ChatGPT ads case studies: what six months of public spend data actually shows

Tarun Kapoor Updated July 2, 2026 11 min read

Six months into the ChatGPT ads era, the public case-study record is small, uneven, and worth reading carefully. The most-cited real-spend case is a single 15-day $60,000 account that returned $89,000 (a 1.49x blended ROAS). The most-cited failure is the native 0.91 percent CTR set against Google Search's 6.4 percent benchmark — a comparison that mostly reveals the wrong metric was chosen. This guide walks through the public cases, category benchmarks, and a four-question framework for judging any ChatGPT ads case study against your own economics.

The short answer

Six months is a small window for a new ad platform, and the case-study base reflects that. There are a handful of publicly documented real-spend accounts, several agency-authored client outcomes with partial disclosure, and a growing body of third-party category benchmarks. Read together, they show a channel that can clear at low-single-digit ROAS in research-heavy verticals when the creative and landing page match, and that punishes advertisers who imported search-buying instincts without adapting to a conversational surface. What they cannot yet show is a stable multi-quarter benchmark; the platform is still moving faster than the case studies.

A note on how to read this guide

Rather than publish anonymized customer stories that cannot be independently verified, this guide summarizes what is documented in public reporting and cites each source. Where the source is a single account, it is labeled as a single account and not generalized. Where a range is reported by a third-party benchmark aggregator, it is labeled as a benchmark range with the source. Two things this page does not do: invent numbers, and blend cases into a single "average outcome." If you need a case study against your own account, use the framework at the end.

Case: 15 days of real spend, ~$60k, 1.49x blended ROAS

Case: the pilot's rocky start and what it actually revealed

Case: the March 2026 volume scale-up

Category benchmarks reported in third-party aggregations

Third-party benchmark aggregators have begun publishing category-level engagement-to-conversion ranges. These are aggregated across accounts of unknown provenance, so treat them as directional pattern-detection, not planning inputs.

Notably absent from most reported benchmarks: high-margin consulting and professional services, financial products, and health, all of which face additional policy review and less benchmark data.

What is working, what is not — patterns across the public cases

A few consistent signals across the reporting:

• Creative tuned to conversation outperforms creative tuned to search. Cases that report headline rewrites toward direct-answer phrasing consistently show improved CPC alongside stable or improved conversion rate.
• Small-budget tests get to signal faster than expected. With CPC at $3 to $5 and no minimum, a $1,000 to $3,000 test buys enough clicks to read relevance within days.
• Attribution setup determines the story. Accounts that instrumented the Conversions API early report meaningfully higher last-click attribution than those relying on UTM parameters alone.
• CTR is a red herring at surface comparison. Every well-instrumented case shows conversion following impression more than click, and reports message-depth or brand-search lift alongside CTR.
• The homepage is the wrong destination. Cases that route ads to dedicated landing pages matched to context hints report materially higher conversion than homepage destinations.

The single most common failure mode: an advertiser who spent $10,000 to $30,000 with default creative, homepage destination, no Conversions API, and no holdout, then concluded the channel does not work based on last-click ROAS. That conclusion is defensible for that setup; it is not a conclusion about the channel.

A four-question framework for judging any ChatGPT ads case study

Whether you are reading an agency post or evaluating a peer's shared numbers, ask four questions before quoting the result:

1. What was the total spend and duration? A case study with "great ROAS" but no spend is a story. Look for real numbers.
2. What category, offer, and margin? Category shapes conversion rate; offer shapes willingness to click; margin shapes what the channel can pay for. Without all three, you cannot generalize.
3. What conversion definition? A "conversion" that means email opt-in is a different animal from one that means purchase. The definition sets the meaning of ROAS.
4. Was any incrementality measurement done? A geo-holdout, a brand-search lift comparison, or even a documented control period. Cases without incrementality report attributed ROAS, which is the ceiling, not the floor.

Answer these four for a case study and the number in the headline is either useful or discardable. For the full attribution and incrementality playbook, see our ROI measurement guide.

Frequently asked questions

Are there real ChatGPT ads case studies published yet?

Yes, but few are documented rigorously. The most useful public case as of mid-2026 is Opascope's 15-day account report of roughly $60,000 in spend against $89,000 in revenue at a 1.49x blended ROAS. Category benchmarks from Sensor Tower show impressions rose approximately 600 percent between early and mid-March 2026 as advertiser volume scaled. Beyond these, most published case studies are agency-authored and mix spend, category, and time period in ways that make direct comparison difficult.

What is the reported ROAS on ChatGPT ads?

The most-cited public number is a 1.49x blended ROAS from a single Opascope 15-day case, at roughly $1.72 per click and 2.35 percent conversion rate. Category benchmark ranges reported by third parties cluster between 1.2x and 3x in the early platform. Treat all figures as directional; ROAS depends on your margin, offer, and conversion rate, not on the channel itself.

Which categories perform best in early ChatGPT ads case studies?

Reported category benchmarks skew toward considered-purchase and research-heavy verticals. Beauty and personal care has clustered at 5.5 to 9 percent engagement-to-conversion, home goods and furniture at 4 to 8 percent, apparel at 3.5 to 6.5 percent, and electronics at 3 to 6 percent, per third-party benchmark aggregators. B2B research reports fewer clicks but higher assisted conversion when paired with brand-search tracking.

What went wrong in the ChatGPT ads pilot?

Two structural issues dominated reporting. Native CTR came in around 0.91 percent against Google Search's 6.4 percent benchmark, prompting misplaced concern given surface differences. Second, several outlets reported that the Ads Manager reporting tool was buggy or missing for parts of the pilot, leaving even native attribution unreliable. Self-serve launch on May 5, 2026 with the Conversions API improved measurement substantially.

Should I trust an agency's ChatGPT ads case study?

Read it against a four-question checklist: what was the total spend and duration, what category and offer, what conversion definition was used, and was any incrementality measurement done. A case study that reports ROAS without spend and duration is a marketing artifact, not a measurement. Prefer studies with a stated methodology over studies with a bigger headline number.

Can a small business replicate the results in case studies?

The economics scale down more cleanly than the reporting. Because the minimum spend is now zero and CPC is available, a small business can run the same auction as a mid-market advertiser at a $1,000 to $3,000 test level. What does not scale is the measurement rigor — small budgets rarely see enough conversions for a proper geo-holdout, so small businesses should judge by attributed cost per conversion and brand-search lift rather than incremental ROAS.

How did ChatGPT ad performance change from launch to self-serve?

Reported CPMs compressed from the flat $60 pilot rate toward a $25 to $60 category-dependent range as inventory expanded, and the minimum spend fell from around $200,000 to $50,000 to zero over three months. Impressions grew rapidly in March 2026 as more advertisers scaled. Native reporting improved when self-serve and the Conversions API launched together on May 5, 2026.

What does a defensible ChatGPT ads case study look like?

Spend, duration, category, offer, conversion definition, and a measurement method that includes either a geo-holdout or a matched brand-search lift comparison. Ideally: pre-registered baseline, single-variable change, and named third-party validation. Case studies missing these elements are useful for pattern recognition but not for planning.

Sources and further reading

• Opascope — ChatGPT Ads Benchmarks: ROAS, CPC, and cost from 15 days of real spend.
• The Keyword — ChatGPT Ads pilot had a rocky start.
• ALM Corp — ChatGPT's Ads Manager: what the $200k entry bar and low CTRs mean.
• Search Engine Land — ChatGPT ads expand as more brands and trigger patterns emerge.
• ScaleWise VA — ChatGPT Ads conversion rate benchmarks by category (2026).
• Context Hints — cost guide, ROI measurement, and ChatGPT vs Google Ads.

Tarun Kapoor

Founder & CEO, Context Hints

Twelve years of media buying across GroupM, WPP, Ogilvy & Mather, and Neil Patel Digital. Has personally owned media for Nestlé, Sage, Qualcomm, Aetna, Weight Watchers, Chubb and Novotel.