Open Exam Prep

A solo, AI-built free exam-prep site with 23,000+ exam pages

Tax and financial adviser Ran Chen used AI coding tools to mass-produce exam pages and practice questions, monetised with display ads on English search traffic, and ten months after launch the founder's own public dashboard lists the site in the ≥ 10,000-dollars-a-month tier.

Ledger card

BSelf-reported

Income: ≥ $10,000; Founder's own disclosure · revenue · monthly · 2026-10 · lower bound only; basis in sources
Status: Active; Rechecked 2026-10-18

Grade B · Self-reported: The founder's own public statement or a media report, with source and date; no platform data to check it against.

Source: Founder's public incubator dashboard (with the ≥ $10K/month tier) · checked 2026-10-18 · next recheck 2027-01

Ledger summary

AI-assisted draft · facts checked by an editor

Ran Chen's day job is tax and financial advice; Open Exam Prep is the side project launched in December 2025: a free exam-prep question bank for English-speaking users that by October 2026 had about 23,000 exam pages, 2.7 million practice questions and roughly 1.3 million monthly page views. Income comes from display ads. On the founder's own public dashboard the site is listed in the "≥ $10,000 a month" tier and states the method: the average daily ad revenue of the last seven full UTC days, times 30. The figure is self-reported, with no payment-provider or ad-network page a third party can check, so it is graded B and only the lower bound is recorded.

Decision notes

AI-assisted analysis · based only on this case's verified facts

1. What problem it solves: Test-takers want free, structured English-language practice, and existing sites are poor or paywalled
2. Who the customers are: English-speaking exam candidates; the payers are advertisers behind the display ads
3. Why customers pay: All study material and questions are free, monetised by display ads on search traffic; ten free AI interactions a day
4. Business model: Subscription · Advertising · Lead generation
5. What AI does in delivery: AI-native · AI content generation · Claude Design · AI coding tools (no specific product named)
6. What public data proves: ≥ $10,000 · revenue · monthly · 2026-10 · Founder's own disclosure
7. Limits of the income data: Grade B · Self-reported; The founder's own public statement or a media report, with source and date; no platform data to check it against; lower bound only; basis in sources; Third-party estimates, user counts, funding and price quotes are never recorded as income
8. What indie developers or small teams can take from it: Pick evergreen keywords, treat content as data, and one person with AI coding tools reaches 20,000+ exam pages; ads earn far more than subscriptions · replicability score 3/5
9. Biggest cost, acquisition barrier or limit on copying it: Income rests almost entirely on Google search traffic and two ad networks; mass AI-generated pages risk being judged low quality; Coding: Medium · Startup money: None · Finding customers: High · On camera: Not needed

Rows 1–3, 8 and 9 are distilled by an editor from the facts recorded for this case and are not the founder's words; rows 4–7 are generated from the case fields.

Timeline

2025-12Site launchedLaunch month on the founder's incubator board
2026-05Founder's video blurb puts the whole AI side business at five to six thousand dollars a monthCovers several sites, not this one alone
2026-10About 23,000 exam pages and 2.7M questions, ≈1.3M monthly page views; listed in the dashboard's ≥ $10,000/month tierFounder-run dashboard; method is the last 7 full UTC days' average daily ad revenue × 30 (grade B)

How it was done

Evergreen keywords only: a keyword-data tool is used to find exam queries where existing sites are poor, search-engine summaries have not yet covered the topic, or everything is behind a paywall.
Content as data: every exam is assembled from one set of templates, the data lives in the code repository, pages are statically generated with Next.js and hosted on Vercel, with no database and hosting mostly within free tiers.
AI coding tools for scale: one person with no employees, using AI to write the code and Claude Design for design, generating pages in bulk until the taxonomy ran to tens of thousands of pages.
Ads first: in the founder's experience ads earn more than selling leads, and leads more than subscriptions; all study material is free, and only AI interactions have a free allowance of ten per day.
Indexing by submission: pages are submitted through Google Search Console and IndexNow; a new site takes about three months before traffic starts.

AI tools used

Step	Tool
Designs the sites and builds them directly	Claude Design
Writes the code (the author says AI writes the Vercel and Next.js stack very well)	AI coding tools (no specific product named)

Failures and risks

Subscriptions failed: in a Zhihu answer the founder says a low-threshold top-up option brought in less than a hundred dollars in six months.
Income depends almost entirely on Google search traffic and two ad networks, AdSense and Adsterra; a policy change at any of them hits revenue directly.
Pages generated in bulk with AI carry the policy risk of being classed as low-quality or spam by search engines.
The playbook assumes English content and US traffic; Chinese-language sites earn far less per thousand views, so the income expectation does not transfer.
Every number comes from the founder: the dashboard gives a tier rather than an exact figure, and the "five to six thousand dollars a month" in a May 2026 video blurb covers several sites combined.

Can it be copied?

3/ 5

Editor's score out of 5, based on barriers, cost and how hard customers are to find.

Coding: Medium
Startup money: None
On camera: Not needed
Finding customers: High

Sources and evidence

Founder's public incubator dashboard (with the ≥ $10K/month tier) · Founder · Simplified Chinese · checked 2026-10-18
Founder's article on building content sites for ad revenue · Founder · Simplified Chinese · published 2026-10-02 · checked 2026-10-18
Founder's reposted Zhihu answer (paywall test result) · Founder · Simplified Chinese · published 2026-10-14 · checked 2026-10-18
Open Exam Prep home page stats (questions, exam pages) · Editor-verified · English · checked 2026-10-18

Verification notes (links, dates viewed and excerpts of at most two sentences) are kept internally and never republished. Pages cite only facts and quotes of at most two sentences, always linked; data a platform's terms forbid reusing is never used as a source.

Rechecks

Date	Status	Evidence grade
2026-10-18	Active	BSelf-reported

Next stepsRead the project guide →CompareCompare other cases in this track →