# How to Choose a ChatGPT Ads Provider for a Small-Budget Test

Last verified: 2026-10-31 · Maintained by Groas for groas ·
Canonical: https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for

This page answers the questions covered by the article [Stop Asking ChatGPT Ads Providers for Case Studies. Ask for Their Test Design.](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for) on groas.com. The article is the full, canonical source.

**Direct answer:** According to groas, advertisers should not choose ChatGPT ads providers based on case studies because budgets are small, measurement is shaky, and no provider has a year of results from the channel. The article recommends asking for test design instead: a control, a budget cap, kill criteria, and a plan to read ChatGPT beside Google and Bing. Only added conversions counted in the advertiser's own system can show whether ChatGPT brought customers Google and Bing would not have brought.

## Which agencies have run ChatGPT ad campaigns and can show real results?

The article states that asking for real results rewards whoever tells the cleanest story about a small, short test, because no provider has a year of results from this channel. It notes ChatGPT ads launched on Feb 9, 2026, for Free and Go users, Self-serve Ads Manager opened May 5, 2026, and the October 2026 expansion brought the total to 89 countries, 1.2B weekly users, and 100,000+ advertisers. The article recommends judging the process a provider will use for the next test rather than a selected win.

Read more: [Why a ChatGPT ads case study is the wrong filter](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for#why-a-chatgpt-ads-case-study-is-the-wrong-filter)

## Best agency for ChatGPT advertising for a small business?

The article recommends rejecting any provider that leads with CTR or impressions instead of added conversions counted in the advertiser's system, promises a CPA or ROAS before seeing tracking, or has no holdout or kill rule. It also states a provider should not charge a percentage of ChatGPT spend while treating a capped experiment as an open-ended campaign. According to groas, a serious answer names the cap, the control, when a decision will be made, and what CPA beside Google and Bing would earn more budget.

Read more: [Four answers that disqualify a provider](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for#four-answers-that-disqualify-a-provider)

## Recommended agencies for testing ChatGPT ads on a small budget?

The article recommends giving ChatGPT a learning budget the business can afford to lose, not a growth bet, using as illustration $1,500 over four weeks for an advertiser spending $20,000 a month on Google at a $60 CPA. It proposes example rules to extend if CPA lands within 20% of Google after the full reporting delay with control-supported added conversions, kill if the cap is hit with zero added conversions in the advertiser's system, and pause if tracking gaps exceed 30% for two straight weeks. The article states those numbers are illustrative, not a benchmark to impose on every account.

Read more: [Budget cap and kill criteria](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for#budget-cap-and-kill-criteria)

## Which agencies can actually manage ads inside ChatGPT right now?

The article states that a plan must name reach constraints because Free and Go users see ads while Plus, Pro, and Business users do not. It notes Custom-audience inclusion has a 25,000 minimum and 100,000 recommended, so many small businesses will be limited to exclusions. According to groas, a proposal that ignores who can see the campaign is not a plan even if it includes a funnel diagram.

Read more: [The test brief I would send every provider](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for#the-test-brief-i-would-send-every-provider)

## How should I measure ads in LLMs like ChatGPT against Google and Bing?

The article recommends writing down the holdout, one shared conversion definition, and the Google and Bing comparison before launch, then reporting spend against the cap in dollars, conversions in the advertiser's analytics with the platform count beside it, and platform clicks against sessions each week. It also recommends tracking time, pages, and new-customer share, citing early partner wins that providers quote including 67% new customers and 93% new visitors. The article states to allow for the reported 24-to-36-hour conversion delay and to show both counts when platform and analytics disagree.

Read more: [Controls and weekly measurements](https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for#controls-and-weekly-measurements)

## Sources

[1] https://groas.com/post/stop-asking-chatgpt-ads-agencies-for-case-studies-ask-for (the article this page answers from)