Search volume numbers and how wrong they are

I’ve picked keywords off a “2,400/mo” search volume number, built a page around it, and watched it pull maybe 40 clicks a month once it ranked. I’ve also done the reverse: gone after something a tool said was “too low to show” and had it become one of the better traffic pages on the site. After enough of these, you stop treating the volume column as a fact and start treating it as a rough sketch drawn by someone who didn’t see the whole picture either.

This isn’t a “keyword tools are useless” article. I still use them every day. But if you’re building a content plan, a link building budget, or a client report around that one number, you should know exactly how shaky the ground under it is.

Where these numbers actually come from

Almost every keyword tool on the market, including the free ones and the paid ones, ultimately traces back to Google Ads data in some form. Google Ads gives advertisers a monthly search volume range for keywords they’re bidding on, through the Keyword Planner interface and its API. Third party tools like Ahrefs, Semrush, and the rest pull from that same well, either directly through the API or by blending it with their own clickstream and panel data, then republish it as a single number instead of a range.

That single number feels precise. It isn’t. It’s a smoothed-out estimate of ad-auction query volume, repackaged for organic keyword research, which is not what it was built for in the first place. Google Ads exists to help advertisers set bids and budgets, not to hand SEOs an accurate count of organic search demand. The fact that we’ve all adopted it as our source of truth is more about convenience than accuracy.

Google rounds it into buckets before you ever see it

If you’ve pulled a keyword report inside Google Ads itself rather than through a third-party tool, you’ve seen this: instead of a number, you get a range like “1K - 10K” or “10K - 100K.” Google does this on purpose, and has said publicly that it rounds and buckets these figures to protect the anonymity of individual advertisers and to smooth out volatility in the underlying auction data.

The tools you actually use for keyword research take that bucketed range and turn it into a fake-precise single number, usually by picking somewhere in the middle or applying their own modeling on top. So a keyword sitting at “1,900” in your tool might genuinely be anywhere from a few hundred to several thousand actual monthly searches. You’re looking at someone’s best guess about a range, dressed up to look like a count.

Third party tools are guessing on top of a guess

Once a keyword tool has that rough Google Ads figure, most of them don’t stop there. They layer on their own clickstream data (aggregated browsing data bought from panels and browser extensions), their own historical trend modeling, and their own keyword clustering logic to fill in gaps and smooth out spikes. Different vendors buy clickstream data from different panels, weight it differently, and update on different refresh cycles.

That’s the real reason you can pull the same keyword in two different tools and get numbers that don’t remotely agree. One tool might show 500, another might show 3,000, for the exact same term in the exact same country. Neither one is lying to you. They’re running different models on overlapping but not identical raw data, and both are several steps removed from an actual server log of what people typed into Google. I’ve stopped expecting these numbers to converge, and I’d tell anyone building a spreadsheet full of “total addressable search volume” that the math is softer than it looks.

Low volume keywords are close to noise

The lower the volume, the less reliable the number gets. A keyword sitting around 20-50 searches a month according to a tool is often not meaningfully different from one showing “0” or not appearing at all, because at that scale, sampling error and rounding swamp the real signal. Google Ads’ own data gets less reliable at low query counts because there’s less advertiser activity generating the underlying auction data the estimate is built from in the first place.

This matters a lot for long-tail content strategy. If you’re building out a site’s blog around dozens of long-tail, low-competition terms (which is a perfectly reasonable strategy), don’t expect the volume estimates on those individual terms to hold up under scrutiny. Some will wildly underperform. Some will pull in traffic the tool never showed at all, because the tool simply didn’t have enough data to render a confident number and defaulted to showing nothing or near-zero.

Seasonality and geography get flattened

Most tools show you a single trailing 12-month average by default. That one number can hide a term that’s dead nine months a year and spikes hard for three, or a term that’s rising month over month and about to be worth a lot more than the average suggests. You have to actually pull up the trend graph, not just the headline number, to catch this.

Geography causes similar distortion. A “United States” volume figure blends a term that might be huge in Texas and nearly invisible in Vermont into one flat average. If you’re running local sites or targeting a specific region, the national number can send you completely the wrong direction on what’s actually worth targeting.

What the number is still good for

None of this means the volume column is worthless. It’s genuinely useful for a few things:

Relative comparison within the same tool, same country setting, same month. If Term A shows 10x the volume of Term B in the same report, that ratio is a lot more trustworthy than either number in isolation, because at least the same modeling assumptions were applied to both.

Spotting a term that’s clearly dead versus one that clearly has demand. The gap between “no volume, no trend, nothing” and “consistent volume across 12 months” is a real, useful signal even if the exact number attached to the second one is fuzzy.

Catching rising trends early, if you look at the month-by-month graph instead of the averaged number. A term climbing steadily for six months is worth more than its current average suggests, and a term declining is worth less.

How I actually use these numbers now

I treat every volume figure as an order-of-magnitude signal, not a forecast. When I’m deciding whether a topic is worth a page on one of my sites, I look at three things together: the volume trend over the trailing year, what the actual search results look like for that query (are they matching informational intent, commercial intent, something else), and whether I can find the term showing up with reasonable consistency across more than one tool. If two tools roughly agree on “this has real, steady demand” even if their exact numbers differ by a factor of two, that’s a much stronger signal than either number alone.

What I don’t do anymore is build a traffic projection spreadsheet that multiplies “expected ranking position” by “expected click-through rate” by “monthly search volume” to produce a revenue forecast. Every input in that chain is an estimate, and multiplying three estimates together doesn’t average out the error, it compounds it. I’ve handed clients reports like that in the past and watched actual traffic land nowhere near the projection, not because the strategy was wrong, but because the whole exercise was built on a number that was never meant to be that precise.

The bottom line

Search volume numbers are a real, usable signal that’s several layers removed from an actual count of what people type into Google. They come from an advertising tool that rounds its data into ranges on purpose, get reprocessed by third-party vendors using their own clickstream data and modeling, and get least reliable exactly where a lot of long-tail content strategy lives, at the low end of the volume scale. Use them to compare terms against each other and to spot real trends over time. Don’t use them to promise a specific number of visitors by a specific date, to yourself or to anyone you’re reporting to.

If you want help building a keyword strategy that treats these numbers with the right amount of skepticism, The SEO Desk is a good place to start.