What Is Bolt.new? Revenue Evidence From 40 Projects
Bolt.new is an AI coding tool used to build or support 40 projects in the ProvenStartups index. The useful answer is not that it makes app creation…
Bolt.new is an AI coding tool used to build or support 40 projects in the ProvenStartups index. The useful answer is not that it makes app creation faster; it is that Bolt appears in real products with outcomes ranging from negligible revenue to verified scale. We would use it to test a narrow paid product, not as proof that a market exists.
Contents
This article separates the tool from the business claim, then checks the full Bolt cohort, comparable outcomes, the main contradiction, and a build decision. Use the links below to skip directly to the evidence or the execution rules.
What Is Bolt.new?
Bolt.new is a vibe-coding product, not a business model. It can be part of the implementation layer, while demand, distribution, pricing, retention, and support remain the founder’s problem. For product instructions, use Bolt.new’s official help center; for business selection, use evidence from shipped products rather than tool demos.
That distinction matters because revenue does not come from generating an interface. The AI Chat Agent case, for example, had a monthly-retainer model but disclosed no revenue number [C]. A technically plausible offer is still unproven when the customer and payment evidence are missing.
The opposite is also visible. The Influencer-Partnered App Venture Studio reached about $67K/mo [V] from one product. Its useful lesson is distribution through a partner, not that an AI coding tool automatically produces that outcome.
ProvenStartups therefore treats Bolt as one attribute in a record. The full project index contains 406 graded ideas, including 266 software or SaaS products and 38 cautionary tales. The evidence class beside a result matters more than the stack named in the build story.
What the 40-Project Cohort Shows
The full matching cohort contains 40 Bolt projects, and 25 are solo-run. Only 10 publish a clean monthly figure, with a median of $67K/mo and a range from $37/mo to $2M/mo. Those are cohort aggregates, not a promise, and an aggregate does not receive an evidence grade of its own.
The mix is broader than “micro-SaaS”:
- ·9 SaaS products and 7 consumer apps
- ·5 AI services, 5 scale references, and 5 cautionary tales
- ·3 digital-publishing products and 3 platform plugins
- ·2 AI websites and 1 ecosystem tool
This distribution says Bolt is used across business shapes, but it does not show that every shape is equally attractive. A solo founder should care that most records do not publish a clean monthly result. Missing revenue is not zero revenue, but it is also not evidence of traction.
The supplied cohort evidence summary records 5 cases as third-party verified [V], with no cases recorded as [F], [C], or [U]. That tally covers fewer records than the 10 clean-monthly cases, so ProvenStartups does not silently assign grades to the remainder. The grading method explains why source quality and revenue magnitude are separate fields.
How Bolt-Related Outcomes Compare
The named cases do not support one standard Bolt outcome. Verified consumer products can reach $2M/mo [V], while an unverified launch reports $12K in its first month [U]. The comparison below is useful because it keeps the number, evidence class, and business type together instead of presenting unrelated screenshots as equivalent proof.
| Case | Reported outcome | Evidence | Category |
|---|---|---|---|
| Viral App Monetization Machine | Cal AI and Lerna: $2M/mo each [V]; three other apps: $700K/mo each [V] | Third-party verified | Consumer App |
| Influencer-Partnered App Venture Studio | About $67K/mo [V] | Third-party verified | AI Service |
| 4-Day AI App Pricing Blueprint | $12K in the case study’s first month [U] | Unverified | AI Website |
| Base44 | $1M ARR in 3 weeks [V]; $80M exit [V] | Third-party verified | Scale Reference |
| Replit | About $160M ARR [V] | Third-party verified | Scale Reference |
| Unblocked Games Site | $15K/mo [C]; sold for $120K [C] | Creator-relayed | AI Website |
These are not six interchangeable templates. Base44 and Replit are scale references with difficulty rated 5/5, while the unblocked-games case is an advertising and site-flip model rated 2/5. We would compare a proposed product with the closest business mechanics, not the largest number in the table.
Where the Data Contradicts the Hype
The popular claim is that Bolt leads the vibe-coding category. ProvenStartups’ data says it does not: among dedicated AI coding tools, Bolt’s 40 projects trail Claude Code’s 50 and Cursor’s 46. Bolt is still third, ahead of Lovable at 19 and Replit at 12, across the site’s 211 distinct tool-mentioning projects.
ChatGPT appears in 100 cases, but it is a broader assistant bucket and is not used for that dedicated coding-tool order. “Second only to Claude Code and Cursor” is therefore the precise claim. Calling Bolt the category leader would be cleaner marketing and worse analysis.
The revenue data also rejects the usual “build in a weekend, then collect MRR” story. In the full site dataset, only 106 of 406 cases publish a clean monthly figure. Of those, 8 are below $1K/mo, while 26 exceed $100K/mo. The distribution has real wins, but publication bias and missing numbers make the middle impossible to ignore.
What We Would Build, and Refuse to Build
We would use Bolt for a narrow product with a reachable buyer, one paid action, and a manual fallback. We would refuse to build another broad “AI does everything” app, a content farm dependent on platform tolerance, or a clone justified only by someone else’s revenue. Tool speed cannot repair weak distribution.
Our build filter is short:
- 1.Name the buyer and the painful recurring task.
- 2.Predefine the payment event before generating the product.
- 3.Ship the smallest workflow that reaches that event.
- 4.Keep fulfillment manual until repeated demand appears.
- 5.Record revenue and source quality separately.
The AI KDP books case is easy to build at difficulty 1/5, yet its benchmark of about $500/mo per qualifying book is unverified [U]. Low implementation difficulty does not remove marketplace, ranking, or proof risk.
We would also reject attractive topline numbers when the operating model is hostile to a solo founder. The cross-border Ozon store group reported about RMB 37K/mo profit [F] on more than RMB 400K/mo GMV [F], but ProvenStartups files it as a cautionary tale. Revenue evidence is an input, not permission to copy.
FAQ
The short answers are simple: Bolt.new is a build tool, the cohort is evidence of use rather than guaranteed success, and source quality changes how much weight a revenue claim deserves. A solo founder should choose a market first, then decide whether Bolt is the cheapest path to a paid test.
Is Bolt.new itself a startup idea?
No. Bolt.new is infrastructure for executing an idea; the startup is the organization and business built around solving a market problem. Wikipedia’s startup entry provides the general definition. For selection, define the buyer, offer, payment event, and acquisition channel before treating generated software as a company.
How much do Bolt-built projects make?
Among the full 40-project cohort, the 10 cases with clean monthly figures range from $37/mo to $2M/mo, with a $67K/mo median. That median describes the matching set, not expected earnings. A concrete high-end reference is the Viral App Monetization Machine’s $2M/mo per leading app [V].
Is Bolt.new a good fit for a solo founder?
It can be: 25 of the 40 matching projects are solo-run. That supports Bolt as a plausible solo execution tool, not as a shortcut to demand. We would use it when one founder can sell, operate, and support the narrow workflow without relying on a large team or hidden manual labor.
What should I verify before copying a Bolt project?
Verify the revenue period, source class, business model, acquisition channel, operating burden, and whether the result belongs to the product or a broader portfolio. Treat $12K in a first month [U] differently from about $67K/mo [V]. Similar-looking numbers can carry very different confidence and repeatability.