RPM drivers
Not every page view can be sold
Your traffic is three populations, and only one of them pays. Counting the other two flatters your dashboard and drags down the RPM you use to judge everything else.
By Mario Bailey · Published
RPM is revenue divided by page views, so it is only as honest as the page-view number under it. On a site we run, a large share of the search visibility was traffic that could never be sold an ad: misdirected foreign navigation and outright machines. Averaging them in makes the site look busier and earn less per view than it really does on the audience that matters.
The three populations
When we separated the search traffic on a business site we run, it fell into three groups that behave nothing alike.
| Population | Behavior |
|---|---|
| Human commercial | Clicks, reads, and can be shown a valuable ad. This is the money. |
| Misdirected foreign navigation | Looking for a portal or login, not your content. One such query drew 6,807 search impressions at a 0.0% click rate. |
| Machine queries | Reference lookups by AI agents and scrapers: contract numbers, legal citations, boolean strings. Bursty, and they click at zero. |
Only the first population sees and clicks an ad worth showing. The other two inflate impression and, where they do land, page-view counts without adding a cent of monetizable attention.
Why this quietly lowers your RPM
Every page view from a visitor who will never be shown a valuable ad still sits in the denominator of your RPM. So a site with a lot of misdirected or machine traffic reports a lower RPM than it actually earns on its real audience, and a publisher who reads that headline concludes the ads are underperforming when the truth is the traffic mix is diluted. The advertiser demand that pays is bidding on the human commercial slice. Measured against just that slice, the same site's monetization looks far healthier.
The chasing trap
The dangerous move is to treat topline traffic growth as the goal and chase whatever lifts the number. Foreign navigational and machine traffic is often the easiest volume to acquire and the least valuable to own; scaling it up grows the page-view count, dilutes the RPM further, and in the worst case trips invalid-traffic flags. Growth that does not add human commercial visitors is not growth you can sell. Judge a site by the trajectory of the population that pays, not by the sum of all three.
How to separate them
Look at click-through rate by query and by country, not at raw impressions. A query or a market with heavy impressions and a click rate near zero is navigational or machine traffic wearing the costume of an audience. One caution: Search Console only samples the query dimension, so it will not show you every query behind your clicks, and you should never project a revenue number from the visible sample alone. Use it to see the shape of the populations, not to size them exactly.
How we measured it
Search Console impressions, clicks, and click-through rate by query and country on a site we run, segmented into the three populations above. The measurement is of search traffic composition; the monetization implication follows from it, since only human commercial visitors are shown and click ads worth bidding on. The click rates are real and our own, and the sampling caveat above is why we read shape, not size.