Page 11 of the deck gives the headline: the margin, and what flows to you. This page is the arithmetic under it, with every row marked measured or our estimate, because the difference matters more than either number.
The AI labs pay people by the hour to correct AI answers. Ours come free, as a by-product.
Worth ~$2,555 a year for each person who actually clicks, at 20 people in every 100 sign-ups clicking, of which we keep 50% for the first 10,000 sign-ups & 75% after that. Worked out year by year in section 6 below. The market itself is $2B to $7B a year today, & not a cent of it counts toward the subscription numbers on this page.
All 5 are built into the product. Not a cent of any of them appears anywhere below. The model stands on subscriptions alone, so every number on this page is a floor.
| Plan | Price | AI cost | Card fees | Profit / mo | Profit / yr | Margin after fees |
|---|---|---|---|---|---|---|
| $5 Certainizer™ | $5.00 | $0.33 | $0.45 | $4.23 | $50.76 | 85% |
| $9 Plus | $9.00 | $0.70 | $0.56 | $7.74 | $92.88 | 86% |
| $19 Pro | $19.00 | $0.70 | $0.85 | $17.45 | $209.40 | 92% |
Card fees at the real published rate, 2.9% + 30¢ a charge. The $9 plan is the thin one & is protected by the 1,000-search monthly fair-use cap already promised in our own marketing.
So 100,000 users means 25,000 on $5, 20,000 on $9 & 10,000 on $19. This split is the single biggest assumption in the model, far bigger than any cost figure. Substitute your own & everything below recalculates.
| Per year | 10K users | 100K users | 1M users |
|---|---|---|---|
| $5 Certainizer™ | $126,900 | $1,269,000 | $12,690,000 |
| $9 Plus | $185,760 | $1,857,600 | $18,576,000 |
| $19 Pro | $209,400 | $2,094,000 | $20,940,000 |
| Gross profit, subscriptions | $522,060 | $5,220,600 | $52,206,000 |
| Hosting & infrastructureest | ($14,400) | ($60,000) | ($300,000) |
| Support & moderationthe advisor bot is trained to handle most of itest | ($9,600) | ($36,000) | ($300,000) |
| Legal, tools, servicesest | ($9,600) | ($24,000) | ($120,000) |
| Core teamthe product is built; the growth team is paid from the payback pool, not salaryest | $0 | $0 | ($960,000) |
| OPERATING PROFIT / YEAR | $488,460 | $5,100,600 | $50,526,000 |
| same figure, per month | $40,705 | $425,050 | $4,210,500 |
| Reaching… | Year 1 of a ramp to that size | Every year after |
|---|---|---|
| 10,000 users | ~$265,000 | $488,460 |
| 100,000 users | ~$2,745,000 | $5,100,600 |
| 1,000,000 users | ~$27,335,000 | $50,526,000 |
* Tested, not guessed. The per-subscriber costs above come from real test runs, re-priced at the engine rates published on 30 July 2026. At the moment of writing the site has not launched & a few of the AI engines are not connected yet, so treat these figures as theoretical until live traffic confirms them. Per-subscriber economics are measured. How many users arrive, & how they split across plans, is our estimate, labelled as such on every row.
Built the same bottom-up way as the rest of this page: real people, real hours, never a slice of a market. This is the AI-lab data line from the 5 excluded above, worked out on its own, entirely separate from the subscription floor.
| Reaching… | Year 1 of a ramp to that size | Every year after |
|---|---|---|
| 10,000 sign-ups | ~$1,280,000 | $2,555,000 |
| 100,000 sign-ups | ~$18,500,000 | $37,047,500 |
| 1,000,000 sign-ups | ~$191,000,000 | $381,972,500 |
The math: ~$2,555 a year for each person who actually clicks (73 hours a year at $35 an hour, the bottom of what this work pays), 20 clicking people in every 100 sign-ups, & we keep 50% of that for the first 10,000 sign-ups & 75% after. Checked a second way, from the low end of Scale AI's own numbers: Scale reportedly makes about $2B a year from human correction work, ~45% of which goes to its ~240,000 contractors, about $3,750 a year each, close to our own $2,555 once the difference in hours worked is accounted for. The “Year 1” column applies a 50% ramp discount, a rounder & more conservative number than the ~54% average discount section 4's own subscription figures above already use. “Every year after” assumes the full sign-up count active for a whole year, same as section 4.