Key takeaways
- The funnel is three kinds of query, not three kinds of customer: people learning that a problem exists, people comparing solutions, and people choosing a supplier.
- Build it from the bottom. Bottom-of-funnel pages are fewer, cheaper and convert, and they tell you what the upper stages should be about.
- Most content programmes do the opposite because the top of the funnel has the biggest volume numbers, which is also where the least intent lives.
- Measuring the funnel in sessions is where it misleads. On this site 99.2% of page-one queries produced zero clicks in the measured period, so a session-based model values most top-of-funnel presence at nothing.
- Pew found users clicked a result on 8% of visits with an AI summary against 15% without, which means the top of the funnel is becoming less clickable rather than less valuable.
- The useful measures are impressions by stage and branded search growth, with conversions attributed to the bottom only. That is less tidy and considerably more honest.
The funnel is the most useful diagram in content marketing and the most abused measurement framework. Both because it is simple enough to put on a slide.
As a way of deciding what to write, it works: it stops you producing six pages for the same stage and none for the others. As a way of judging whether the writing worked, it quietly breaks.
What follows is how to use it for the first job, why it fails at the second, and what to measure instead.
What is the SEO funnel?
Short answer: Three categories of search query, grouped by how close the searcher is to buying. People discovering that they have a problem, people comparing ways of solving it, and people choosing between specific suppliers. The same person moves through all three.
The important correction is that these are query types rather than audience segments. One buyer runs all three kinds of search, often within a week.
| Stage | Query shape | Page that answers it |
|---|---|---|
| Problem aware | Why is my X doing Y | Explanatory article |
| Solution aware | How to fix X, X versus Z | Comparison or method article |
| Supplier aware | X agency, X services, best X | Service page or listicle |
Notice that the third row is where your commercial pages live and the first two rows are where almost all content budget goes. That inversion is the standard mistake.
Why should you build it from the bottom?
Short answer: Because the bottom is cheaper, smaller and actually converts. Supplier-aware queries need a handful of pages rather than dozens, the people searching them are already ready to buy, and what they ask you tells you what the upper stages should be about.
The bottom also fails fastest, which is useful. If your service page cannot convert people already searching for a supplier, more awareness content will not help.
- Make the supplier-aware pages good. Your service pages, your pricing, your comparison content. Usually under ten pages.
- Check they convert the demand they already receive, using your own impression and enquiry data rather than an assumption.
- Then build the solution-aware layer, choosing topics from the questions bottom-of-funnel prospects actually asked you.
- Then build the problem-aware layer, and only to the extent you can afford to treat it as a long-term investment rather than a channel.
Step two is the gate most programmes skip. Spending on awareness while the commercial pages leak is paying to fill a bucket with a hole in it.
Why does measuring the funnel in sessions mislead?
Short answer: Because the top of the funnel increasingly produces visibility without visits. An explanatory article can be read as an AI summary or a snippet and send nobody, which a session-based model records as a failure rather than as an answered question.
Our own data is the sharpest version of this we can offer. Across the queries where this site reached page one in the period we measured, 99.2% produced zero clicks.
That is a figure we published rather than one we would have chosen. It means a session-based funnel model would have valued almost our entire page-one presence at nothing.
Pew found the same direction externally: users clicked a result on 8% of visits where an AI summary appeared against 15% where none did. Ahrefs found comparable click loss on top-ranking pages.
The conclusion is not that top-of-funnel content is worthless. It is that sessions are the wrong unit for it, and a model built on them will recommend cutting the stage that is working.
What should you measure at each stage?
Short answer: Impressions at the top, because that is what presence looks like when clicks are falling. Position and clicks in the middle, where comparison still drives a visit. Conversions only at the bottom, because that is the only stage where attribution is defensible.
| Stage | Primary measure | Do not use |
|---|---|---|
| Problem aware | Impressions, branded search growth | Sessions or conversion rate |
| Solution aware | Clicks and average position | Last-click revenue |
| Supplier aware | Qualified enquiries | Impressions alone |
The right-hand column matters as much as the middle one. Applying a bottom-of-funnel measure to a top-of-funnel page is how good content gets cancelled.
Branded search growth is the most useful single proxy for top-of-funnel effect. It is free in Search Console and it rises when people have encountered you somewhere that sent no referrer.
How many pages does each stage need?
Short answer: Few at the bottom, more in the middle, and potentially unlimited at the top. The exact ratio matters far less than the sequence does: an unconverted bottom makes every single page above it less valuable, however many of them you publish.
In practice most businesses need under ten supplier-aware pages, a few dozen solution-aware ones, and as much problem-aware content as they can sustain without diluting.
Dilution is the real constraint at the top. A hundred thin explanatory pages is worse than twenty good ones, because each one is a weaker answer competing for the same attention.
The test for whether to write another is whether it could be the best thing on the internet about its question. If not, it is volume rather than content.
Frequently asked questions
Is the SEO funnel the same as the marketing funnel?
It is the search-specific version of the same idea, and it differs in one way that matters: the stages are query types rather than audience segments, so one person generates searches at all three stages rather than progressing through them as a cohort.
Should I write top-of-funnel content if it does not convert?
Yes, as a long-term investment measured in impressions and branded search rather than conversions. What you should not do is judge it by a bottom-of-funnel measure, which is how it reliably gets cut while working.
How do I know if my bottom of funnel is leaking?
Look at impressions against enquiries for your service pages in Search Console. Substantial impressions with no enquiries means the demand is arriving and the page is not converting, which no amount of upper-funnel content fixes.
Does AI search break the funnel?
It breaks the measurement rather than the model. The stages still describe how people buy; what has changed is that the top two stages increasingly get answered without a click, so session-based measurement understates them.
What is the most common funnel mistake?
Building top-down. The volume numbers are biggest at the top, which makes it the natural place to start and the worst one, because nothing above a leaking commercial page is worth much.
Sources and further reading
- Google Search Central, getting started with Search Console
- Pew Research Center, Google users are less likely to click on links when an AI summary appears (22 July 2025)
- Google Search Central, creating helpful, reliable, people-first content
- Google Search Central, AI features and your website
- Ahrefs, AI Overviews reduce clicks to top-ranking pages
Check the bottom before funding the top
Send us the impression and enquiry numbers for your service pages. If the demand is already arriving and not converting, we will tell you that before we propose writing anything.