Apps Built With Gemini That Have Revenue Evidence
ProvenStartups tracks 26 projects that name Gemini somewhere in the build, and 19 are solo-run. Twelve disclose a clean monthly figure, from $300/mo to…
ProvenStartups tracks 26 projects that name Gemini somewhere in the build, and 19 are solo-run. Twelve disclose a clean monthly figure, from $300/mo to $600K+/mo, median $12.5K/mo. The honest headline: almost none were *built* with Gemini. Here it is usually the model inside the product, not the agent at the keyboard — and that changes what you should copy.
Contents
How many apps built with Gemini have revenue evidence?
Twenty-six, and the evidence mix is thin at the top. Three are third-party verified (✅), fourteen are founder-reported (🗣), eight are creator-relayed (📎), and one is unverified but high-potential (🔮). Nineteen of the 26 are solo or one-person operations. Read the grade before you read the number.
Those grades carry weight. ✅ means someone outside the business saw the figure — Mine Marketing refreshed QuickBooks live on stream to show $140K/mo. 🗣 means the operator said it and nobody checked. 📎 means a creator repeating a number they never saw. 🔮 means unproven.
The category spread is wider than "Gemini app" suggests: six SaaS products, four consumer apps, four AI content businesses, three AI services, three digital publishing plays. For scale, the Built With AI hub holds 50 Claude Code cases and a comparable Cursor cohort — read the Gemini set as a sample, not a census.
Which Gemini projects report the clearest revenue?
These nine publish the cleanest figures in the set. Three carry third-party verification; the rest are founder- or creator-reported. Do not average the column: a $600K+/mo video SaaS and a $300/mo bedtime-story email are not points on the same curve, and one figure covers a single month.
| Project | Reported revenue | Evidence | Tier |
|---|---|---|---|
| Revid | $600K+/mo | 🗣 Founder-reported | Tier 2 |
| Mine Marketing | $140K/mo, under 50 clients, 6 people | ✅ Verified | Tier 2 |
| nano-banana.ai | ≈$115K/mo net profit, single month | 📎 Creator-relayed | Tier 1 |
| Prayer Lock | $21K/mo, 58K downloads in 6 months | ✅ Verified | Tier 1 |
| BlogToPin | $16K MRR, 400+ subscribers, ~10% churn | 🗣 Founder-reported | Tier 1 |
| Aura | $15,000 MRR, 21,700+ users in ~1 month | 📎 Creator-relayed | Tier 1 |
| Max's 28-app portfolio | $10K/mo, 1,000+ subscriptions, 4–5K DAU | 🗣 Founder-reported | Tier 1 |
| Arborist exam prep | $5K+ MRR, $31,000 in 8 months | 🗣 Founder-reported | Tier 1 |
| Dreamtales | ~$300/mo, 33–34 paying subscribers | ✅ Verified | Tier 1 |
The three ✅ rows deserve the most weight, and they are the least glamorous entries: a phone-blocker app, an agency selling websites to local businesses, and a bedtime-story email with 33 subscribers. The $600K+/mo top line rests on one founder's word.
Is anyone actually shipping with Gemini CLI?
No. Not one of the 26 cases names Gemini CLI, Antigravity, or Jules as the tool that wrote the product. Gemini appears in four other roles instead, and sorting the cohort by role is the only way this set makes sense.
1. Gemini as the product's engine. The largest and most useful group. Prayer Lock runs Gemini Flash 2.5 for the prayers, picked because it is cheap and fast. BlogToPin uses Gemini for pin text, fal for images. Stamp Identifier sends the photo to "OpenAI or Gemini — either works," which tells you how much the model choice mattered.
2. Gemini as the image supply. Nano Banana, Google's image model, does the work here — not Gemini the assistant. That covers nano-banana.ai, the seek-and-find and learn-to-draw books on KDP at an estimated $711–800+/day across three titles, and the historical time-lapse channel, where Nano Banana Pro makes the frames.
3. Gemini as one model among several at the keyboard. Revid's founder uses Cursor for code, GPT for complex logic, Gemini Pro for UI. Aura's method: start with Claude, switch to Gemini or GPT when a task fails. The anonymous 3D web studio takes first-pass landing pages from Google AI Studio, then hands backend work to Claude Code or Codex. Clarvo ships a GEMINI.md beside its CLAUDE.md — a hedge, not a commitment.
4. Gemini as the market. AEO Service bills $2,000/mo to audit whether a client appears in ChatGPT and Gemini answers. Distribute.io seeds Reddit so LLMs cite you, at 76 paying users and roughly $10,000 MRR. Chris Barrera's channel treats Gemini as a threat: YouTube uses it to detect reused AI voices, which he calls a demonetisation trigger.
Which three cases are worth reading in full?
Prayer Lock, the arborist exam app, and nano-banana.ai. They mark the three honest shapes in this set: a verified consumer app with a boring model bill, a weekend build that kept paying, and a domain snipe nobody can repeat. Two are copyable; one is a lottery ticket you already missed.
Prayer Lock reached $21K/mo and 58K downloads in six months (✅). The app took three days to write. Everything else was distribution: he copied competitor Prayscreen, found one video format that worked, pumped volume, and squeezed onboarding. Gemini Flash is a line item, not the story.
The arborist exam app reports $5K+ MRR and $31,000 total in eight months, on $140 total lifetime cost (🗣). Eighteen apps shipped in that time; one hit — flashcards for a tree-certification test, built in a weekend with zero marketing. Sit with the $140.
nano-banana.ai posted roughly $115K/mo net profit in one month (📎). The founder registered the domain on August 14, 2025; Google launched Nano Banana on August 26 — beaten by twelve days with a thin image-generation wrapper. The asset was the name, and the window is closed.
What repeats across the cohort, and what fails?
The operators who make money treat Gemini as a swappable line item and spend their effort on distribution. The weak records are the ones whose whole thesis is the model. Nine of 26 sit in Tier 1; four sit in Tier 3, where the headline is a target, a valuation, or someone else's channel.
Model-agnosticism is the default here, not a precaution. Dreamtales chains Claude, GPT and Gemini behind one email; Legora hot-swaps Claude, Gemini, GPT and Mistral by design. When businesses build the swap in on purpose, betting your differentiation on one model is the opposite of what this data shows.
Be hardest on traffic-as-revenue. Humanize AI Pro pulls 2.6M organic visits a month at DR 48 with 1,700+ referring domains (Ahrefs) — and reports no revenue. The $200K–$500K/mo figure circulating about it is a creator's 1%-conversion arithmetic, not a bank statement. Same caution on the Modiversity clone: the $38,000/mo AdSense belongs to Modiversity, and the cloned n8n-to-Gemini pipeline has no revenue attached at all.
What we would copy: pick a paid problem with a reachable buyer, use the cheapest model that clears the bar, keep the swap cheap. Mine Marketing is the cleanest template — a Gemini prompt bulk-scrapes businesses with a Google profile and no website (27% of US businesses have none), then the site gets built and sent before the pitch.
More in Built With AI, including apps built with n8n and our AI coding tools revenue ranking.
Frequently asked questions
The short version: this cohort documents Gemini inside products far more often than Gemini building them. Use it to pick a problem shape and an evidence threshold, not a toolchain.
Can you build an entire app with Gemini?
Nothing in this 26-case set proves it. No project names Gemini CLI, Antigravity or Jules as the builder. Where a coding agent is named it is Cursor, Claude Code, Codex or Lovable, with Gemini brought in for UI passes or when another model fails. Gemini's strongest showing is as the API inside a shipped feature.
Which Gemini app has the strongest revenue evidence?
Mine Marketing, at $140K/mo with under 50 clients and six people, verified by refreshing QuickBooks live on stream. Prayer Lock is the strongest solo benchmark at $21K/mo with 58K downloads in six months, also third-party verified. Both beat higher headline numbers in the set, which rest on a founder's word alone.
Is Nano Banana the same thing as Gemini?
Nano Banana is Google's image generation model, distinct from Gemini as a text assistant or coding agent. Three cases here run on image generation rather than code assistance: a domain-snipe website, a print-on-demand book business, and a faceless YouTube channel. Grouping them under "built with Gemini" is technically defensible and practically misleading.
How much do these Gemini projects make?
Twelve of the 26 disclose a clean monthly figure. The range is $300/mo to $600K+/mo and the median is $12.5K/mo. Do not apply that median to the whole set: fourteen projects give no clean monthly number, and businesses willing to publish results skew upward, so the disclosed group is not a fair sample.
Should I pick Gemini over Claude Code or Cursor?
This dataset cannot settle that, and anyone claiming otherwise is selling something. These are 26 case records, not a controlled benchmark. What the cohort does show is that operators hedge: several keep instruction files for multiple models so they can swap providers when quality or availability shifts. Build for the swap, not the brand.