Search volumes are an essential piece of information for planning any work with search engines. We all use search volume to prioritize keywords, decide which topics to cover, and maybe even decide which products to build.
Search volumes are not accurate, and they’ve not been accurate for a decade. The volumes you see everywhere – including from Google – are estimates.
Academically this makes sense, but the seriousness with which we treat “50 people a month searched for this keyword in April, but then it dropped to 30 in May” does not reflect the reality that these volumes are estimates.
It’s important to understand this. It’s also important to urgently add significantly more nuance to your understanding of search volumes in order to thrive in a generative AI world.
Search volumes are not accurate
Google used to provide accurate search volumes through the AdWords API. These were removed in 2013 in the name of privacy. An accurate replacement has not been provided.
Most of the search volume data you see today is taken from Google’s Ads Keyword Planner. That this data is an estimate becomes more obvious if you’re using the Planner itself, where you see huge ranges:
But most SEO software gets its search volume data from the Planner’s API, which returns much more specific numbers. These numbers get presented to you as “truth”. My search volume range of 1k to 10k per month changes to a very specific 2,600 per month through the API, with an alert that interest is up 45%!
From this data you might make the decision to create – or not – a post about the best coffee grinders. But we’re collectively ignoring that the data is not accurate.
Recognizing the limitations of Google’s data, some SaaS products will make their own estimates. Ahrefs, Semrush, and SpyFu are the major providers offering this. The way they do this is quite interesting: they take Google’s data and then enrich it with clickstream data.
Clickstream data comes from the murky world of ads, trackers, and wholesale data purchasing. You can buy wholesale data sets that track users’ activity across the internet. If you filter this appropriately, you can use it as a proxy for search volumes.
This is an interesting solution and has merit, but it’s impossible to properly assess the accuracy of any of the methods as there is no overall source of truth. The clickstream solution adds more nuance, but it’s hit-and-miss on accuracy for low search volume topics in particular.
Here’s how the search volume differs for the keyword “best coffee grinder” across the different data sources:
- Keyword Planner: 1k to 10k.
- Keyword Planner API: 2,600.
- Ahrefs: 2,000.
- Semrush: 2,400.
- Spyfu: 1,600.
The Planner offers a 10x range, and the SaaS solutions have a 38.5% difference. You might reasonably pick this keyword over another with “2,000” search volume, when in fact because of the margin of error the “lower” volume keyword could, in reality, have more searches. It’s vital to understand this in order to interpret and triage the data more effectively.
Not all searches are equal, but search volume presents them as such
An additional challenge is that not all searches are equal. In a world of generative AI, zero-click searches, and featured snippets, you cannot treat all searches as equal by just looking at volume. You desperately need more nuance than is offered by mainstream tools.
Google is opening up more and more actions for users to take when making a search. We’ve long had ads alongside organic results, and even featured snippets are over 10 years old.
But, a search in 2024 has more possible outcomes than ever, and once you throw in improved generative AI in search – which is coming – whether or not your search can be answered in 2 short sentences becomes critical.
To keep using my coffee grinder example, these searches are all counted the same:
- A search that results in an ad click.
- A search on desktop that results in multiple sessions and a purchase.
- A search on mobile that results in research that is revisited later.
- A search that results in no click as a result of using Google’s featured snippet or AI generated answer.
1 of those – the purchase – is a truly valuable search if you’re trying to sell coffee grinders. Search volume metrics count the rest as though they’re of equal value to you.
You therefore need to start thinking about the complexity of the questions a searcher has against your target keywords. If there is a simple answer then the high search volume may not result in clicks today, and it’s going to result in fewer and fewer clicks over time. On the other hand, if searchers need your expertise, experience, and authority, and can trust your detailed and high-quality answers to their questions, this type of content can thrive.
Simple additional metrics to consider
One of my complaints about SEO software – that we solve internally at Ellipsis through surfacing metrics important to us through FALCON AI – is that it gives you the feeling of being data-driven, whilst not truly giving you the information or insight you need to make a data-driven decision.
We propose a couple of points you should always consider:
- Can this search be answered for the vast majority of searchers in 2 sentences?
- How many additional elements are on the SERP? Is this therefore likely a no-click search?
You can answer the former from understanding the subject matter and making a judgment call. You can answer the latter from viewing and analyzing a true neutral version of the SERP.
And we propose additional data you should get hold of:
- A low and high estimate of search volume, using data from different sources.
- % of searches likely to be zero-click.
- How many clicks each month a search gets.
- The share of searches on desktop vs. mobile.
If, like me, you love spending money on all the data you can access in order to get the best results possible, the additional data will be a treat. The clickstream sources I mentioned above, in addition to data from Google, is your source for these.
It then obviously takes more effort to make a decision about which topics to target, but the results should make the whole difference.
Search volume needs more nuance and rigor
Search volumes are not accurate, but understanding this and adding significantly more rigor to the keyword selection process can result in higher quality outcomes.
Search volume must be understood as an estimate and a range. It must be clear that all clicks are not equal. You need your content to thrive in a generative AI world.
Understand all of these, and you can optimize for the future unfolding in front of us, whilst your competition is still searching for answers.
