Free calculator
AI screening cost calculator
Every AI stock tool has done this multiplication privately and none of them publish it, because the honest version explains why the universe on offer is smaller than the one you wanted. Here it is, with sliders.
The universe you want an opinion on. A global equity screen is 8,000 plus; most products in this category quietly cover a few hundred.
Fetch, parse, derive, compare to last period, look for one-offs, write it up. Six is the floor for anything you would act on.
Cents. A short grounded call on a mid-sized model, mid-2026. This has fallen by an order of magnitude in two years and may again, which is the one number here worth arguing about.
22 is every weekday. 1 is monthly, which is what most tools do while describing the result as current.
How much of a repeat pass is answered from what you already computed. Generously modelled: the cache discounts every pass after the first.
Changes the clock, not the bill. Worth knowing, because the clock is what decides whether a company page can be read on demand.
It is a straight line. There are no economies of scale in the model calls, because they happen once per company and do not amortise. That is the whole economics of the category in one shape.
The trade nobody states out loud: the same budget buys 2,500 companies at 6 checks, or about 625 companies at 24 checks. Breadth and depth are the same money. When a tool tells you its universe, it has told you which side of that it chose, and it is usually a more honest signal than anything on the pricing page.
Your numbers, the curve, and the note on what the same budget buys in depth.
No spam, no selling your address, unsubscribe in one click. The tools stay free either way.
Straight answers
Why does covering more companies cost proportionally more?
Because the expensive part is per company and does not amortise. Fetching and parsing get cheaper at scale, but the model calls that read each company happen once per company however many you cover, so the cost line through the origin is close to straight.
What is a realistic number of checks per company?
Six is the floor for anything you would act on: fetch the filings, extract the statements, derive the ratios, compare against the previous period, look for one-off items, and write the result. Products doing one or two are summarising rather than analysing.
Does caching solve this?
It reduces repeat passes, not the first one, and it introduces a different problem: a cached answer has an age. This calculator models the cache generously, discounting every pass after the first, and the bill is still substantial at a wide universe.
How can I tell what universe a tool actually covers well?
Ask it about a company outside the obvious list: a mid-cap listed outside the US, or a recent listing. A tool that is excellent on the first ten names you try and vague on the eleventh is pre-generating answers for popular companies, which is a legitimate strategy and almost never disclosed.
Is the cost per check figure realistic?
It is a defensible order of magnitude for a short grounded call in mid-2026 and it is a slider precisely because it is the least stable number here. It has fallen by an order of magnitude in two years and may again, which would change the conclusion without changing the structure.
Educational information, not financial advice. Figures current as of July 2026 where dated; allowances and rates change, so check the source before acting.