Article No. 80
How to Read Search Volume Data (And When It’s Lying to You)
Abstract
You already have a keyword with a number next to it. Maybe it's 2,400, maybe it's 90, maybe it's a range. The question this guide answers isn't where to find...
On this page
You already have a keyword with a number next to it. Maybe it’s 2,400, maybe it’s 90, maybe it’s a range. The question this guide answers isn’t where to find keywords, it’s how to judge whether the number in front of you means what it looks like it means. That’s a different skill from discovery, and it’s one most people never learn explicitly, they just start trusting whatever number the tool shows.
What “Search Volume” Actually Is
No keyword tool, including Google’s own Keyword Planner, gives you a census of exact searches. Keyword Planner’s numbers come from Google’s ad-auction system, real but incomplete for accounts without active, ongoing spend. Ahrefs and Semrush numbers are statistical estimates built from clickstream data, panels of real, anonymized user behavior purchased from third-party data providers, blended with other signals to model volume for keywords the panel data alone can’t fully cover.
Every one of these is a model, not a direct read of Google’s internal logs. Treat every reported number as directionally useful, not precise to the unit.
Why the Same Keyword Shows Different Volume in Different Tools
Search for the same keyword in Keyword Planner, Ahrefs, and Semrush, and you’ll frequently get three different numbers. This isn’t a bug in one of them, it’s a consequence of different data sources and different modeling choices:
- Keyword Planner draws on Google’s own ad-auction data, which is real click and impression data tied to paid search, not organic search behavior directly.
- Ahrefs combines Google Keyword Planner data, Google Trends data, and other third-party sources, using clickstream signals to split apart keyword variants that Keyword Planner groups together (its own research found this approach roughly accurate for about 60% of studied keywords when checked against Search Console impressions, versus about 45% for Keyword Planner’s own reported figures for the same comparison). Ahrefs is explicit that its numbers, like every other tool’s, are an estimate: “only Google has access to the exact data.”
- Semrush builds its estimates from a panel of clickstream providers processed through machine-learning models, refreshed monthly, with historical data going back to January 2012.
Different panels, different sampling windows, and different modeling assumptions produce different numbers for the same query. Neither tool is “wrong” in an absolute sense; both are estimating the same real thing from different partial views of it.
In practice, this means a keyword showing 1,200 in one tool and 800 in another isn’t a contradiction worth agonizing over. Both numbers are estimates of the same underlying reality, produced by different models with different blind spots. What matters more than picking the “correct” tool is consistency: use the same tool across your keyword set so the numbers are at least comparable to each other, and treat the absolute figure as an order-of-magnitude signal rather than an exact count. If a client compares your number against a different vendor’s report and asks which one is “right,” the honest answer is neither is wrong: point to the mechanism above (different panels, different models) rather than defending your figure as more correct.
Refresh cadence adds another layer of variance. Semrush updates its volume figures monthly, which means a number pulled today can shift by the time you check again next quarter, not because your keyword changed but because the underlying model recalculated against a newer data window. Don’t treat a volume figure as a fixed fact; treat it as a snapshot that will drift over time, particularly for competitive or trending terms.
Read the Trend, Not Just the Average
Most tools default to showing a 12-month average, which flattens out exactly the information that matters most for planning:
- Seasonal terms (“tax preparation,” “holiday shipping deadlines”) show huge monthly swings that a flat annual average hides entirely. A term averaging 1,000/month might be 50 in July and 8,000 in April.
- Declining categories can carry a healthy-looking average built almost entirely from search behavior 8 to 10 months ago, with the current trajectory pointing sharply down.
- Emerging terms often show a thin or unreliable average simply because they’re new, not because demand is genuinely low.
Pull up the monthly trend line, not just the headline number, before deciding a term is or isn’t worth targeting.
Regional and Local Volume Distortion
National-level tool defaults systematically underreport local demand. A term with real, meaningful search volume in a specific metro area can show as low or “no data” volume in a tool’s national database, simply because the national aggregate doesn’t reflect what’s happening in that one market. If your business serves a specific city or region, always check the tool’s local/regional volume setting rather than trusting the national default, and cross-reference against your own Search Console data filtered to that geography, which reflects your actual local audience regardless of what a national panel shows. A term showing “not enough data” nationally can show a real, actionable 90 searches a month once the same tool is filtered down to a single metro area, the exact gap a national default hides.
Why Real Keywords Show “0” or “Not Enough Data”
This is one of the most misread signals in keyword research. A term showing zero or “not enough data” in a keyword tool is not proof that nobody searches it. It’s a known limitation of how these tools aggregate and threshold data: long-tail queries are grouped, sampled, and filtered before they’re surfaced, and terms below a certain sampling threshold get suppressed from the report entirely rather than shown as a small but real number.
There’s no honest, universally applicable percentage for how much real search volume this represents across all tools and all niches, and any source citing a precise figure for “how much traffic zero-volume keywords actually drive” should be treated skeptically. What’s well established is the mechanism: aggregation and sampling limits mean some real, traffic-driving queries simply don’t clear the bar to be reported. This is the exact gap the discovery methods in a companion guide are built to work around, since “no data” in a tool is a data-collection limit, not a demand signal.
Building an Opportunity Score
Once you have volume data you trust (or have appropriately discounted), combine it with three other inputs rather than ranking a list by volume alone:
Opportunity = Relevance x Volume x Difficulty x Business Value
| Input | What it captures |
|---|---|
| Relevance | Does this term actually match what you offer? |
| Volume | Real demand, read as trend and range, not a single trusted point number |
| Difficulty | How realistic is ranking, given current competition and your site's authority |
| Business value | What happens if you rank: a lead, a sale, an ad impression, or just traffic with no downstream value |
A term with strong volume and low difficulty but poor business value (informational curiosity searches with no path to a conversion) can score lower overall than a smaller-volume term that converts. The logic matters more than the exact weighting; the point is refusing to let volume alone drive the decision.
Common Misreads to Avoid
- Chasing head-term volume with no realistic ranking path. A 40,000/month keyword is worthless to you if your site has no realistic chance of ranking in the top results for it within a useful timeframe.
- Ignoring zero-volume-but-real-demand terms entirely. Filtering your keyword list purely by reported volume throws out legitimate opportunity, particularly for niche or emerging terms.
- Treating which of your own pages should rank for a given volume number as a data problem. It’s usually a mapping problem instead, and if a query is already splitting attention across more than one of your own pages, that’s a cannibalization issue with its own diagnosis-and-fix process, not something a volume number will resolve.
Reading volume data well is mostly a discipline of not trusting any single number in isolation: check the source, check the trend, check the region, and treat the reported figure as a starting estimate to sanity-check rather than a fact to build a content calendar on.