Company grid
Vendor comparison insight
The same products lined up company by company, with initials in place of names. Open the grid →
Why the nine questions matter. Without them you compare vendors on presentation, and the risk is that the cheapest-sounding option is simply the one measuring least. CiteAngle answers all nine on this page, with competitors read side by side with your brand in the same run and every reading sealed — that is the reason we published a grid we are on.
One number frames the rest of this page. Our own panel runs 50 buyer questions across 15 engines, 7 times each — every planned reading landing on 17 named surfaces, each sealed with the conditions it ran under. Ask every vendor on your list for the same four figures. A roster that cannot be counted cannot be compared.
How to read this page
Everything compared here comes from what each company published on its own pages. On 2026-08-01 we downloaded and archived those pages in Korea, the United States and Japan, and citations stay inside that archive. Company names are withheld — what we sell is our own measurement, not somebody else's flaws, and one archived page is thin ground on which to judge a whole product.
"Not found in what we read" means exactly that: absent from the public pages we archived. It is not a claim that the capability is missing. The company-by-company grid, with initials in place of names, sits in the vendor comparison insight.
The nine questions
| Ask this | Why it decides the purchase | What CiteAngle answers |
|---|---|---|
| How many places do you ask? | With a narrow roster, "your brand is absent" and "our roster is narrow" look identical. | 15 engines get asked and 17 surfaces get counted — an engine is a service we send the question to, a surface is a screen where its answer shows up. Each one is named on the coverage map. |
| Do you read search results and AI answers together? | Buyers move between both screens. Measure one and the prescription has nowhere to land. | Both sit in the same grid, on the same yardstick. In the pages we archived we did not find another product describing that combination. |
| Do you cover the surfaces my market actually uses? | A US brand expanding into Korea or Japan meets screens that a US-only roster never opens. | Korean and Japanese answer surfaces are measured natively, in-language, on the same protocol as the US roster. |
| How many times do you ask the same question? | Generative answers vary run to run. A single pass buys you that day's luck. | Every question is repeated, the repetition count is published as a number, and shares are reported with the interval around them. |
| Can I verify the conditions afterwards? | Recorded conditions are what make the next reading comparable and let "we improved" survive an audit. | Conditions are sealed with a hash at measurement time, and every public figure is listed with its source and as-of date in the claims registry. |
| Is the price published? | When the number arrives only after a sales call, the comparison stalls before it starts. | List prices are published in all three markets, and the calculator adds up any combination you pick. |
| Does the engagement stop at the number? | A dashboard reports the same score next month unless somebody changes the pages behind it. | The refit sprint, the category sprint and the monthly programs sit on the same ladder as the audits. Scope is on the programs page. |
| How often do you read it again? | One reading is a coordinate; a series is a trend. They are different purchases. | The audit fixes the coordinate, and monitoring continues it under conditions held constant — the list of what stays fixed is on the monitoring page. |
| Can I take the data with me? | If the evidence lives only inside a vendor dashboard, it leaves when the contract does. | Reports ship as a hosted page, a PDF and JSON, with a signed share link. Published datasets are open at machine-readable addresses. |
Where the archived pages actually diverged
Line the archive up on one axis and the differences cluster in three places.
Scope of what gets counted. Most products read AI answer surfaces only. A description of search results and AI answers held in one grid, on one yardstick, did not appear in what we archived. Rosters that reach into Korean or Japanese surfaces were rare, and where they existed they carried a single surface each.
Treatment of variance. Repeating a question is not unique to us — several pages say they repeat, and a few publish the count. What most of them sell instead is the refresh cadence: daily, every few days, weekly. Refreshing often and asking repeatedly are different instruments. One thickens the time axis; the other makes a single point trustworthy enough to act on.
Verification machinery. A sealed record of measurement conditions, published as a ledger anyone can check, did not turn up in the archive. At the other end of the range, some pages carry growth percentages with no source attached at all. A company selling numbers should open the provenance of those numbers first.
Read next
Our evidence
Published method
How we answer the nine questions above, and the procedure behind each answer. Open the method →
Run the numbers
Cost calculator
Audit, add-ons and program in one total, straight off the published price list. Open the calculator →