One row per buyer question, one column per engine. From $5/mo.
An example board, the brand is invented, the shape is not. One row per question you configure, one column per engine your plan covers, and a straight answer in every cell.
| Buyer question | ChatGPT | Perplexity | Google AI | Gemini | Claude |
|---|---|---|---|---|---|
| cited | named | absent | named | cited | |
| named | cited | named | absent | named | |
| absent | named | absent | absent | named | |
| cited | cited | named | named | absent |
Four parts, every 7 days. The board above is the first of them. The other three are counted across a week of a customer’s own probes, so no public endpoint serves them and this console does not draw one.
A board is only as good as the questions on it, so the product reads your site and proposes three. These are the three it proposed for 3 real product sites, exactly as https://citerank.thecompound.tech/api/site-info answered on the capture date.
The third slot is <brand> alternatives and is meant to carry the name. The first two are category questions and are not: the subject they are built from is the first two words of the page title, and a product title starts with the product. Marked in caution ink above. A category question that names the brand measures whether the engine repeats a name the buyer already typed, which is not what this product measures, and the landing’s own example chips are unbranded for that reason.
This app drops the brand’s own words before it picks a subject. Run over the same three captured pages, its generator returns best system product tools, best make product tools, best email developers tools. Still a keyword pass over a page the visitor wrote, and the visitor edits it. It no longer asks an engine to repeat a name the buyer already typed.
| Monthly, in US dollars. Every plan re-asks every question on every engine it covers, once a week, and mails the report on a Monday. Cancel from the Stripe receipt. | ||||
|---|---|---|---|---|
| Plan | Price | Brands | Questions | Columns of the board |
| --- | --- | --- | --- | --- |
| SoloOne product, ten questions, the three engines your buyers actually open. | $5 | 1 | 10 | ChatGPTPerplexityGoogle AICompetitor share of voice |
| FounderThree products, twenty-five questions, all five engines. | $12 | 3 | 25 | ChatGPTPerplexityGoogle AIGeminiClaudeCited-source list |
| StudioSix products. The plan for people who ship more than one thing. | $19 | 6 | 50 | ChatGPTPerplexityGoogle AIGeminiClaudeCited-source list |
Probes run on a shared, scheduled browser pool rather than on demand, that is the design decision the price rests on, and the reason results land weekly rather than the moment you click.
There is nothing to sign in to. The dashboard link emailed on purchase is the only credential this product issues, which is why it is worth bookmarking and worth not pasting anywhere public.
Not a discount, a different bill of materials. Profound's $99 Starter tracks ChatGPT and nothing else. AthenaHQ's Essential is free across all five engines but meters you at one credit per AI response against $25 of starting credit, which makes it a trial of five engines rather than a plan that keeps answering. Otterly's $29 Lite covers four engines and sells Claude, Gemini and Google AI Mode as add-ons on top of it. Solo is $5 for ten of your own questions on ChatGPT, Perplexity and Google AI Overviews, re-asked every week with nothing metered; Founder is $12 with Claude and Gemini inside the plan rather than billed beside it. Prices read off each vendor's own pricing page on 13 August 2026.
No. Studio is $19 for six brands and one report each, six products on one bill, which is the case this plan exists for. It is also where a per-prompt budget starts to bite elsewhere: Otterly's $29 Lite is fifteen search prompts in one workspace, and six products have to share them.
Answer engines do not change their minds hourly, and a single run is genuinely unreliable, only about 2.2% of cited domains recur across three runs of the same prompt. A weekly cycle across every engine is what makes a trend mean something. If you need on-demand probing at volume, the enterprise tools are priced for it and are the right answer.
Cited means the engine linked a page on your domain, that is the state that sends traffic. Named means it typed your brand into the answer without linking you. Absent means neither: somebody else answered that question. The three are never collapsed into one "visible" figure, because they are three different jobs, named-without-cited is usually the one you can fix this week, and absent tells you whose page to go and read.
No. This tool measures; it does not change what an engine says. What changes the answer is being mentioned somewhere other than your own domain, and the useful half of the report is the list of pages the engine read instead of yours. That list is a work queue, but the work is yours.
The report says so. A probe that could not run is never reported as an absence of citation, because "not cited" and "could not ask" are different facts and reading one as the other will have you fixing something that was never broken.
Put the address you subscribed with into the dashboard-link page and it is emailed to you again, to that address and nowhere else. There is no password to reset because there is no password: the link is the credential, which is why it is worth bookmarking and worth not pasting anywhere public. A cancelled subscription still gets its link, the history stays readable.
From the receipt Stripe emails you, or by writing to citerank@thecompound.tech. Cancelling stops the next charge and stops new probes; your history stays readable. Two of those are arguments against buying. A page of answers that only sells is a page nobody believes. The cheapest plan is Solo at $5 a month; Founder at $12 is the cheapest that asks all 5 engines. Last answer recorded by the pool: 2026-09-12.