The impact of Google's /goto on SERP scraping and rankings: I crossed 693 keywords with Search Console and one in nine changes in the top 10

On 10 August I wrote here that Google was testing a login wall past page 2, and that the risk was not ranking but measurement. I ended by asking you to take inventory: which decisions you make with third-party data that scrapes Google, and what you would do tomorrow if that access disappeared.

The login wall came to nothing. But while I was watching that door, Google was already closing another one, and this one is fully deployed.

The google.com/goto redirect and why it makes rank tracking expensive

As a result of this change, result links no longer point directly at the destination. They point at google.com/goto with the real URL encoded inside, and from there a 302 sends you to the page. Alex Greenland first spotted it on 23 June; through July it showed up in organic listings and in ads; on 26 August Google confirmed it, describing it as a technical measure against abuse of its services.

The encoding cannot be undone locally, and Google refuses HEAD requests, so asking for the header is not an option: you need a full GET per link to learn where it goes. Derek Perkins, of Nozzle, measured it: resolving every link in a five-page ranking costs between 500 and 1,000 requests. That same day he measured coverage across several providers and found it close to 100%.

Seen as one piece, the pattern has been the same for a year. In September 2025 Google retired the num=100 parameter, which let you pull a hundred results in one request, and overnight you had to make ten where one had been enough. In December it sued SerpApi. Now it routes its own links. None of the three moves touches your rankings. All three touch what it costs to know what they are.

A real comparison across 693 keywords: rank tracker vs Search Console

If rank tracking gets more expensive or less precise, the obvious plan B is Search Console, so the key question is not whether it works, but how much the answer changes when you switch sources.

I crossed both sources over the same portfolio: 693 keywords across 16 domains, all on Google Spain, in the week of 8 to 14 September. From the rank tracker I take the median of that week's daily positions. From Search Console, the average position for the same query, filtered to Spain, over the same window.

Worth saying before showing anything: they are not the same metric. The tracker queries a clean, unpersonalised SERP from a fixed location. Search Console averages real impressions from mobile and desktop, across all of Spain, personalisation included. Neither is the true position. I am not measuring which one is right, I am measuring how much your report moves when you swap one for the other.

For the rough question, the "roughly where am I" one, Search Console holds up: half the keywords land within two positions and three quarters within three.

The problem is with the question that actually ends up in a report. One in nine keywords changes its top 10 answer depending on who you ask. In 33 of the 402 the tracker puts you inside and Search Console leaves you out; in 13 it is the other way round. If you switch sources mid-project, that jump will look like a ranking movement, and it is not.

There is also a bias with a fixed direction: Search Console reports a worse position than the tracker in 72.4% of cases, with a median of seven tenths. It makes sense, because it averages impressions the tracker never sees, but it is worth knowing before you explain to anyone why "we have dropped".

The location bias: where local tracking really breaks

Looking at the region-specific data, two of the sixteen domains track with the location pinned to Barcelona. There the median difference doubles, from 1.1 to 2.0 positions, and large disagreements quadruple: from 1.6% to 6.4%. That is expected, because the tracker asks for a Barcelona SERP and Search Console averages the whole country, but I did not expect the size of it.

What does surprise me is that agreement on the top 10 stays the same in both cases, at 88.5%. In other words: locally you are not wrong more often, you are wrong harder.

Blind spots: what Search Console will never give you

Across the whole sample, 85 of the 693 keywords sit in the top 100 according to the tracker and have not a single impression in Search Console. It sounds like a Google blind spot until you look at whose they are: all 85 come from the same project, and they are very long queries along the lines of "best women's running shoe shops madrid".

So it is not that Search Console is blind. It is that there are no searches behind those keywords to measure. That is a warning about how we build tracking lists, not about the tool. But it leaves a practical consequence: if you track prompt-shaped queries for your GEO strategy, Search Console will not be your plan B for those, because it only reports what people actually search.

It happens the other way round too, though less: in 33 keywords Search Console records impressions and the tracker cannot find the domain in the top 100. One query every couple of weeks against a seven-day average.

My recommendation: calibrate your sources before the tracker fails

Despite these technical obstacles, I do not think rank tracking is going away. The tools will adapt, as they adapted to num=100. What I do think is that it will cost more, and that the cost will arrive as less frequency, less depth or fewer keywords.

So what I would do this week is the exercise I have just done: export your positions from both sources over the same keywords and keep the difference. While both are still alive, that table is free. The day one of them falls over you can no longer build it, and then you will not be able to tell a source change from a real drop.

It is the same idea I left written in August, only now I have the number: one in nine. If your portfolio looks like mine, that is the size of the error you will inherit the day you switch sources without having calibrated first.

Google is testing something that, if it rolls out, changes how we measure. Instead of the usual captcha, some users are running into a different message after moving past the first pages of results: "Sign in to verify you are human and see more results". For now it is a limited test. It may come to nothing. But it is worth understanding what it would imply.

What impact would a login wall past page 2 of Google have?

The impact is not where most people would look for it. It is not a ranking problem. It is a data access problem. The entire rank tracking ecosystem, every SERP API, every visibility study published in the industry, is built on being able to query Google results without being authenticated. If Google puts a login wall past page 2, that access gets substantially harder.

And this is where the pattern of recent months shows. Google already took queries away in Search Console when the AI reports arrived. It gave us an AI performance report that only shows impressions, with no clicks and no CTR. And now it is testing restricting access to the results themselves. Every move goes in the same direction: less data available to anyone who wants to measure independently.

My take on depending on a single external source to measure rankings

My criterion, and we have been working on this with Mencoro for months: if your measurement depends 100% on a single external source that can close from one day to the next, you do not have measurement. You have a dependency.

What I would do this week is not panic over a test that may come to nothing. It is to take inventory: which decisions you make with third-party data that scrapes Google, and what you would do tomorrow if that access disappeared. The answer to that question is worth more than the test itself.