You get one weekly 5-minute check-in. The number that moves week to week is your AI Visibility Score, and we measure it the same way every time so you can trust the direction of the arrow.
Every scan asks ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode the same fixed set of prompts tuned to your bucket and category — and asks each one 2–5 times per engine (more on higher plans). A single AI answer is a coin flip: the same question to the same engine flips run-to-run more often than not. Repeats turn a coin flip into a rate. We count how often you're cited, where you appear in the answer, the sentiment around the mention, and which domain the citation came from. Those signals roll into a 0–100 score. The full algorithm lives in our published methodology — we publish it so anyone can audit it.
Because the score is a rate over repeat runs, it comes with a likely range — the band the number would land in if we ran the whole scan again. Week-to-week movement inside that band is normal wiggle, not a result, and we label it that way. We won't ping you over noise.
Every customer-approved action is measured against fixed windows:
Measured lift = the T+30 score − the T-0 score, and we only record it when the change is bigger than the combined likely-range noise of the two measurements. Within noise → we report nothing rather than a fake number; the field stays empty and we say "within normal wiggle." No clever math beyond that gate. No baseline assumption. No counterfactual modelling. When a lift is real, we show it next to a one-line explanation of which engine moved and why.
We do not blend predicted lift with measured lift (they live in separate fields, never the same number). We do not backfill measurements for actions taken before May 23, 2026 — the pre_push_score column wasn't there yet, so we say "not measurable" rather than invent a value. And we don't smooth or stabilize the score. If your score drops 8 points in a week, you see the 8-point drop — alongside whether that drop is outside your normal wiggle band or indistinguishable from run-to-run noise.
Email measurements@iseen.ai with an optimization ID. Within 48 hours we send back the pre-score row, the T+30 paired scan IDs (plus the T+7 early-signal readout), the per-engine breakdown behind the delta with its likely ranges, and redacted raw query/response samples. If we can't show you the math, we don't bill you for the lift report.
If we measure honestly for you, we measure honestly for ourselves. This card stays live and dated forever, no matter what the numbers say.
The honest placeholder. Real MRR, customer count, and churn appear here the moment they exist. No projections, no "soon," no hockey-stick teaser graphs.
Feature, pricing, and methodology across the four tools we get asked about most. Where we couldn't verify a competitor's data from public sources, the cell reads Verify.
Sources: vendor pricing pages and product docs as of 2026-05-27. If a competitor publishes a methodology page like this one, we'll update the table and link to it.