Google confirmed on 26 August 2026 that it is rolling out google.com/goto passthrough links inside its search results. The anchor behind an organic listing used to carry the destination address. It now carries an encoded wrapper that issues a 302 to the real page.
A Google spokesperson told Search Engine Roundtable: "We have a long history of deploying technical measures against evolving forms of abuse, and we regularly take steps to protect our services and users." The statement names abuse. It does not name scraping, rank trackers, or AI companies, and it arrived with no timetable and no technical documentation.
This is no longer a test. Derek Perkins of the rank tracking firm Nozzle reported near total coverage across several residential IP providers on the same day, after roughly four months of intermittent testing. Alex Greenland logged the first public sighting on 23 June. Brodie Clark documented it across both paid and organic listings on 2 July.
The cost is the story
For most of Google's history, a tool could read a results page and parse each destination address out of it without ever contacting the destination site. That assumption sits underneath rank trackers, share of voice dashboards, competitive intelligence platforms, and a long tail of scripts that marketing teams run in house.
The wrapper removes it. Perkins says the links cannot be decoded locally, so a tool has to follow every redirect to learn where it goes. Two obvious workarounds died in the same thread. HEAD requests, which would retrieve the redirect target cheaply without the body, are not allowed through. A GET that stops at the redirect does work, but volume is the constraint: resolving all the links for one five page ranking takes between 500 and 1,000 requests. Rate limiting becomes the binding limit, rather than bandwidth or storage.
One question decides how bad this gets, and nobody answered it. Dominic Woodman asked whether the redirect is static per page. If it is, a large operator builds a lookup table once and reuses it. If it is not, this is a permanent per query tax on every tool that reads a results page.
The second problem is worse than the cost
Joe Manna asked whether resolving a goto link registers as a click. Perkins did not know, and said it would be interesting to see whether Google correlates those events with client side analytics.
Sit with that for a second. If resolution registers server side, the tool measuring your visibility becomes a source of the traffic it is measuring. And the same wrapper doubles as a detection surface for automated clients, separate from its job as a link.
What this does not touch
The manual test. Asking ChatGPT, Perplexity, or Google the question a real customer would type, then reading what comes back, never depended on parsing a results page. We published the five question version of that test well before any of this, and the argument there was that a paid monitoring tool automates the same test rather than replacing it.
That was a claim about signal quality. It has quietly become a claim about whether the automated version still runs reliably at all.
None of this makes rank tracking useless. Vendors adapt, and they always have. But if a dashboard is the only evidence you hold that your site is being found, it is worth knowing three things: the pipe feeding it just narrowed, nobody has said how permanently, and your vendor cannot tell you either yet.
Sources: Search Engine Roundtable and PPC Land.
Top comments (0)