Short answer, as of October 2026: the number of prompt-checks you need each month is prompts × engines × runs per prompt × measurement rounds. A single business tracking 15 prompts properly across six engines uses roughly 200 a month; an agency with ten clients on the same pattern needs about 2,000. On DigiRank Expert that maps to the $99/mo Starter plan (200 checks) and the $249/mo Agency plan (2,500 checks) respectively.
A prompt-check is one prompt, sent to one AI engine, one time. Almost every AI visibility tracker meters on some version of it, and almost every buyer underestimates how quickly the multiplication runs away. This guide is the arithmetic, done once, so the plan you pick fits the tracking you intend to do.
The formula
Checks per month = prompts × engines × runs × rounds
- Prompts — the questions in your set.
- Engines — the assistants you check each question in.
- Runs — how many times each prompt is asked per round. AI answers vary, so one run is a sample, not a measurement. The reasoning is in why AI answers change between runs.
- Rounds — how many times a month you repeat the whole exercise.
Each factor multiplies the others. Doubling the prompt set and going from monthly to weekly rounds is an eightfold increase, not a fourfold one, once you count the extra weeks.
Why runs are the factor people forget
A tracker that asks each prompt once will show a tidy grid of "cited" and "not cited". For any prompt where you are one of several plausible answers, that grid flips from week to week without anything changing. The fix is repetition: three to five runs per unbranded prompt, so you can tell whether you are named most of the time, about half the time, or rarely.
Branded prompts, the ones that include your name, are the exception. They are stable and one run is usually enough. That difference is the main lever for saving budget, and it is set out in branded vs unbranded prompts.
Worked budgets
All figures assume six engines unless stated.
| Scenario | Branded prompts (×1 run) | Unbranded prompts (×3 runs) | Rounds per month | Checks per month |
|---|---|---|---|---|
| Solo business, minimal | 5 | 6 | 1 | 138 |
| Solo business, sound | 6 | 9 | 1 | 198 |
| Solo business, three engines only | 6 | 15 | 1 | 153 |
| In-house team, two product lines | 8 | 20 | 2 | 816 |
| In-house team, five brands | 30 | 60 | 2 | 2,520 |
| Agency, 10 clients on the "sound" set | 60 | 90 | 1 | 1,980 |
| Agency, 15 clients, fortnightly | 90 | 135 | 2 | 5,940 |
How to read a row: the "sound" solo set is 6 branded prompts × 6 engines × 1 run (36), plus 9 unbranded prompts × 6 engines × 3 runs (162), for 198.
Two things stand out. The solo rows all fit inside a 200-check allowance, but only just, and only at one round a month. And the agency rows show why per-client arithmetic matters: fifteen clients measured fortnightly is nearly six thousand checks, which is a different plan from ten clients measured monthly.
Matching the budget to a plan
DigiRank Expert's allowances, from the pricing page:
- Starter, $99/mo — 1 brand, 1 campaign, 200 prompt-checks a month. Fits the solo "sound" set with two checks to spare.
- Agency, $249/mo — up to 15 client tenants, 10 campaigns, 2,500 prompt-checks a month. Fits ten clients on the sound set, with about 500 left for re-tests.
- Pro, $499/mo — up to 5 brands, unlimited campaigns, 10,000 prompt-checks a month. Fits the five-brand in-house row nearly four times over, which is what makes weekly rounds and deeper prompt sets practical.
- Scale, $999/mo — unlimited tenants, campaigns and prompt-checks.
Agency, Pro and Scale include a 14-day trial; Starter does not. A fuller walk through what each plan contains is in DigiRank Expert pricing plans explained.
Whatever tracker you are evaluating, do this sum before you compare prices. A plan that is cheaper per month and half as generous per check costs more for the same measurement.
What to cut when the allowance is short
Cut in this order. Each step saves a lot and costs little.
1. Rounds. Go from weekly or fortnightly to monthly. Month-on-month is the comparison that carries meaning anyway; weekly lines mostly show noise. This alone halves or quarters usage.
2. Runs on branded prompts. Take them to one run. They rarely vary.
3. Engines. Keep the two or three your customers use and pause the rest. Which those are depends on your audience, and the way to decide is in which AI engines matter for your business. Halving the engines halves the bill.
4. Near-duplicate prompts. "Best accountant for freelancers" and "top accountant for freelancers" return much the same answer. Keep one.
5. Prompts you already own. If you have been named in nearly every run for three months, move that prompt to a quarterly spot check.
What not to cut: runs on unbranded prompts. Dropping from three runs to one saves two thirds of the budget and makes every remaining number unreliable. It is the false economy of AI tracking.
What to add when you have room
Spend spare checks in this order.
1. Re-tests after a change. When you rewrite a page, run its prompts five times before and five times after. This is the only way to learn whether the work did anything, and it is where a reserve earns its keep.
2. A competitor's branded prompts. Seeing how assistants describe your main rival is cheap and informative.
3. More intents, not more phrasings. Add a cost question, a comparison question and a problem question for a service line you were not covering. How to choose them is in how to build a prompt set.
4. Local variants, if you serve several areas and the answers differ by place. Multi-location tracking multiplies quickly; the way to keep it under control is in multi-location AI visibility tracking.
Agency allocation
For agencies the allowance is a shared pool, and it needs a rule.
Size each client's share to its prompt set, not evenly. A single-service local business needs the 198-check set. A client with four service lines needs more. Write the number into the client's scope so it is not renegotiated monthly.
Hold back about ten per cent. Unallocated checks pay for the before-and-after tests that prove your work, and for the new client who starts mid-month.
Use rounds as the service tier. Monthly measurement on a standard retainer, fortnightly on a premium one. It is an honest difference in cost and the client can see it. Pricing the service around it is covered in packaging a GEO service line.
Review quarterly. Retire owned prompts, rotate in new ones, and re-check that the pool still covers the client list.
A two-minute planning exercise
- Write down the prompts you would track today. Mark each branded or unbranded.
- Count the engines you care about.
- Multiply: branded × engines × 1, plus unbranded × engines × 3.
- Decide monthly or fortnightly, and multiply again.
- Add ten per cent.
The result is the allowance to shop for. If it is far larger than you expected, the prompt list is too long, and trimming it will improve the tracking as well as the cost. If it is under 200, a single-brand plan is enough to start. The tracker module and its Opportunities tab, which lists the prompts where you are missing, are described on the features page.
Frequently asked questions
What is a prompt-check? One prompt, sent to one AI engine, one time. It is the unit most AI visibility trackers use to meter usage. Tracking ten prompts across six engines once uses 60 prompt-checks. Running the same set three times uses 180.
How many prompt-checks does a small business need per month? Around 150 to 250 covers a sound small-business set: six branded prompts run once and eight to ten unbranded prompts run three times, across six engines, measured monthly. Tracking fewer engines or fewer prompts brings it lower.
How many prompts should I track for AI visibility? Most single businesses are well served by 12 to 20 prompts. A smaller set measured with repeated runs gives more reliable readings than a large set measured once, because AI answers vary between runs.
Is it better to track more prompts or run each prompt more times? Run each prompt more times. A single run cannot tell a brand named nine times in ten from one named four times in ten. Three runs per prompt is the practical minimum for unbranded prompts; add prompts only after that is covered.
Do I need to track all six AI engines? Not always. If your allowance is tight, track the two or three engines your customers use most and add the others later. Dropping from six engines to three halves the checks used without changing the prompt set.
How should an agency split a prompt-check allowance between clients? Give each client a fixed share sized to its prompt set, keep around ten per cent unallocated for re-tests after content changes, and review the split quarterly. An even split is simple, but clients with more service lines or locations usually need more.
How much does AI visibility tracking cost with DigiRank Expert? As of October 2026, AI tracking is included from the Starter plan at $99 per month with 200 prompt-checks. Agency is $249 per month with 2,500, Pro is $499 per month with 10,000, and Scale is $999 per month with no cap on prompt-checks.