How Cal AI went from $0 to $5.7M a month, mostly on creators

The short version

Cal AI launched in May 2024 with one feature: point your phone at a meal and get the calories and macros back. Its founders were Zach Yadegari and Henry Langmack, high-school friends, and Blake Anderson, who had already built viral apps such as Umax. They raised no outside money (Forbes).

• $5.7M in one month (January 2026), founder-stated (My First Million).
• Over $40M in sales in the 12 months before it was acquired, and 15M+ downloads (MyFitnessPal press release, TechCrunch).
• Bought by MyFitnessPal. Closed December 2025, announced March 2026. The price was never disclosed. You will hear "about $100M"; that is a podcast host's guess, not a reported figure.

The interesting part is the order of the engines. Creators alone took them to about $2M a month, and then growth flattened. Paid ads were the second engine, and they only worked once the creative changed.

Why it was built to be filmed

The insight was small and specific. Zach quit a mainstream calorie app after three days because weighing and typing every meal was "unsustainable" (CNBC). Taking a photo takes a few seconds, and you can watch it work on camera.

That matters more than it sounds. The value of Cal AI is visible in under three seconds of video: food, scan, numbers. Most apps need a voiceover to explain themselves. This one did not, which made it a perfect product for creators to show without it feeling like an ad.

It was also cheap. Cal AI charged about $30 a year against roughly $80 to $120 for incumbents, and MyFitnessPal's CEO later described it as the app "for those preferring speed over accuracy" (TechCrunch).

The first 1,000 users

1. The first ~100 came from Zach's own TikToks (founder interview).
2. The first ~1,000 came from fitness creators posting Instagram Stories, not Reels. Stories are cheap, carry a link and let you test a niche quickly. They validated it with two creators.
3. Cold outreach at volume. Roughly one creator in a hundred replies to a DM. Opening with "paid promo" made the message show up in the inbox preview (My First Million).
4. About ten creators on retainer after month one, four posts each a month, growing every month after that.

The revenue timeline

Gross monthly revenue before Apple and Google take their cut, as stated by the founders unless noted.

One breakdown worth knowing: in July 2025, out of roughly $3M in the month, about $1.2M to $1.4M went on marketing, about $1M was profit and payroll was $100K to $200K (Internet Empires). Apple pays about 45 days after a sale, so scaling spend burned cash for about six months even while the business was profitable (CNBC).

Engine one: a creator roster on retainer

"We got to about $2 million a month just from influencer marketing. After that we kind of plateaued." Zach Yadegari (My First Million)

This was not a few lucky viral posts. It was a system:

• Monthly retainers, four posts a month, paid upfront. Creators usually want paying in advance, so pure pay-per-view deals rarely happen.
• The roster grew from about 10 creators after month one to about 150 by November 2024 and about 250 by June 2025 (Nathan Latka).
• 20+ virtual assistants ran the outreach with follow-ups on day 1, 4, 7 and 10, tracked in a spreadsheet and later Airtable (Founders Arm).
• The format was "what I eat in a day". The creator scans a meal mid-video, often without saying the app's name. The scan shows three results in a row, and the team even removed a streak pop-up from creator builds because a fourth beat broke the rhythm (Sub Club).
• Not every niche works. Bodybuilders and diet creators did. Mukbang creators flopped: their audience does not want to think about calories.

The creator pricing math

This is the most copyable part of the playbook, described in detail in an interview with Brett Malinowski:

1. Estimate what 1,000 views earn you. They started by assuming about $5 of revenue per 1,000 views.
2. Pay less than that. Target $2 to $3 per 1,000 views.
3. Price the creator from their average views. A creator averaging 50,000 views at $2 per thousand gets an opening offer of $100 a video.
4. Bundle and pay upfront: four videos at once for a lower rate.
5. Protect yourself with a minimum-view clause: if the posts underperform, the creator keeps posting until they hit the agreed views.

They preferred flat fees over per-view deals. One creator went from about 10,000 views to a 7M-view video in month three; on a flat fee, that spike is pure upside (Sub Club). The most-cited single bet: $500K to sponsor a MrBeast second-channel video, which paid back about $350K to $400K directly and was justified as credibility for later deals.

Engine two: paid ads, after one expensive mistake

When creators plateaued, Cal AI hired an agency to run ads. The agency re-ran influencer videos as ads, and it lost money. The integrations had been made deliberately subtle, so they did not sell hard enough as ads. Zach took over the buying himself and later built an in-house team (Brett Malinowski).

The fix was creative, not targeting:

• Affiliate creators made videos specifically to run as ads, paid a share of revenue through a platform called Tribe.
• "Most of our videos actually aren't even people talking." The winning format was gym footage and music, then scanning food, with a one-line caption hook.
• Simple campaign structure, lots of creative. Meta's ad library tracker shows Cal AI going from 4 active ads in June 2025 to about 800 by July 2026 (AdWhispr, a third-party estimate).
• Every campaign pointed at its own App Store product page, so paid revenue could be separated from creator-driven revenue.

Ad spend reached "a little over a million" in January 2026. The founders describe paid ads as "the kill shot" on people the creators had already warmed up, and creator spend still ran to a few million a month.

The onboarding and paywall

Traffic is only half of it. Cal AI's founders said 20 to 25% of installs started a trial or paid (Nathan Latka). How they got there:

• A long quiz on purpose. About 28 steps: goal, weight, pace, diet, then a "your custom plan is ready" reveal (ScreensDesign). Adding questions that changed nothing still raised conversion, because people value what they invest time in.
• Animations between screens nearly doubled the onboarding length and lifted conversion about 10%.
• Sign-in moved to the end. The founders called it "the biggest by far".
• A hard paywall with the annual plan ($29.99, shown as $2.49 a month) pre-selected next to $9.99 monthly, and a three-day trial (Lazyweb capture). Dismissing it triggered a cheaper one-time offer.
• Constant testing. One full-time hire's only job was paywall experiments.

What went wrong

A fair teardown includes the costs:

• Accuracy. A study presented in July 2026 tested 102 lab-prepared meals on four tracking apps, including Cal AI. All four underestimated by 250 to 345 calories a meal (ScienceDaily). These are conference results, not yet peer reviewed.
• Billing complaints. A law firm is recruiting users who say a "free trial" turned into an upfront annual charge (Zimmerman Reed). That is an investigation, not a ruling.
• Retention. Calorie tracking is notorious for churn. The founders expected 25 to 50% of annual subscribers to renew.
• An ownership dispute. An early promoter sued in January 2026 claiming a 25% stake; the founders dispute it (Daily Mail).
• Creative ethics. Creators used a special build showing long streaks so it "looks good on camera". If you copy this, disclosure rules still apply.

What you can copy, and where we help

You do not need Cal AI's budget to use its structure. You need a product people can see working in a few seconds, a repeatable way to put it in front of the right creators, and a funnel that converts the traffic.

Onboarding and paywall changes are product work. We can plan the creative and the tests around them, but your team owns the build. For the visual breakdown we used to plan this page, see the Cal AI teardown board on Miro.

Before you brief us