Google Keyword Planner for SEO, and the four things it will not tell you

Keyword Planner is the origin of most search volume data in the SEO industry, and it is not an SEO tool. It is a bidding aid inside Google Ads, and almost every confusing thing about it follows from that one fact.

Here is the sentence to carry around. Google defines its main metric as “the average number of times people have searched for a keyword and its close variants based on the month range as well as the location and Search Network settings you selected”. Four qualifiers, one number, and the column heading says none of it.

1. The number is a cluster, not a keyword

Close variants are folded in. That means two different phrasings of the same idea frequently report the identical figure, because they are not two keywords with similar demand. They are one measured cluster, displayed twice.

In our last sizing run this was everywhere. A phrase and its reordering came back the same. A term and its everyday synonym came back the same. Both are Google collapsing variants into a single reading, exactly as documented.

The consequence is arithmetic and it is expensive:

• Never sum two phrasings of the same thing. A plan that adds a term to its synonym has invented demand, and the invented figure is usually the one that justified the project.
• A cluster is worth its largest single phrasing, not the total of its rows.
• Format variants are a real exception. Adding a platform or a file type to a query, the “in Excel” or “for Google Sheets” ending, does return its own separate figure. Those are genuinely different searches, not variants of one.

If you have ever built a content plan by exporting a hundred keywords and totalling a cluster column, some part of that total was the same search counted several times.

2. The ranges are a display, not a data limit

Accounts without meaningful spend generally see bucketed ranges instead of numbers. Two candidates that both land in 1K to 10K are indistinguishable, which is precisely the decision you opened the tool to make.

Google does not document this condition in the Keyword Planner help. It is consistently reported by advertisers and easy to reproduce, and we are naming it as observed behaviour rather than as a published rule, because the distinction matters when you repeat it to a client.

What is documented is more useful. The Google Ads API’s historical metrics method returns the same measurement as a single integer, averaged over the past twelve months, alongside a month by month breakdown and the list of close variants that were folded in. Same data, no buckets.

So the practical answer to bucketing is not a workaround inside the interface. It is to get the number from the source that returns numbers, either through the API directly or through one of the many tools that resell it.

3. Zero is a result, and the tool will still suggest it

The keyword ideas panel is a suggestion engine. It will offer phrasings that read perfectly, look like exactly the page you were about to write, and have no measurable demand behind them at all.

We priced a batch of a few thousand candidate keywords in our last research run. Roughly 43% of them came back with no volume. Several sat near the top of our own shortlist beforehand, ranked by a popularity proxy we had built ourselves, and the measurement retired them in one request.

That is the single highest return thing paid volume data does. Not confirming the winners, which you can usually guess. Killing the plausible losers before anybody writes them.

The habit that follows: size the shortlist before you write, and write the whole raw response to a file. Paid data you did not keep is paid data you will buy twice, and a figure with no file behind it cannot be checked when somebody eventually questions it.

4. The twelve month average hides the shape

The default is an average across twelve months. A term that tripled in the last quarter and a term that collapsed can report the same average, and the average is the column everybody reads.

The monthly breakdown is right there and mostly ignored. Use it for anything seasonal, anything driven by a product cycle, and anything in a category that is currently changing fast, which at the moment includes most of this one.

What it is genuinely good at

None of the above makes it a bad tool. It makes it a tool that is honest about being built for advertisers, in documentation most people do not read.

Discovery. The ideas panel is grounded in real query data and it is free. It will surface phrasings you would not have thought of, which is the expensive half of keyword research.

Relative sizing. Even bucketed, it separates a real topic from a hobby topic reliably.

Location and network settings. These are first class controls, and they matter more than the volume figure for anybody selling in one country or one language. Most third party tools expose a thinner version of the same thing.

Seasonality. The monthly breakdown is the best free source of it.

Treat it as the discovery and sanity check layer, then pay once for real numbers on the shortlist that survives. Which tool you pay matters less than most roundups suggest, because nearly all of them are reselling this same data. If you want to run the expansion conversationally instead, the keyword research prompt and the clustering prompt cover finding the phrasings and grouping them into pages.

The two questions it was never built to answer

Does the click survive? Volume counts searches, not visits. A query that is fully answered on the results page still reports its volume in full, and the method for reading around that is judgement about intent rather than anything in a data column.

Does an assistant name you? There is no ad auction behind a prompt, so there is no volume data about prompts for anyone to sell. The measurement for that half is a rate rather than a number: run the questions your buyer would plausibly ask, and count how often you are named. Our AI visibility checker does that, and the free tools cover the floor underneath it, which is whether anything can fetch and read your page at all. For where you sit on the typed half, a rank tracker remains the right instrument.

Keyword Planner tells you what people typed into a search box last year, averaged, clustered and bucketed. That is genuinely useful and it is a smaller claim than the industry usually makes on its behalf.

Sources

Read 16 August 2026.

• Google Ads Help, about Keyword Planner forecasts, for the definition of average monthly searches and its close variant, location and network qualifiers.
• Google Ads Help, use Keyword Planner, for what the tool is built to do.
• Google Ads API, generate historical metrics, for the twelve month average returned as a single value with a monthly breakdown and the close variants behind it.
• The bucketed range behaviour on low spend accounts is widely reported by advertisers and is not documented in Google’s published Keyword Planner help, which is why it is described here as observed rather than stated.