What a lone reading costs. Without a second look taken under identical conditions, a swing in the answers is indistinguishable from a result, and you risk paying for work nobody can show moved anything. CiteAngle fixes the conditions in advance and seals each reading with them — 50 questions, 21 engines, 7 runs — which is the reason two of our readings can be subtracted from each other at all.
A second cost arrives with the first surprise. Without a series, a bad month cannot be told apart from ordinary answer variance, so teams react to noise: work gets stopped that was working, and the loss never appears as a line anywhere.
Here is the problem a single reading leaves behind. Generative answers move on their own: ask the same question a week apart and the names in the reply can change without anything on your site changing at all. So one number cannot tell you whether the work you paid for did anything — it only tells you what that day looked like. The moment somebody in the room asks how this compares with last quarter, a lone coordinate has no answer, and a second reading taken under different conditions is worse than no reading, because it looks like evidence.
That is the whole reason this product exists, and it is why the fixed list below matters more than the cadence. Questions, engines, run count, judging rule and reporting window are locked before the first reading and stay locked, and each reading is sealed with the conditions it ran under. Competitors are read inside the same answers, on the same questions, in the same runs, so the line you steer by is a position on a field. And readings the grid planned but could not take are printed as a count instead of dropped, since a denominator that quietly shrinks would manufacture an improvement out of nothing.
How is this different from an audit?
An audit answers where you are. Monitoring answers whether that is moving, and which way. That second answer is worth something only if the second reading was taken exactly like the first — same questions, same places, same count. Keeping the setup steady between runs is the hard part, and it is the part we own.
What exactly stays fixed?
Five things get sealed at the first run and held for every run after it. Sealed means we take a fingerprint of the setup and the raw data the moment a run ends, so any later edit shows. Change one of the five and the next reading opens a new baseline — a new starting line — instead of counting as movement against the old one.
- The question setThe panel you sign off on is locked to a version. Add a question later and it opens a new series rather than stretching the old one.
- The places we readThe same 20 places an answer can show up, named just as they were named at the first run. If a vendor renames one, we log the rename rather than quietly folding it in.
- How many times we askSeven runs per question on every place, every time. Asking once and asking seven times give you different numbers, so the count sits in the contract.
- Judging rulesA citation is the AI listing your link as a source. A mention is your name in the answer text with no link. Which is which — and what we do with an answer too vague to call — goes in writing before the run, not after we see it.
- When we sealEach run closes with a receipt: which service, which model, the time of capture, the settings for each run. A later reading gets set against a record no one can quietly edit.
The full method behind these five — why seven runs, how we work out the ranges — is published on the methodology page.
What lands every month
Two rhythms, one grid. The monthly read is narrow, so it catches movement early. Each read is wide enough to settle the call for that month.
Where the observation cycle stands right now
18 days 22:11:47 to the next cycle
Every time the month turns, the whole grid runs again. This clock is read off the calendar, so you can watch the rhythm before you start a series.
A month, as it is planned
Month 1
Baseline sealed
You sign off on the panel and we lock it to a version. The first run closes with its receipt, and that record becomes the t0 — the starting point every later reading is set against.
Month 2
Spot read
The same panel, run again on the same rules. Movement month over month, sources that showed up and sources that fell away, plus a written note on what to watch next.
Month end
Delta report and lens reading
A full-width run that shows the status of every planned observation. It joins the two spot reads on one panel, then closes with a decision deck and the whole change-card ledger.
Do you want the reading, or the reading plus the moves?
Some teams want the reading and will do the work themselves. Others want the reading and the moves in one loop. We sell those as two lanes so the scope stays easy to read.
Reading lane
Competitive Signal — the reading, monthly
A fresh read on your rivals every month, settled by a delta report each month, with change cards for what is new, what dropped and what got worse. Your team does the work. This is the wide-angle Panorama on a monthly loop — the whole grid re-measured every month, with thirty competitors judged in the same answers and the month-on-month movement that a single reading cannot produce.
Reading plus moves
Growth Operations and Category Leadership
The reading, a written decision memo, and the moves themselves with a receipt when each one lands. One monthly loop that settles into a change figure every month. These carry the audit credit as well.
Week-by-week output for the second lane sits on the programs page. Monthly and yearly prices are on the pricing page.
How often should you read it again?
Match your own decision rhythm, not the calendar. AI answers move day to day. Our repeated runs see it, and published outside studies see the same thing, so one audit gives you a single point. A verdict on change needs waves of reads on the same method. If you set budgets every month, a wave on the same grid is what makes each month's number hold up.
A baseline can only be built forward. Nobody can put your questions to last month’s engines. The sooner the first run, the longer the series you own.
What do we need to start a series?
A sealed baseline on the same domain, taken in the thirty days before the program starts, and the question panel you signed off on. If a recent paid audit already sits on that domain, it becomes the t0 and the series runs from there.
Why one reading goes stale
The measured case for re-reading, in full.