We checked the "$0 ad spend" AI growth stories. Here is what broke.

What "checked" means here

zero2first stores every case in the same five fields: startup cost, launch channel, GEO & AI-search visibility, the source of the first paying customer, and a 30-day playbook. On top of those, every number carries an evidence level.

The four levels are deliberately unflattering. Multi-source means a filing, a press announcement, or independent reporting corroborates the figure. Self-reported means the founder or company said it and nobody audited it. Disputed means the numbers or the narrative are publicly contested. Thin means a single interview or post and nothing else.

Nothing in the index is paid placement, and no case is removed for being awkward. That is the whole point: a story you can dissect is worth more than a clean story you cannot check.

Base44: the deal is real, the framing is not

The headline number — "$80M, six months, solo founder" — is the most repeated fact in this space. The deal itself is the one thing in my entire index that reaches multi-source: Wix's SEC filing puts total consideration at $92.158M, made up of $18.058M in cash and $74.1M contingent, with $25M of that tied to employee retention.

So the "$80M" is not the total, and the rest of the framing is weaker than the number. At close the team was 6 to 8 people, not one. The founder was a repeat founder who had previously co-founded and run a company that raised roughly $125M — which is real capital, just not capital in this company.

The growth claims are the part that never leaves the founder's own posts: 10,000 users in three weeks, a $189K profit in May 2025. Those are self-reported. They may well be true. I could not find them anywhere the founder does not control, and that is the difference between a number you can build on and a number you can only repeat.

Kuse.ai: when "ARR" is not ARR

The second case is the cleanest example of a number migrating. The revenue figure moved across two public statements: $9M in an August 2025 press release, then $10M+ on a podcast in November 2025. Both were company statements, so both are self-reported — but the direction of travel is what made the claim famous: "$10M ARR in 60 days, no VC, zero marketing spend."

Then, in August 2026, the company's own CEO said the quiet part out loud: people are calling GMV "ARR". If that were e-commerce, he noted, $10M would not be a big number. That single sentence is worth more than any growth chart, because it tells you what the denominator actually was.

Two other details matter more than they look. First, the most-cited article about the company misidentifies the founder — it names the head of marketing, in what was a contributor piece rather than reported journalism. Viral copy propagates that error, including into AI answers. Second, "zero marketing spend" turned out to describe zero *ad* spend: the actual distribution was a Threads account matrix running hundreds of accounts, which is labour, not zero.

Swan AI: one interview, no second source

The third case is included precisely because it is thin. The claims are ~$1M ARR, $0 customer acquisition cost, and a three-person team, all from a single podcast interview. There are no revenue documents and no third-party corroboration, and even the timeline is stated inconsistently — nine weeks in one telling, six months in another.

I kept it because the method is representative of a real 2025–2026 pattern: the founder's own account is the channel, and content is produced in bulk by an agentic workflow. The method is worth studying. The revenue figure should not be quoted as a benchmark, and in the index it never is.

The three patterns that keep showing up

Across the three cases, the same three things break the story, in the same order.

• —"Bootstrapped, no outside capital" hiding prior resources — a previous exit, a funded company, or an audience built elsewhere. The company may be bootstrapped; the founder is not starting from zero.
• —Revenue claims that change between public statements. Not always downward, and not usually dishonest — but a number that moves is a number with a definition problem, and the definition is usually the interesting part.
• —A "solo founder" who already had the audience. Distribution that looks like a product insight is often an audience that took years to build somewhere else.

A 20-minute check you can run yourself

You do not need a research team to do this. Four steps catch most of it.

• —Find the earliest public claim, not the most recent summary. Summaries inherit errors; the original post or press release usually still exists and tells you what was actually said.
• —Ask whether the number appears anywhere the founder does not control. A filing, a regulatory document, an independent report, a customer disclosure. If every mention traces back to one post, it is self-reported no matter how often it is repeated.
• —Find the denominator. ARR, GMV, gross bookings, and revenue are four different numbers, and founders are rarely lying when they pick the flattering one — they are just picking it.
• —Check who was actually on the team on the date in question. Headcount at close, co-founder credits, and prior companies are usually public and take two minutes to confirm.

Why the disputed cases stay in

It would be easy to publish only the three cleanest stories and call it a curated index. It would also be useless, because the stories people actually repeat are the messy ones — and the value is in knowing which numbers you can build a plan on.

So the index keeps disputed cases with their labels attached. If a founder sends better sourcing, the case gets upgraded and the change is noted. If a figure turns out to be wrong, it gets corrected rather than quietly deleted.

Three cases and zero verified revenue figures is a thin starting point. It is also an honest one, and it is the reason the index gets more useful as it grows.