When B2B software brands rushed into ChatGPT’s new ad inventory this spring, retail and DTC marketers largely stayed on the sideline — and the trade press framed it as slow adoption. It’s not. It’s a conversion-first evaluation framework doing exactly what it’s supposed to do: screening out channels that can’t yet prove unit economics at the household level.
The data backs up the instinct. Eight of the top 10 advertisers on ChatGPT are B2B SaaS, productivity, or developer-focused companies, according to Similarweb. No major retailer cracked the top 10. That’s not coincidence. It reflects a structural mismatch between what ChatGPT currently offers commerce brands and what performance marketing budgets actually require.
Why Retailers Sat Out — And Why That Was the Right Call
Retail’s absence from ChatGPT advertising isn’t primarily about unfamiliarity with the platform. It’s about a specific bad experience with it.
ChatGPT’s Instant Checkout feature launched in September 2025, letting users complete purchases inside the chat interface without visiting a retailer’s site. Six months later, it was effectively unwound. OpenAI announced in March 2026 that it would shift focus to product discovery and let retailers use their own checkout experiences, after internal acknowledgment that users were researching products but not completing purchases inside the platform.
Walmart provided the clearest evidence. The retailer made approximately 200,000 products available through Instant Checkout starting in November 2025. Walmart’s EVP of product and design described purchases completed inside ChatGPT as converting at one-third the rate of click-through transactions to Walmart.com, and called the in-chat experience “unsatisfying” — no familiar brand environment, no visible return policies, no standard security signals. Even after Walmart adapted its approach by embedding its own shopping assistant into ChatGPT and Google Gemini, conversion rates only recovered to roughly 70% of Walmart.com‘s baseline. ChatGPT works as a discovery surface. It doesn’t work as a replacement for the brand’s own checkout.
That history matters for how retailers should evaluate ChatGPT ads today. The platform is maturing, and AI-referred traffic does convert at higher rates once shoppers reach a retailer’s own site. Criteo has reported that users referred from ChatGPT are converting at close to twice the rate of other referral channels in certain retail categories. That’s a real signal worth watching. But it’s not yet the same thing as a channel with a closed attribution loop, a published incrementality methodology, and validated unit economics at scale for commerce brands. Until those pieces are in place, the evaluation framework for a performance budget has to hold firm.
ChatGPT Ads Can’t Close the Attribution Loop — Programmatic Direct Mail Can
Conversational AI ad impressions are served to logged-in users within a chat interface, but advertisers currently cannot match an exposed user back to a specific household, a specific transaction, or a specific customer record in their CRM. The platform launched pixel-based attribution and a Conversions API alongside its self-serve Ads Manager in May 2026, which covers click-through conversions — but reporting remains aggregated by design, with no user-level data and no demographic breakdown. For the significant share of influence that doesn’t result in a direct click — users who see an ad, continue researching, then convert later through branded search — there’s no measurement mechanism.
Programmatic direct mail operates differently at the foundation. Every send is logged to a verified postal address, and matchback attribution ties that address directly to your transaction file. Every conversion is independently verifiable against your own data with no inference layer, no device graph dependency, no modeled estimates standing in for actual measurement. The attribution denominator is entirely composed of real households at real addresses.
Holdout-Validated ROAS: What Retail Campaigns Actually Deliver
Retail programmatic direct mail campaigns consistently deliver strong ROAS with holdout-validated incrementality. These aren’t modeled estimates or platform-reported figures. They’re calculated from controlled holdout groups where a randomized subset of the audience is physically withheld from receiving mail, and the lift in conversions between the mailed and held-out groups is measured against actual purchase data.
One outdoor retail brand running a Postie campaign achieved 3,951% ROAS with $651,825 in incremental revenue, measured through holdout testing that isolated only the revenue that wouldn’t have occurred without the campaign. A candle and home fragrance brand achieved a 4X jump in conversion rate and 34.3% higher AOV while mailing only one-third of its standard audience by concentrating send volume on households identified as being in a genuine buying window.
ChatGPT’s ad platform has published no incrementality methodology, no holdout framework, and no third-party-verified lift studies for retail verticals. Until that changes, there’s no credible way to compare its performance against a channel with a proven measurement infrastructure and no defensible way to hold it to the same CPA standard.
Acquisition Costs You Can Actually Benchmark Across Cycles
Retail and DTC brands running acquisition campaigns through Postie see CPAs that can be tracked at the cohort level across campaign cycles, optimized through ML-driven audience scoring, and compared against blended digital CPAs with a shared denominator: cost per acquired customer who actually transacts.
ChatGPT ads launched with CPMs around $60 for early enterprise advertisers, and while the self-serve Ads Manager removed the $50,000 minimum spend threshold in May 2026, no published conversion-rate benchmarks for commerce categories exist. CPA projection for retail remains speculative. For performance marketers who need to hold channels to the same unit-economic standard, that’s a disqualifying gap — not permanently, but right now.
Defined Attribution Windows Make Payback Periods Precise
Because programmatic direct mail campaigns have discrete in-home delivery dates, you can set 30, 60, or 90-day attribution windows and calculate payback with precision. Conversational AI ad exposure has no equivalent delivery timestamp, no standard attribution window, and no mechanism for isolating the ad’s contribution from the organic intent signal that prompted the user’s query in the first place.
That’s not an edge case. It’s a structural limitation of the format. If you can’t isolate the channel’s incremental contribution from existing demand, you can’t accurately price the spend, and you can’t confidently defend what you’re getting for it.
Budget Reallocation Requires Proof, Not Buzz
The question retail CMOs should be asking isn’t “why aren’t we testing ChatGPT ads?” It’s “why would we move budget from a channel with proven, independently verifiable unit economics to one that cannot yet demonstrate the measurement prerequisites we require?”
That’s not resistance to innovation. It’s fiduciary rigor applied to media spend. ChatGPT’s shopping influence is real and growing — 53% of US consumers now use AI tools to research products, and 28% turn to AI for shopping research every day, according to Pacvue’s 2026 Funnel Rewired report. Ignoring that trend would be a mistake. But there’s a meaningful difference between monitoring a channel’s development and allocating performance budget to it before the measurement infrastructure exists.
The same conversion-first framework that keeps performance teams disciplined on paid search and paid social should apply to every emerging channel. Right now, ChatGPT ads don’t clear the bar for retail performance budgets and retail’s collective decision to hold back while software brands run the early-mover experiments is exactly the right call.
Postie gives retail and DTC performance teams deterministic targeting, matchback attribution, and holdout-validated incrementality in a single platform.
