Where do keyword volume numbers actually come from?

Open any SEO tool and you will see that a keyword gets 2,400 searches a month and has a difficulty of 38 out of 100.

Both numbers feel like measurements. Neither is one, in any tool, and the reason is worth understanding before you plan a year of content on them.

Nobody has the real number

Search engines do not publish per-keyword search volume to third parties. Every tool therefore models it — from clickstream panels, from sampled advertising data, from its own crawling — and the models disagree with each other, often by multiples, on the same term.

Difficulty is worse, because it is not even modelling an observable quantity. It is a vendor's own composite of link and authority signals, which is why a term rated 30 by one tool is rated 60 by another. There is no external truth to check either against.

None of that makes them worthless. It makes them relative within one tool — useful for ordering two terms against each other in the same product, and misleading the moment you treat them as counts and build a forecast on them.

What ours says about itself

With no data source connected, the keyword numbers here are labelled placeholders, in the interface, not estimates.

That word was chosen deliberately and it replaced a softer one. The earlier wording called them relative proxies, which implies the ordering means something. It does not: these figures are derived from a hash of the keyword string, so the term the screen calls easiest may genuinely be the hardest on the list.

An estimate that is uncorrelated with the thing it estimates is not an estimate. It is a number of the right shape in the right place, and the only honest name for it is placeholder.

Why ship a placeholder at all

So the interface is complete and testable before a data source is wired in — the screens, the exports and the tests all exist and work, and connecting a provider changes the numbers without changing anything around them.

The alternative, which is what most products do, is to ship the same scaffolding with the word "estimate" on it. The customer then plans against figures with no information in them, and nobody ever finds out, because an SEO forecast that does not come true has a hundred available explanations.

The data you already have, which is real

Search Console shows the queries your site actually appears for, with real impressions, real clicks and real positions. It is a much smaller picture of your market and it is entirely true.

For a local business that is usually sufficient: the terms you already rank for on page two are the cheapest wins available, and they are visible for free.

The order of operations that works

Specific pages about things you actually do, each saying something only you can say. Then look at what people were already searching for when they found you, and write more of that.

That sequence needs no volume estimate at all, which is convenient, because nobody has one.

The question to ask any SEO vendor

Where does this number come from, and what happens to it if your panel changes?

A vendor who answers plainly is worth listening to about the rest. A vendor who presents a modelled figure as a measurement has told you how they treat every other number on the screen.