Open three keyword tools, type the same term, and you'll get three different search volumes. It's not that two of them are broken — it's that nobody actually measures search volume directly. Google doesn't publish it. Every number you've ever seen is a model.
The only ground truth is a bucket
The closest thing to an official number is Google Keyword Planner, built for advertisers. It returns ranges ("1K–10K"), merges close variants, and rounds aggressively. Useful as an anchor — useless as a precise figure.
So how do the big tools get to a specific number?
They combine signals and model the gaps:
- Keyword Planner for an absolute-ish baseline.
- Clickstream data — anonymized browsing panels bought from third parties — to de-bucket the ranges and estimate real clicks.
- Google Trends for relative popularity and seasonality.
- Modeling to extrapolate to the long tail no panel has data for.
Different panels + different modeling = different answers. None is "the truth"; they're competing estimates.
Treat search volume as a confidence-graded estimate, not a fact. The honest question isn't "what's the number?" but "how grounded is this number?"
A better way to read volume
This is why Citenix attaches a confidence grade to every estimate:
- Measured — anchored by real first-party demand (your own Search Console impressions).
- Calibrated — anchored by a provider's bucketed data, de-bucketed with supporting signals.
- Modeled — extrapolated from semantic neighbours when no direct signal exists.
A keyword that says "1,200 / Measured" deserves more trust than "1,200 / Modeled" — and a tool that hides that distinction is selling you false precision.
What to actually do with volume
Use it for relative prioritization, not absolute forecasting. Pair it with intent and difficulty, and weight your own Search Console data above any third-party estimate — it's the only number that's truly yours.