How CitedIndex sizes and prices a tracking job
One job for every tool, sized so its monthly number can be trusted to within 5 points, and priced on each tool's own pricing page.
Method 1 from
The job
A monthly AI visibility number you can trust to within 5 points takes about 80 buyer prompts a day, or about 160 a week. That is the job we price on every tool.
| Daily | 80 buyer prompts, each run once a day, 30 runs a month |
|---|---|
| Weekly | 160 buyer prompts, each run once a week, 4.33 runs a month |
| Engines | ChatGPT, Google AI Overviews and Google AI Mode |
| Answers a month | 7,200 daily or 2,079 weekly, on three engines |
| Brands | one brand tracked |
| Prices | the tool's own pricing page, US dollars a month, before VAT |
Each tool is priced on whichever of the two it runs for less, and the table says which. The daily job sees a month-on-month change of about 2 points; the weekly job only changes above about 5.
How the job is sized
The size is derived from published noise data (Peec, Sep 2026; checked against Graphite, Sielinski and Gumshoe). We did not measure that noise ourselves; we credit it and publish the arithmetic, so anyone can put in their own figures.
Inputs, Peec's variances for non-branded prompts, since buyer questions don't name a brand:
| Input | Value |
|---|---|
| Per-run variance σr² | Value0.07 |
| Between-prompt variance σb² | Value0.048 |
| Shared daily shock ρ | Value0.015 |
How close a month's share is, for P prompts run r times in the month, at 95%:
1.96 · √(σb²/P + σr²/(r·P) + ρ·σr²/r)
The smallest month-on-month change that stands out from noise, with the same prompts both months:
1.96 · √(2·(σr²/(r·P) + ρ·σr²/r))
| Prompts | Month within (± points) | Change seen above (points) | Answers a month on 3 engines |
|---|---|---|---|
| 25 a day | Month within (± points)8.9 | Change seen above (points)3.1 | Answers a month on 3 engines2,250 |
| 50 a day | Month within (± points)6.3 | Change seen above (points)2.5 | Answers a month on 3 engines4,500 |
| 80 a day | Month within (± points)5.1 | Change seen above (points)2.2 | Answers a month on 3 engines7,200 |
| 100 a day | Month within (± points)4.5 | Change seen above (points)2.1 | Answers a month on 3 engines9,000 |
| 50 a week | Month within (± points)7.7 | Change seen above (points)6.6 | Answers a month on 3 engines650 |
| 100 a week | Month within (± points)5.8 | Change seen above (points)5.6 | Answers a month on 3 engines1,299 |
| 160 a week | Month within (± points)5.0 | Change seen above (points)5.1 | Answers a month on 3 engines2,079 |
The fewest prompts that stay within 5 points are 82 a day and 158 a week; we round them to 80 and 160. Branded prompts need fewer, about 52 a day. Adding prompts narrows the number faster than adding runs.
The inputs, the minimums and every row are in the job-size CSV.
Each tool's job figure, with its plan, the plan's own price, its engines and the same job on all five engines, is in the job CSV.
Words we use
- Share: the share of a month's answers on one engine, to a brand's fixed set of questions, that name the brand.
- Trusted to within 5 points: the month's share lies within 5 points of the brand's share over every question its buyers could ask, 95% of the time. At a 25% share, 5 points is a fifth of a brand's presence, about the smallest change a team would act on.
- All five engines: the same prompts in ChatGPT, AI Overviews, AI Mode, Gemini and Perplexity, on the same plans. It gets its own median.
Which plan counts
The cheapest public plan whose printed terms cover the job, at its regular monthly price after any first-month offer. A plan sold only by the year counts at its monthly share. Free plans and trials don't count.
- Prompts: the job's prompts held against the plan's prompt allowance.
- Credits and answers: one prompt in one engine on one run is one answer. Credits count at the rate the page states; with no stated rate, the tool says too little to price.
- Engine add-ons: an engine sold apart from the plan is added at its own monthly price.
- Past the allowance: extra answers at the plan's own top-up price, or the next plan up when that costs less.
Euro prices are converted to US dollars at the rate printed under each chart.
The engine marks
Six marks: one per engine (ChatGPT, AI Overviews, AI Mode, Gemini and Perplexity), then all five at once. Each is the tool's grade in the engine grid: what its cheapest paid plan covers, read from its pricing page. A tool sold by quote only, or one still under review, shows as unknown, never as covered.
No figure
A tool gets no job price when every plan stops below the job, when only a quote covers it, or when its page says too little. It stays in the table with the reason, outside the median.
Who is in the figures
We list tools people look for: at least 200 US brand searches a month and named in at least 3 AI answers to buyer questions, from the 128 we track; or named in 15 of those answers alone. Software only in the figures: agencies and services that do the work for you are listed apart, outside every count and median. No content-writing tools.
A maker can also pay to enter the research. It is then read, priced and counted like any other tool. Paying changes no number, rank or word about it.
Pricing pages and AI crawlers
Once a month, from our own server, we fetch each listed tool's pricing page as a browser and as GPTBot, ChatGPT-User and PerplexityBot. We read its robots.txt and check that the price is in the page without scripts. The bot names come from our own address, so a site that checks a real crawler's address may treat it differently.
Price history
Each pricing page is read again at least once a month. Months before our first read come from the Internet Archive's copies of the same page, back to January 2024, read the same way. Each of those months is labelled "from archived pages" and links to its copy. A month is charted only when at least 5 tools have a reading for it.
Limits
- The job size rests on one vendor's published noise, checked against three others. From our third month, our own weekly sample tests it in this niche; if it puts the job more than 20% off, we size it again and date the change.
- The noise differs by brand and niche. The job is sized for a typical set of buyer questions, not for yours.
Reusing the data
Charts, tables and CSVs are published under CC BY 4.0; credit CitedIndex.
Spotted a wrong price?
Send the tool's page that shows the right one to hello@citedindex.com. We change the chart once that page confirms it.
Every change to the job and how we price it
Each change to the job or the rules, and what it moved.
- Version 1, : Each tool priced on one job, sized from published noise data to keep a month's share within 5 points: 80 buyer prompts a day, checked daily, in ChatGPT, Google AI Overviews and AI Mode, on the cheapest plan that runs it, engine add-ons counted, read from its own pricing page. A tool that only runs weekly is shown on 160 prompts a week, outside the median.
Reads
- 22 tools listed, each read from its own pricing page
- Prices read up to