The one-glance read on who they are and how they grow. Each point is verifiable from the receipts above.
An AI homework-help app owned by Singapore-based D3 Dimension Technology, Zuoyebang's international arm, letting students photograph or type any question for step-by-step answers.
Creator-led and paid-reinforced: TikTok creator videos and app-install ads carry acquisition, not a founder's personal following.
As of 2026-07-12, monthly visits sit at about 114,600, even as third-party app trackers put Android downloads above 26 million by early 2026.
As of 2026-07-12, they're testing an "AI Study Buddy and Tutor" repositioning on Google, running alongside its long-standing proven-winner ads.
The website is shrinking while the real funnel, paid App Store deep links and TikTok creator clips, routes students straight into the app instead of through questionai.com.
The order the channels came online. Sequence is strategy: what they did first, and what they layered on once demand existed.
Estimated demand, the channel split behind it, and the keywords and referrers doing the work. Directional modeling, not audited analytics.
| Keyword | Volume | Weight | CPC |
|---|---|---|---|
| question ai | 13k | $1.10 | |
| spelling bee passage miranda | 540 | - | |
| bicopying convention computers | 0 | - | |
| after the storm by orhan pamuk | 0 | - | |
| "which statement best describes the claims made by president andrew jackson in this passage?" | 440 | - |
An estimated 114,600 monthly visits, down 42.7% over the last three months as of 2026-07-12, split across three dominant channels: Direct at 46.7%, Search Organic at 28.7%, and Referrals at 18.6%. The rest, social organic, paid search, generative-AI referrals, display ads, and email, together total under 6%. Given that most acquisition here runs through paid App Store deep links and TikTok clips rather than the homepage, that large direct share plausibly reflects students who already clicked an ad or a creator link off-platform and are re-opening a bookmarked tab, not pure brand loyalty.
The branded term "question ai" pulls about 12,800 monthly searches, dwarfing the demand behind the literal homework-question strings the site actually ranks for, like a 440-search/month query quoting a passage from Andrew Jackson's writing. Referrals concentrate on pearson.com and bettermarks.com, both online courseware platforms whose assigned homework likely sends stuck students looking up individual questions here. Its closest named rivals, gauthmath.com (called out directly in its own paid ads), solvely.ai, and quizard.ai, are all the same scan-a-problem-get-an-answer category fighting for identical query intent.
The specific pages earning their organic search traffic, and the pattern behind why they rank. Adapt the format, not the topic.
Their organic footprint is almost entirely the homepage: it alone accounts for roughly 79,600 of the ~102,000 monthly organic visits a keyword-rank estimate assigns to their top five pages, a near-total concentration in one URL. That keyword-based figure runs well above the ~32,900 organic-search visits implied by the channel-mix panel above, so read the two as different measurement lenses rather than one comparable number. The other four ranked pages are auto-generated answer pages hosting the literal text of individual homework questions, phrasing like a biology prompt about "free flow oxygen select apply laryngeal" or an assignment about "alcohol tolerance impacted," each pulling only a handful of visits on its own. The pattern is programmatic long-tail capture: every question a student has ever typed into the tool becomes its own indexed page, so the site ranks for thousands of impossibly specific, zero-competition search strings no manual content team would ever think to target.
Not traffic share. How much weight the growth system actually puts on each channel, with a one-line read on the role it plays.
For founder-led SaaS the breakdown shifts from ads to traction: where the first users came from, how the founder grows it in the open, and the compounding organic surface.
Question AI's traction reads less like an indie launch and more like a large download-chart campaign layered onto an already-built parent product.
International rollout happened in 2023 in the US and Indonesia, an overseas extension of Zuoyebang's existing China homework-help product rather than a from-zero founder launch. The earliest creator UGC in evidence, an October 2023 Indonesian-market TikTok clip pulling about 81,800 engagements, lines up with that entry window.
Sensor Tower's mid-2024 US AI-app download report ranked Question AI third overall and first among education apps, behind only ChatGPT and Microsoft Copilot, per third-party app-intelligence reporting dated 2024-07-30.
The biggest traction spikes visible in the evidence sit in April-May 2025 (the 840,000 and 316,000-engagement clips above) and cluster around back-to-school hashtags, suggesting seasonal creator pushes rather than one single breakout moment.
There is no founder posting in the open here: Question AI is run by a corporate operator, and its acquisition engine runs through paid creators rather than one person's audience.
Public filings name Hou Jianbin and Li Boyang, Zuoyebang's own founders, as directors of the Singapore entity behind Question AI, but neither has a verified personal account tied to this product, so there is no founder build-in-public track to point to.
The brand's own TikTok posts, math-trick clips under the Question.ai handle, draw modest, steady engagement in the 900-8,000 range, a house style rather than a personality-driven account.
Instead of a founder's following, creators are seeded directly. One August 2025 TikTok post carries a named promo code, "studyv," for unlocking premium features, an affiliate-style incentive run through third parties.
The compounding asset here is structural, not editorial: every homework question a user submits becomes a permanent, indexed page, and the paid stack runs alongside it purely to route people to the App Store.
Rather than writing comparison or how-to content, the product auto-publishes a results page per question (the pattern covered above), so the content library grows for free with every user session instead of needing a writer.
Its Google ads point to purpose-built pages like /free/chatgpt, /pdf/reader, and /math/ai_solver, extending paid capture beyond its own brand name into ChatGPT-adjacent and PDF-tool search intent, a demand-generation move rather than pure brand defense.
One proven-winner Google ad is headlined "Better Than Gauthmath, Free AI," bidding straight onto a competitor's brand name (expanded as Play 2 below), and a Meta video ad opens with "POV: So this is why everyone's deleting ChatGPT," hijacking incumbent fatigue as a curiosity hook (expanded as Play 3 below).
the homepage's only outbound links go to its own TikTok and Instagram, so the creator-code seeding above runs through direct outreach rather than a public partner program.
This founder-led SaaS also runs paid acquisition. Here are the live ads doing the work, each with the X-ray and a play you can adapt.
Why it works. Lets a raw, unscripted-feeling reaction carry the pain point before the product appears, mirroring how students actually vent online.
Opening with a raw emotional hook of crying over homework (0:00-0:03). Open with a low-fi, emotionally raw reaction to the exact frustration your product removes, before showing the product at all.
Why it works. Rides an established math-meme format to earn attention from students who know the joke, then pivots into a real step-by-step demo.
Meme-style 'POV' Hook (0:00 - 'The way my brother teaches me math') Open with a recognizable 'wrong way' meme from your niche, then use your product to reveal the actual correct way.
Why it works. Times an aspirational perfect-grades outcome to the start-of-year reset moment, when students are most open to a new habit.
Identity Labeling (0:04): 'SMART GIRL' text overlay. Time an aspirational outcome hook to a calendar moment your audience associates with fresh starts, like semester start.
Why it works. An unrelated, absurd physical-comedy clip earns watch time long enough for a plain speed claim to land before the app demo.
Bizarre Pattern Interrupt (0:00 - 'Question.AI can eliminate problems in seconds.') Pair a jarring, unrelated clip with a caption that reframes it as a metaphor for your product's benefit, then cut to the demo.
Why it works. Names the exact last-minute panic moment right before an assignment is due, when a student is most likely to reach for help.
POV meme hook at 0:00 ('POV: the work is due by the end of class but you barley paid attention 😂'). Write hooks with the same casual typos and POV phrasing your audience uses on their own accounts, not clean brand grammar.
Why it works. Pairs a familiar frustrated-crossing-out visual with comparison-style caption shorthand to set up a before and after against the app.
Binary framing with 'NO' (0:03) and 'YES' (0:04) overlays. Use a comparison-fragment caption like 'this feeling>>>' over a frustration visual, then resolve it with your product ranked higher.
The channels are not separate. They are one system where each stage feeds the next. Here is the read, then the plays to run tomorrow.
The loop starts on TikTok and in app-install ads and ends in the App Store, with the website's long-tail pages acting as a slow-burn backstop rather than the front door.
paid creators and brand TikTok clips build problem-recognition, while 19 concurrently active Meta and Google ads (as of 2026-07-12) push straight to the App Store listing, so most new users likely never touch questionai.com at all.
reported pricing, not independently confirmed on the live pricing page, runs a free tier plus roughly $8.99-10/month or about $60/year (about $4.99/month effective) for unlimited use, against a monthly cap near 30 solved questions, tying the paywall to volume of use rather than a feature gate.
against reported 2025 weekly-active-user levels in the low millions and a roughly $5-8 blended monthly price, even a low single-digit-percent paid conversion would put monthly recurring revenue in the high six to low seven figures, an order-of-magnitude estimate only.
there's no visible community or forum layer in the evidence and no personal founder audience, so the loop depends entirely on paid budget and third-party creators rather than any owned audience the company controls directly.
free-to-paid conversion rate, churn, CAC, and verified subscriber counts.
The proofQuestion AI turns every submitted homework question into its own permanent, indexed page, the mechanic behind the long-tail organic pattern covered above.
The adaptationIf your product processes unique user input, a converter, a calculator, a prompt tool, wire it to auto-publish a public results page per input the moment it's generated, with the input text itself as the page title. Start by picking your single highest-volume input type and shipping one templated page for it; the mechanism is that each page targets a search string only that exact user ever typed, so you rank for demand no manual content plan would ever list.
Cost: $0 · Time to signal: weeks · Works pre-PMF: yes
The prooftheir longest-running Google ad is headlined "Better Than Gauthmath, Free AI," bidding on a named competitor.
The adaptationidentify the one or two competitor brand terms your prospective customers already search, and run a small paid campaign with a direct comparison headline pointed at a comparison landing page you build first. First step: write one page titled "[Your Product] vs [Competitor]" before spending anything, then point a $50-100 test campaign at it.
Cost: under $500 · Time to signal: days to weeks · Works pre-PMF: conditional, only if there's a well-known rival your audience already searches for by name, and check the platform's trademark-bidding policy in your category first
The prooftheir top Meta ad opens with "POV: So this is why everyone's deleting ChatGPT," hijacking rival fatigue as a curiosity hook.
The adaptationmonitor incumbent-tool complaints, outages, price hikes, policy backlash, on X, Reddit, and TikTok in your category, and ship a same-day reactive short-form video framing your tool as the alternative while the moment is live. Post it from whatever account you already have; the mechanism is that people are actively searching for an alternative right at that moment, so timing beats production quality.
Cost: $0 · Time to signal: days · Works pre-PMF: yes, but only around a real, visible incumbent stumble, not as an evergreen format
The proofa TikTok creator post carries a named promo code for unlocking premium features, an informal creator-affiliate mechanic run without a public partner page.
The adaptationDM 10-20 small creators in your niche offering a personalized, trackable discount code in exchange for one organic-style video, no upfront cash. First step: pick creators whose recent posts already show your target customer in the comments, not just high follower counts.
Cost: $0-under $500 · Time to signal: weeks · Works pre-PMF: yes
The prooftheir reported pricing caps the monthly plan at roughly 30 solved questions while the annual plan is unlimited, tying the paywall to volume of use.
The adaptationif you're building freemium, gate on a usage counter tied to your core action (N free generations, exports, or lookups) instead of hiding features behind a paywall. Set the free cap at the point where a typical user has already felt the product work at least once, so the upgrade prompt lands right after value, not before it.
Cost: $0 · Time to signal: weeks · Works pre-PMF: conditional, needs at least a little usage data to calibrate where the cap should sit Not transferable at an earlier stage: the 19 concurrently active paid ads and the parent company's existing download-chart position and creator relationships assume a media budget and distribution most first-time founders won't have; the auto-page, creator-seeding, and usage-cap mechanics scale down to zero budget fine.
Every morning we take one company that is actually growing and break down where its customers come from: the ads still running after a year, the channel doing the real work, and the play you can run this week.
Systemaic · directional intelligence. Traffic, spend, and reach figures are SimilarWeb-style estimates and qualitative reads of public data, not audited numbers. Built on real public receipts.
Live ad libraries: meta ad library · google ad library
Launch archives: Hacker News: Two senators question Air Force nominee's SpaceX connections · Hacker News: Dems question Airbnb over Xinjiang rentals on land owned by
Traffic, spend, and revenue figures are estimates as noted in the report; the links above are the primary public artifacts.