I run more than a hundred websites. For years, if I wanted to know whether one of them mattered, I opened Search Console and looked at clicks. Low clicks meant the site was quiet, maybe not worth much attention yet. That was the whole test.
It turns out that test was only ever answering half the question. Search Console can only tell you what a human clicked. It has no way to tell you whether anything else is reading the page, and lately, a lot of something else is.
Why this matters right now
Every server already logs every request that hits it, human or not. Most of that log gets ignored because it is noisy ... uptime monitors, SEO crawlers checking on competitors, outright scanners probing for weak points. Buried in the same log, though, are requests from GPTBot, ClaudeBot, PerplexityBot, and a growing list of other AI crawlers actually reading your pages to build the answers they give people later.
None of that shows up in Search Console. It never will ... GSC only knows about Google's own index. If you are only checking clicks, you are blind to an entire second audience that is already forming an opinion of your site.
What most people get wrong
The natural assumption is that no clicks means no visibility, full stop. I made that assumption for a long time. It is wrong often enough to matter, and it is wrong in a specific, checkable way: a page can be nearly invisible to Google search traffic and heavily read by AI systems at the same time. Those are two different instruments measuring two different things. Neither one is the whole picture, and most of us are only looking at one of them.
What I actually found
I finally sat down and joined my own Search Console data against my own server logs across the whole portfolio ... ninety days of GSC, forty-eight days of raw log data, every request already tagged by what kind of bot it was. One property, NaturalHerbLibrary.com, made the point better than I expected. Ninety days of Search Console: 31 clicks, 1,411 impressions. Numbers that quiet, I would normally write the site off as not working yet.
Same window, the server logs: 19,045 requests from nine separate AI crawlers ... OpenAI, Anthropic, Perplexity, Amazon, Meta, and more, all reading that domain on their own, with no human search click involved at all. That is not a fluke reading. Zoomed out across the whole portfolio, the niche that looked weakest on Search Console, health and longevity content, turned out to be the single most AI-read category I own, by a wide margin over every other niche.
One instrument said that corner of my portfolio was quiet. The other said it was the loudest thing I have. Both were telling the truth about what they measure. I just was not asking the second one anything.
A version of this you can run yourself
You do not need a data warehouse to start seeing this. Most hosts keep raw access logs, and every AI crawler identifies itself plainly in its user-agent string. Pull your logs, grep for GPTBot, ClaudeBot, and PerplexityBot to start, and count. Compare that count against what Search Console is telling you about the same pages over the same window.
If the two numbers roughly agree, fine, you already had the full picture. If they do not, and for a lot of sites they will not, you have just found an audience you were not accounting for at all.
What to do with it
Stop reading Search Console as the whole scoreboard. It is one instrument, not the instrument. A page with real AI-crawler activity and weak clicks is not a failure, it is a different kind of early. And a page with real impressions and zero clicks, which I also found in this same pass, is telling you something too ... that one is a listing problem, not a reach problem, because the reach is already proven.
Track both. The full breakdown, the methodology, and the complete portfolio numbers are in the case study on RealSEOLife, and I coined a term for the metric itself, Bot Visibility, over there as well. This is not a one-time check for me either. I plan to keep running it.
