in ChatGPT Ads Manager splitting the figure into View-Through Attribution and Click-Through Attribution (Search Engine Roundtable). OpenAI has not announced this specific attribution change, and we have not independently observed it in an account, so treat it as a spotted interface change of unknown rollout scope, not a confirmed platform-wide default. If the screenshot reflects a live change, its meaning is straightforward: view-through credit records a conversion when someone merely saw your ad and later converted, no click required. On most channels that is a familiar, if generous, setting. On ChatGPT Ads it would land on a surface that still has no click-level conversion data and no third-party attribution validation (Auspia), so the headline conversion number gets bigger and softer at the same time.
The incumbent way
Most teams will treat the Conversions column the way they treat every other conversion column: it is in the report, so it is a conversion. The ChatGPT Ads total gets pasted into the same ROAS and CAC model that governs Google and Meta, and the channels get compared head to head. That worked while every platform in the comparison reported roughly the same kind of number. It stops working the moment one of them quietly starts counting a different thing.
As recently as May, independent field guides said ChatGPT Ads had no view-through attribution at all (Choice OMG). So if the reported tooltip is real, it points to a new capability rather than a relabel of something that already existed, which is why it is worth reacting to now instead of waiting for OpenAI to document it. The platform itself is live, self-serve, and US-wide, with a pixel and Conversions API that OpenAI documents first-party (OpenAI); a 30-day first-party-cookie window is described in earlier trade reporting rather than by OpenAI (AI Advantage Agency). What it does not have is a disclosed view-through methodology. The look-back window, the threshold that qualifies a view, and the logic that de-duplicates a view-through conversion against a click-through one are all undocumented.
The cost of standing still
Leave reporting on the blended default and the first cost is a bigger number you cannot decompose. View-through attribution almost always enlarges a channel's reported conversions relative to a strictly click-based channel, because it counts an audience the click-based channel never gets credit for. The ChatGPT Ads total rises, and nothing in the interface tells you how much of the rise is real demand versus incidental exposure.
Carry that number into a cross-channel comparison and the second cost arrives: ChatGPT Ads looks artificially efficient next to Google and Meta. Its conversions are padded by views theirs do not count, so its cost-per-conversion looks lower and its ROAS looks higher. Budget follows the flattering number, and it shifts toward the channel with the softest, least-auditable measurement.
Carry it a full quarter and the third cost is a reconciliation you cannot run. When someone finally asks how much of the ChatGPT Ads contribution was click-driven, there is no click-level path to inspect, because the platform does not expose one (Auspia). You cannot diagnose the gap on the platform that created it. The allocation was wrong, and the tool that would prove it wrong does not have the column you need.
What the evidence shows
The structural problem is simple to state. If the reported change is live, the vendor that sells you the impression is also the vendor that decides, without a click and without an outside auditor, that the impression converted. The evidence for the change itself is a single practitioner screenshot reported by Search Engine Roundtable, not an OpenAI announcement and not something we verified in an account (Search Engine Roundtable). The platform's measurement limits, by contrast, are documented independently: no click-level conversion data, no third-party validation (Auspia). Put those together and view-through credit on ChatGPT Ads would be a number defined by the seller, graded by the seller, and unverifiable by you.
This is not an isolated quirk. It is the same measurement-trust question landing on channel after channel this month. Google put an aggregate figure on AI-search clicks that its own Search Console cannot itemize. GA4 has been folding AI-referred sessions into buckets that hide their origin. The common thread is a widening gap between what a platform reports and what a business can independently confirm. View-through attribution on a self-graded ad channel is the paid-media version of exactly that gap.
The binary marketing directors now face
That is not a reporting-hygiene nitpick. It is a governance decision about which numbers are allowed to move a budget.
Path one: report the default. Paste the blended Conversions figure into ROAS and let view-through credit ride. It is simple, and it is wrong in the direction that flatters the channel selling you the ads.
Path two: split and normalize. Whenever the split appears in your account, pull view-through and click-through separately, model your cross-channel ROAS on the click-through-equivalent number, and footnote view-through as a directional, vendor-defined estimate until OpenAI publishes the window and de-duplication logic. Reconcile both against server-side and UTM data you own. Treat ChatGPT Ads as a promising channel worth testing, not a number worth forecasting against on faith.
Verdict
Choose path two. A conversion you cannot audit is not a conversion you can budget against. View-through and click-through are two different numbers, and only one of them survives the question your board will eventually ask about where the figure came from. Until OpenAI documents how view-through is counted, the honest number in the deck is the click-through one, and the view-through total belongs in a footnote, not a forecast.
Sources
- ChatGPT Ads Manager conversions value, VTA/CTA tooltip sighting (Search Engine Roundtable)
- New ways to buy ChatGPT ads, Ads Manager, pixel, and Conversions API (OpenAI)
- ChatGPT Ads guide 2026, no click-level data and no third-party validation (Auspia)
- ChatGPT Ads update, pixel and 30-day cookie window (AI Advantage Agency)
- ChatGPT Ads 2026 field guide, no view-through attribution as of May (Choice OMG)
Magnet's attribution and measurement practice builds the corrected cross-channel view: clean data collection, an attribution model that separates verified conversions from vendor-estimated ones, and reporting where budget follows evidence rather than platform vanity metrics. If your next-quarter allocation is about to run on a ChatGPT Ads number nobody can audit, talk to Magnet about an AI-channel measurement audit that normalizes view-through against click-through first.
