What a listing must do, and what we refuse to do.
The measurement recipe behind every listing, written so you can reproduce it yourself — and the three rules that no amount of money is allowed to bend.
How a listing gets its citation status
Name one query. Every listing is measured against a single stated query — the one the product would want to be found for. A listing measured on no named query is not a listing, it is a banner.
Build the pool, and publish it. We run that query through a web search, expand it into fan-out sub-queries, and collect the unique sources. That short list is what an assistant would actually see. It goes on the listing, so you can see who you were up against.
Show the model the page, not the snippet. A measurement that never lets the model read the page can say nothing about the page. This is the single most important design decision here, and we got it wrong once before and published the correction.
Run it six times, once in reverse. Five runs in pool order, one with the pool reversed, and a cache-bypass nonce on every request. Reversing the order is how we test whether position, rather than content, is doing the work.
Count it against the right denominator. The number on a listing is citations divided by the number of runs we could actually read for that query — not by the total number of runs anywhere, and not by the number of runs the source happened to appear in. Two of the thirty runs in this study failed to parse and are excluded from both sides of the fraction, here and on every listing. Getting this wrong inflates or deflates every figure on the page.
Publish the count, including zero. A directory that only shows you the listings that won is an advertisement, not a directory. 26 of the 48 sources in our own sample were cited zero times, and every one of those zeros is published.
The model is gpt-oss-20b. It is independent of us, reproducible, and free to run. It is also not ChatGPT, and we do not present its citation preferences as if they were ChatGPT's. If you want a number that means something about a specific platform, that is a different and more expensive measurement.
Three rules money cannot bend
A paid placement is labelled, always. It says sponsored on the listing itself, in the visible text, and the link carries rel=sponsored. A placement that is not labelled is not for sale.
A paid placement never enters a citation-ordered list. If a list is sorted by citation count, sponsorship cannot buy a place in it, because that would make the sort meaningless. Sponsored slots sit in their own labelled block, or nowhere.
Payment never changes a measured number. What you can buy from us is labour — someone going into your site and making the fixes. The number that comes out afterwards is whatever the next measurement says, including if it says zero again.
What we will not do, at any price
We will not sell you a single visibility score. Three tools measuring the same brand against the same forty prompts have been observed to report 22, 19 and 24 hits. A number that moves depending on who is measuring is not a number you can act on, so we do not make one.
We will not promise that your citation count will rise. It depends on your content, the platform algorithms and your competitors, and we only measure — we do not control. What we promise is that we test on schedule, that we report what we find including a zero, and that we do the work we were paid to do.
There is one price list, and it is the same for everyone. We sell to one market, so we do not run regional pricing. We do not read your IP to guess what you can afford, we do not show two different numbers to two people on the same plan, and there is no cheaper version of this site to find with a VPN.
We will not sell a rank. If we ever publish a list ordered by citation count, money cannot move a position in it. That is the whole reason a list ordered by citation count is worth reading.
When a listing is re-measured
Once a month, same recipe. Same query, same six runs, same model. If the pool has moved since last time, the listing says so, because a rebuilt pool is a different contest.
The history is kept, not overwritten. A listing shows its previous measurements as well as its current one. A single number with no history cannot tell you whether anything you changed did anything.
A zero is a complete report. If a month comes back with no citations, the listing shows a zero and that is a finished delivery. It is also the finding most likely to be worth acting on, which is why we do not hide it.
This page is the standard. If a listing of ours ever breaks one of these rules, the listing is wrong and the rule is right — tell us and we will correct the listing in public.
Questions about the standard
Can I pay to be ranked higher in the directory?
No. A paid placement is labelled as sponsored, carries rel=sponsored, and is excluded from any list that is ordered by citation count. There is no price at which payment changes a rank.
Can I pay to have my citation status changed?
No. The status is the output of a measurement you can reproduce yourself. Paying us buys labour — someone making the fixes on your site — never the result.
What happens if my listing is measured at zero?
We publish the zero. A measured zero is a complete delivery, not a failure, and it is the finding most likely to be worth acting on.
See the standard applied to a real page
A listing that came back 6 of 6
- Every field with its real value
- Two amber checks it failed anyway
- The JSON-LD it ships to machines
Run the same checks on your own page
- Eight checks, every result shown
- Every fix as copy-ready code
- No account, nothing uploaded