ChatGPT $100 Credits Were a Clever Customer-Acquisition Bet

ChatGPT $100 credits were less a giveaway than a user-acquisition loop disguised as a thank-you.

Most giveaways create a private transaction. A company gives you money, usage, or a discount. You take it. The interaction ends.

This campaign asked people to publish their opinion first.

Share what you love about GPT-5.6 Sol, or explain why you switched. Submit the public X post. Get $100 in ChatGPT credits. The first 10,000 eligible participants got the offer.

There was a catch. Claiming the credits required an existing paid ChatGPT Go, Plus, or Pro account. People could not qualify by subscribing on the submission date.

OpenAI was not handing free credit straight to brand-new users. It rewarded existing subscribers, then used their public posts to reach everyone else.

That small change turned the reward into a distribution system.

The Price of Entry Was a Public Opinion

The campaign did not ask users to repost an ad. It asked them to describe something they had built or explain why the product had won them over.

That distinction matters. A repost tells people a campaign exists. A specific story tells them what the product did.

A friend saying "this restaurant is good" barely registers. Hearing exactly which dish justified a 40-minute wait can change where you eat tonight.

One developer discussed shipping a project. Another explained how GPT-5.6 handled longer tasks. Someone else compared it with a tool whose limits blocked real work.

The campaign collected all 3 kinds of proof in public.

The claim process was simple:

1. Post on X.

2. Add the ChatGPT email and direct post URL.

3. Confirm an active Go, Plus, or Pro plan.

4. Wait for verification.

5. Receive the credits.

The people claiming the reward were already paid customers, not cold prospects sampling ChatGPT for free.

The reward gave people a reason to write the post. The product gave them something to write about.

Here is the first part of the loop: a useful product creates the opinion, and the incentive gets that opinion published.

X Put Every Testimonial Into Circulation

The channel fit the audience. Mark Zuckerberg owns social networks with roughly 3 billion daily active users, yet he posts on X when he wants developers to notice Meta's AI models.

That includes Muse Spark. The AI-model audience on X differs from the audience Meta reaches on Instagram.

Developers already use X to share projects, compare models, and debate which tools deserve their time.

Handing out gym passes outside a gym catches people already thinking about exercise. Doing the same in the frozen-food aisle means first persuading shoppers to care.

OpenAI did not need to teach the audience a new behavior. It placed the offer inside a stream where that behavior was normal.

X's recommendation feed carried Codex posts beyond each participant's followers.

People who had never used Codex saw project demos, model comparisons, and switching stories from accounts they did not follow.

Then the numbers stacked up.

The campaign tweet reached 6.24 million impressions, with 13.7K replies, 3.7K reposts, and 15.6K likes.

Those replies were not empty engagement. They were the product's sales pitch, written by people who had used it.

The giveaway traveled because every participant supplied a fresh example. Each example gave the next reader another reason to investigate.

The public post did the work of a paid ad. It created awareness, supplied a claim, and pointed the audience toward the product in a single unit.

The Posts Recruited New Users, Not the Credits

The credits did not capture many new users directly. Eligibility required an existing paid ChatGPT Go, Plus, or Pro account. The recipients already paid OpenAI, and many already used Codex.

The credits had a different job. They pushed current customers to use the product more and gave them a reason to publish proof of that usage.

The public posts did the customer acquisition. X's recommendation algorithm filled developer feeds with Codex projects, usage charts, and Claude comparisons.

A non-user did not receive the $100. They still saw what paid users were building, again and again.

The play was simple: reward current customers for creating credible ads.

Put those ads where prospective customers spend time. Let product demonstrations create demand.

Acquisition came from the reach generated by the posts, not from the credit claims.

The Product Had to Earn Day 2

Credits only work as a growth tool when the product earns a second session.

GPT-5.6 Sol felt unusually good, and it cost less than Anthropic's Fable 5, a strong high-end competitor.

The reward lowered the cost of trying the product. Product quality gave people a reason to return after the first test.

OpenAI also reset Codex usage daily during this period. Those resets were separate from the $100 promotion, but they extended the growth moment by giving existing users more chances to run long tasks.

That changed the campaign's economics. OpenAI was buying more than a post. It encouraged extra sessions that produced the next project demo or switching story.

The screenshots show what came from those sessions: apps, games, code, and long-running projects built with Codex and GPT-5.6.

The Hockey Stick Came From Several Forces Pointing the Same Way

Stefan Streichsbier's growth chart maps Codex's weekly active users from under 500K in early February to an estimated 9 million by July 16.

Codex now has over 10 million weekly active users.

The campaign deserves credit for the sharp wave of public sharing. It made the product visible where developers talk about tools.

But the campaign did not work alone.

Product quality made the praise believable. Low pricing and enough usage made switching easier. Extra usage resets kept people building. Tighter Claude limits gave some developers another reason to test Codex.

Each force pushed the same way.

A packed bar is rarely the work of one drink special. Good music keeps people inside, fair prices make the second round easy, and the line outside tells passersby that something worth joining is happening.

Calling this a simple giveaway misses the mechanism. The credits rewarded existing paid users. Those users generated public proof.

X distributed that proof to prospective users. Some tried Codex.

The loop looked like this:

Paid users → credit-backed usage → public proof → X distribution → new user interest → more Codex usage.

The campaign's output became its next input. People did not just receive a reward. They supplied the evidence that made the next person curious.

That is the lesson from ChatGPT's $100 credits. A reward can make users explain a product in public, while the product makes those explanations believable.

Do both, and the giveaway starts paying for itself through distribution.

Top Tweets of the day

1/

Source: @keean_edward

Most people are so averse to learning, after they leave school and college, that if you know even a little bit about AI, you are way ahead of the rest of the world.

Always talk to normies and most don't have a clue on how you can use an app.

2/

Source: @RhysSullivan

The future of how you can simply fix feature requests or bugs via social media directly.

3/

Source: @ivesparrowai

Brilliant use case. You just have to curate the taste so it finds only gold.

Before AI, this took hours of scrolling & saving.

Rabbit Holes

1. The 7 deadly sins of bloated AI spend

2. How Gumclaw Works

3. Chegg and StackOverflow were both basically destroyed by LLMs, what other websites / companies have already seen their traffic go to 0 because of LLMs?

What’d ya think of today’s newsletter? Hit ‘reply’ and let me know.

Do me a favor and share it in your company's Slack #marketing channel.

First time? Subscribe.

Follow me on X.

More Startup Spells 🪄

1. InVideo's "RT + Comment" Giveaway on X, Explained (LINK)

2. Red Bull's Extreme Stunts Are the Best Marketing Playbook Nobody Copies (LINK)

3. Michelin's €26.0 Billion Content Marketing Playbook (From 1900) (LINK)

4. How Old Spice Ad Speaks To Women But Targets Men (LINK)

5. Jenni AI: A Masterclass in Bundling B2C Apps (LINK)