You can’t measure SEO the way you did two years ago. Rank trackers can show you a different Google than your customers see, and Search Console hides the query behind most of your clicks. Two guests on Unscripted SEO, Ernesto Ortiz of Structural SEO and Matt Goren of RunOctopus, gave me the clearest answer I’ve heard to the obvious next question: so what do you measure now? Here’s their answer, plus the routine I’d run on a small business site this week.

Rank trackers are reading a different Google than your customers
A rank tracker is only useful if the results it scrapes match the results a person sees. Ernesto told me he isn’t sure that holds any more. Tools that send automated requests to Google have been getting results that don’t line up with what a human gets for the same search, and he hasn’t decided whether that’s a bug or something Google is doing on purpose.
“The fact that results for automated requests aren’t the same as the ones you see as a human makes the decision-making process for SEOs harder and you have to go back to manual because how are you gonna trust results that are thirty percent off?”
— Ernesto Ortiz, Structural SEO, on Unscripted SEO Special Edition
That “thirty percent” is Ernesto’s read from his own checks, not a published measurement, and you should treat it that way. But you don’t need the exact number to change how you work. If your tracker says you moved from position 6 to position 3, check that query by hand in a clean browser before you put it in a client report. If the two don’t match, the tracker is a trend line at best, not a fact.
Here’s the uncomfortable part. Most of us built our reporting on rank data because it was cheap and it looked precise. Precise and accurate were never the same thing.
Search Console hides the query behind most of your clicks
The other half of the problem is Google Search Console. You get the clicks. You don’t always get the query that produced them. Google anonymizes queries that too few people search, and on a typical site that’s most of the traffic.
I measured this across 38 Search Console properties, 12 months (October 5, 2025 to October 4, 2026), 4.06 million clicks. On the average site, 63.6% of clicks arrived with no query attached. The median site was worse, at 71.8%. Product pages were the darkest content type in the study, with a median of 83% of clicks dark. The full method and the breakdown by site type are in my dark clicks study of Google Search Console on SEO Arcade.
So when you open the Performance report and sort by query, you are usually looking at the minority of your traffic. Your top 20 queries can tell you a story that’s true for those queries and wrong for the site. If you want to see how badly your own visits skew the visible part, I wrote up how operator queries inflate your Search Console baseline after catching myself in the data.
The long tail you can’t see is the part AI is growing
Ernesto’s point was that hidden queries aren’t new. Google has always held some back. What changed is the share of searches that fall below the threshold, because people now ask much longer, much more specific questions.
“I think what the challenging part now is the long tail is now becoming more and more big because people are able to do just more niche questions to their specific situation because AI is able to resolve them.”
His example was shampoo. Ask Google for the best shampoo for a 30-year-old with a specific problem in a specific climate and you used to get junk. Ask ChatGPT and you get an answer. Once people learn they can ask that way, they keep asking that way, in Google too. Every one of those long questions is close to unique, which means it’s exactly the kind of query Search Console won’t name.
The trap is reading that missing data as missing demand. Matt was blunt about it.
“I’m willing to let people get stressed about the anonymized searches and ditch them because I want all of those.”
He walked through the math with me on a hypothetical: a few hundred questions, each searched a few dozen times a month, adds up to thousands of chances to be found that never appear in your query report. When other site owners give up on terms that show zero, those are the terms left on the table for you. (I covered the rest of Matt’s argument, including why ChatGPT sends people to the local shop instead of the chain, in what Matt Goren taught me about AI search for small business.)
What to measure when the query data goes dark
You can’t get the hidden queries back. You can stop depending on them. Here’s what I’d put in the report instead, ordered from most trustworthy to least.
- Clicks and conversions by page, not by query. Search Console still counts every click to a page, named query or not. A page whose clicks grow while its visible queries stay flat is gaining in the long tail you can’t see. That’s a real signal.
- Leads you can trace. Form fills, calls and booked appointments from organic landing pages. If you have to explain results to an owner or a CEO, this is the number that survives the meeting. Rich M. Smith and I got into that in how to report SEO results to a CEO.
- Manual spot checks of the queries that matter most. Ernesto’s fix for rank drift was to go back to manual. Pick 10 queries that drive revenue and check them by hand once a month, logged out, in the location your customers search from.
- AI answers, checked by hand. Matt’s routine is to run the questions your customers ask through the chat tools and record whether you show up.
- Rank tracker trends, last. Keep the tracker for direction over months. Don’t quote a single position from it as fact.
“I think it’s important to be running these types of queries through the chats and seeing where you’re at.”
Item four is the one most small businesses skip, and it’s the cheapest. You don’t need a tool to start. You need a list of questions and a spreadsheet. I’d keep the same questions every month so you’re comparing like with like, and note which competitors get named when you don’t. If you’re answering to an executive and need help framing that kind of evidence, The SEO Advisory has a good piece on proving SEO value after attribution broke.
One honest limit: none of this gives you the precision rank tracking used to promise. A monthly spot check on 10 queries is a sample, not a census. But it’s a sample of the real results, which beats a census of the wrong ones.
How to measure SEO without query data: the monthly routine
If you want a routine you can actually keep, here’s the one I’d hand a small business owner. It takes about an hour a month.
- Write down 20 customer questions. Use the ones your front desk or installer answers every week, in the customer’s words. Ernesto would add the product questions: who it’s for, who it’s not for, what it includes and what it doesn’t. I wrote about that in what Ernesto Ortiz taught me about saying who a product is not for.
- Check each one by hand in Google and in at least one AI chat tool. Record whether you show up and who does.
- Pull clicks by page from Search Console and compare against last month. Ignore the query tab for this step.
- Match pages to leads. Which landing pages produced a call, a form or a sale?
- Fix the page with the gap. Where a question has no good answer on your site, write one. Where a page gets clicks but no leads, fix the page.
You’ll notice the routine doesn’t start with a keyword tool. That’s on purpose. When the tools and the reports both get blurrier, the questions your customers actually ask are the one data source that hasn’t degraded.
So here’s the question to take into next month’s report: if you deleted every rank position and every named query from it, what would be left, and would it still tell you whether SEO is working?
Quotes are from the raw Riverside transcripts of each episode, trimmed only for false starts and repeated words. Full conversations: Ernesto Ortiz’s special edition recap and Matt Goren’s episode recap on Unscripted SEO.
