In the snapshot from November 14, 2026, QuizAPI's monthly recurring revenue was $3.98, with 2 active subscriptions. On the same day, it was listed for sale on TrustMRR, had received 8 acquisition offers, and 515 interested buyers had recently viewed the page. [1]
What it does is concrete: developers can put a quiz into an existing page without building a question bank or writing grading themselves. Fetching questions, grading, embedding, and score callbacks go through a REST API and a script tag. [2]
A quiz has to pass five gates
For example, you want to add ten JavaScript multiple-choice questions to your own course page. The usual approach is to collect the questions yourself, write a grading function, and then put together an interface that can show progress and results. QuizAPI's path is: pull questions from the API, and a script tag renders the quiz directly into the page; when the user finishes, the host page receives quizapi:complete, containing the score, the total number of questions, and the time taken; for per-question detail there is also quizapi:answer, with the question number, selected option, and whether it was right or wrong. If you do not want to touch styles, switch to an iFrame, with zero configuration. [3]
The real work is on the question-generation side. The AI entry point on the homepage lets you generate a full quiz by topic, setting difficulty, number of questions, and format; according to the website, it produces content that can go live in seconds. [2]
Image: Source QuizAPI[2], accessed 2026-11-14, original interface retained.
But generation is not the end point. An engineering article dated April 8, 2026 laid out the full pipeline: the LLM first produces a batch of candidate questions; structural validation, where each question must have exactly 4 options and only 1 correct answer; answer verification, where conceptual questions call the model once more to check and code-output questions try actually running the code; then quality scoring, where questions scoring below 75 are regenerated or discarded; those that pass 75 enter a human review queue, and only after passing that gate are they published. [4]
Two details are worth pausing on.
First, they generate about 50% more than the target number and then filter, which the article says is cheaper and faster than repeatedly regenerating failed questions. This is the operator's explanation of their own process, not an independent cost audit, but it shows that what they fear is not the cost of a single generation, but the cost of failed retries.
Second, a shortcoming the article itself admits: difficulty is subjective, and no matter how the model is prompted, it tends to gravitate toward medium difficulty; hard questions still have to be calibrated by humans. And if you ask the same model with the same settings to check the answers, it will agree with itself; changing the model or temperature makes it easier to catch errors. The step in this pipeline that looks least like AI is precisely the step where it most wants to prove it is reliable. The homepage still says 'Now in public beta'.
Who is paying
TrustMRR writes the target audience as developers. The product shape points to the same group: developers who want to put a quiz into their own site but are unwilling to maintain the question bank and grading logic themselves. As for who actually presses the subscribe button, the public profile does not say; it only shows 2 active subscriptions.
The three monthly tiers are $0, $1.99, and $9.99, split mainly by usage: API calls from 500 to 100,000 per month, and AI generations from 3 to 100. The most expensive tier says 'unlimited number of quizzes,' but calls and generations are still capped—it charges by usage, not per person. [2]
Facing the same thing—'generate a quiz and then use it'—MagicForm's public path is to generate from materials such as text, a topic, YouTube videos, or PDFs, then edit and export to Google Forms, PDF, or a share link. MagicForm[5] Both can generate; the difference is where delivery happens: one hands the quiz to you, the other embeds the quiz into your page and then hands answer events to your own code. If you choose the script tag, the quiz renders on the customer's page and the score is received by the customer's code; but the platform's leaderboard and answer analytics also record these responses—the data does not land on only one side.
The customer-acquisition actions are also there: the blog has a stack of integration tutorials—Laravel, Next.js, Vue, embed guides, tracking student progress with webhook. Blog[6] This is a typical developer content entry point, but how much traffic, how many sign-ups, and how many payments the blog brings in is not in the public material.
Beside $3.98
Back to that set of numbers. As of the snapshot dated November 14, 2026, MRR was $3.98, active subscriptions were 2, and MRR growth in the past 30 days was 0.0%; revenue in the past 30 days was $3.98, for the past 12 months $84.72, and all time $270. In the monthly records, June and July 2026 received $17.57 and $19.39 respectively, and the monthly records from August through October were all $3.98. These are recorded by the Stripe interface and are payment totals, not profit; bills for LLM calls, hosting, human review, and payment processing fees are not disclosed, so whether this business makes money cannot be seen from this page. [1]
The revenue multiple column in the profile says 0.0x. This page does not give an asking price for the listing, so this multiple cannot be used to judge whether it is expensive or cheap.
The part about people is sparse: founder Bobby Iliev, location Bulgaria, founding date March 2, 2020. There is no statement from him about the reason for listing, operating condition, or sale progress, nor information on time invested or who is behind the 2 active subscriptions. Whether the listing is a voluntary exit or something else cannot be read from it.
From the public documentation and articles, several choices can be read: high-scoring questions are placed in a human review queue before publication; the lowest paid tier is set at $1.99, split by monthly API calls and number of AI generations. As for why the founder set it this way, the public material has no explanation from him.
There is a mismatch worth pointing out. This pipeline solves 'AI-generated questions are unreliable,' while a developer who wants to add a quiz to a page may first need to solve only 'I do not want to maintain the question bank and grading myself.' The former is a quality problem; the latter is a workload problem. QuizAPI touches both: the question bank and grading go through the API, and AI question generation gets a separate quality pipeline.
This pipeline splits the accuracy of a question into multiple steps; and the currently public monthly recurring revenue is $3.98, corresponding to 2 active subscriptions.